清末民初老照片

合集下载

博物馆临时陈列布展的实践探讨以“紫禁藏影———故宫博物院藏老照片展”为例

□于沁博物馆临时陈列布展的实践探讨摘要：基于山西省人民政府与故宫博物院签订的合作框架协议，太原市博物馆与故宫博物院合作在太原市博物馆办展，举办了“紫禁风华———2018太原·故宫文物展”。

本文以其中展览之一的“紫禁藏影———故宫博物院藏老照片展”为例对展览进行总结和探究，包括设计到布展的全过程。

关键词：博物馆老照片陈列设计形式设计太原市博物馆是太原市的地标性建筑之一，位于长风文化商务区。

博物馆总建筑面积64400平方米，其中博物馆主体建筑面积52740平方米，平台建筑面积11660平方米。

本次展览以“紫禁风华———2018太原·故宫文物展”为主题。

共设七大展区：《普天同庆———清代万寿盛典展》《凝华焕彩———故宫博物院藏宫廷珐琅器精品展》《笔走龙蛇———傅山及明末清初名家行草书特展》《日丽紫禁———明清官式建筑展》《紫禁藏影———故宫博物院藏老照片展》《御瓷撷英———明代御窑瓷器展》《数字故宫———万象紫禁》。

此次展览综合运用实物、文字、图片、雕塑和现代信息技术，生动、立体、多方位地展示了其独特的历史文化和深厚的人文底蕴。

太原市博物馆于2018年10月1日开放运行，自开馆以来，以其独特的陈列方式，简洁大气的陈展空间，舒适宜人的参观环境吸引了众多国内外游客，成为了太原市又一处旅游热点（图一）。

为了追求展览内容的准确性、科学性、严谨性以及展览设计与制作的艺术性，太原市博物馆实行展项责任制。

我有幸成为了“紫禁藏影———故宫博物院藏老照片展”的展项负责人，通过全程参与，更加全面了解一个展览从开始筹备到完美的展出所要做的准备工作和付出的努力。

一、布展内容设计19世纪40年代，随着摄影术的传入，使昔日的皇家禁地第一次以影像方式进入人们视线。

摄于清末民初的老照片，忠实地记录了皇家建筑的恢宏与高敞，也承载着一个历史时代的记忆与沧桑。

20世纪初，人物影像作为老照片中最为重要的一类，上至达官显要，下至普罗众生，不仅体现出百年前社会各阶层的人物百态，而且透过他们的衣着、神态、人物关系及拍摄场景，亦可再现晚清至近代社会的生活原貌、政治活动及历史事件等。

一幅清末民初保定社会生活的历史图卷——记王宗玺拍摄的一组老照片

科技含量而论，以河北乾义面粉公司和
保定电灯有限公司这两大民营企业为
图一民国初年的保定火车站
最。这两家企业的筹建和发展，都与当时保定商会会长樊榕有密切关系。樊榕（８１９２年）字荫荪，１６ —１４，号退安，河北清苑人。他是光绪癸巳（８３１９年）举人，山西知县。民国后秉承实业授
隶 ”１［４。
在樊达汇、樊榕父子两代人的努力
之下，（） “ 樊家大饶裕，遂创设商会于保
定，广建学堂，与开四民之智，国家利凡
图四保定城外运河一景
用厚生诸事，出巨资助之成立”。屡樊达
一
７ —
维普资讯
王宗玺（８０— ９３年）字尊信，１９１４，辽宁海城人，清光绪朝兵部右侍郎王英楷之子。王
英楷，海城县志》《有传，家住当时的保定城关后卫街¨ Ｉ。王宗玺１１年出任北洋政府航空９６署总工程师，是中国第一个设计三引擎飞机
华兵工厂帮办，是一位军事工程技术专家【２１。由于工作需要，他又兼工摄影，这些照片就是
地，是北方政治、军事、经济、文化中心，也是
南北交通的枢纽。袁世凯主政以后，为筹集军
资、练兵，不断扩大保定火车站的规模，使火车站成为各种军事物资的集散地。照片反映
余张，内容涉及清末民初河北保定社会生活
的方方面面，包括火车站、城外运河码头、城
的专家；９５，１２年应孙传方邀请，出任上海龙六十万，口密集，人市场也开始繁荣起来。码
他在工作之余拍摄的。下面挑选其中有代表性的介绍如下，供河北学术界研究参考。

IDEO_Expertise

BUSINESS DESIGNLearning the business of baby formula prep for Yoomi.Whenever a company designs a new product, service, or experience, it is essentially designing its business. When done well, Business Design creates offerings that inspire organizations and excite customers. At IDEO, we combine design thinking and traditional corporate strategies to help clients create avenues for market growth. By shifting focus from linear practices to iterative design processes, we can shed light on new options and explore the various alternatives. Our methods include qualitative and quantitative research, business model prototyping, data visualization, organizational design, and IP liberation.Over the past five years, we've used Business Design to help clients address a broad range of challenges, such as: designing the next-generation news offering for a major publishing company; creating a personal finance service built on lifestyle and community; building new, sustainable growth model for an online education company; launching an online employment service designed for the millennial workforce; redesigning a direct sales business for emerging markets; and developing an OTC retail business for a pharmaceuticals manufacturer.We believe that every part of a business can be prototyped, including projected financials. We help clients design all the way to market, iterating potential solutions from initial concept through final consumer delivery. We look at innovation across industries to find inspiration and viable new patterns of thinking and operatingNew Venture DesignAs organic growth becomes increasingly difficult, businesses must find new means of attracting customers and driving growth. We help companies identify how their organization can do so through design. By rethinking business models and developing innovative offerings and supporting strategies, we support clients as they experiment with new categories, channels, or business models to fuel sustainable growth.Growth and Innovation StrategyInnovation plays on the strengths and the culture of a company, so its approach and alignment can be vital to its success. Our Business and Organizational Design teams work closely with companies to develop a strategy for innovation and draw roadmaps for future growth. We assist business leaders in figuring out what innovation means for their organizations, how will it impact their business, and what it will take to engage and mobilize their organizations.FEED ME BOTTLESHELPING AN INNOVATIVE U.K. START-UP BRING ITS AWARD-WINNING BABY-BOTTLE CONCEPT TO MARKETProject Date: 2009Farah and Jim Shaikh founded FeedMe after their own experience with heating up baby bottle feeds - from stumbling around the kitchen in the wee hours to begging restaurant waiters for hot water - inspired them to find a more pleasant, efficient way to do so. They conceived a self-warming device that would warm formula or breast milk to the ideal temperature in just one minute. Their early prototype, designed with IDEO, won the 2008 Design and Engineering category of the London Technology Fund's Competition and a 2007 Department of Trade and Industry Research Award.The U.K.-based entrepreneurs approached IDEO London for help in realizing a final product and creating a holistic and aligned brand, marketing, and business strategy to promote its market adoption. Working closely with the Shaikhs, IDEO set out to understand the needs of new parents and to determine how they might be persuaded to use an unfamiliar, albeit innovative, technology. Through research, the design team discovered that emotionally driven thinking is sometimes illogical: Mothers want the bottles they purchase for their babies to seem natural, yet look trendy with benefits focused on baby rather than their own convenience, so they don't feel guilty. These and other motivations were synthesized into a series of principles that informed the design stage.IDEO, the Shaikhs, and their engineering team then developed a series of bottle and branding typologies for assessment by mothers and fathers-and subsequently narrowed these down to three options that were fully prototyped. Each offered less fuss than traditional bottles (particularly for caregivers other than moms), was easy to use and aesthetically pleasing, and relied on existing market language (to downplay the "new technology" angle and provide reassurance to users). The final product and branding approach was chosen after the team gathered additional input from new parents.The resulting 5-ounce Yoomi bottle works on a similar principle to hand-warming gel packs. Its rechargeable warmer heats milk to 32-34 degrees Celsius in 60 seconds-and makes it easy to combine breast and formula feeding. "The button activates the process that turns the non-toxic solution from liquid to solid, during which it gives off heat. It can be recharged after use by immersion in boiling water or a steam sterilizer," Jim Shaikh explained to the U.K.'s Daily Mail. The bottle features a naturally shaped teat, an ergonomic design, and materials that are 100% safe and BPA-free.(continued on next page)Yoomi creates an entirely new product category by integrating bottles and warming stations into a single solution. The team developed a business strategy for Yoomi's pricing, promotion, and launch after conducting a full market analysis (in the U.K. and the U.S.) and a series of workshops with the client.IDEO's brand and product design clearly show the product as a complement to breastfeeding (or as a natural alternative to breastfeeding for those that can't), making the bottle a sensible choice prior to birth as it keeps mothers' options open. Mothers tend to be highly loyal to products that they use early so it is envisaged that they will continue to use the Yoomi suite of products until their child no longer drinks from a bottle, hence the forthcoming launch of a 9oz bottle.Yoomi, launched in November 2009, is initially being offered for £22 (plus £4.50 for two teats) exclusively through John Lewis, a leading U.K. retailer. Sales volumes so far have been ahead of target, and an international launch is forthcoming. Clare Byam-Cook, a breast-feeding counselor and former nurse, says that Yoomi's safety and convenience is bound to make life easier for parents. She told the Daily Mail that "the great thing about Yoomi is that mothers can put breast or formula milk into the bottle and know that it will be heated to the perfect temperature that is safe for their baby."SERVICE DESIGN, IDENTITY AND MARKET STRATEGYHELPING THE EMERGING WORKFORCE AND EMPLOYERS SHARE WHAT THEYHAVE TO OFFER AND FIND THE RIGHT OPPORTUNITIESProject Date: 2009The US job market can be difficult to crack, especially for media-savvy 20-somethings whose work experience is limited and whose skills are a hard sell on a traditional resume. A pair of young entrepreneurs saw an opportunity and set out to provide their peers with a reliable online source for career advice, job listings, and networking opportunities. Problem was, they weren't really sure how to go about it. A start-up investor strongly recommended IDEO, and the project was born.The first order of business was to re-evaluate KODA's business plan and general concept. Originally, the founders envisioned a "Monster meets MySpace" site for 18- to 24-year-olds looking to establish their personal brands. Because understanding Generation Y and potential employers would be crucial to KODA's success, the creative team interviewed a diverse group of individuals seeking their first job or internship, as well as various recruiters representing companies of all sizes. The research produced ample data about young workers and the corporate stereotypes that cause many recruiters to resist hiring them. It also suggested that a website which helped frame the experiences of young talent and balanced the interests of applicants and recruiters was relevant and needed. Next, working closely with KODA, IDEO's creative team conceived and designed the new company's service offering and market strategy. It also provided KODA with a business plan and model, visual identity, and communication materials. IDEO and KODA partnered with software development company Pivotal Labs for the site's technology build out. The end result: the first site to enable social recruiting, creating space and context for applicants and companies to share information about themselves.Unlike most job-search sites, which are purely transactional (i.e., companies post positions they need to fill, candidates search and apply), lets job seekers and recruiters co-exist in a transparent space that enables self-selection. The site promotes this transparency by asking for information not traditionally included on a resume and by giving both parties a "safe" space in which to discuss who they are and what they have to offer.Other key features include:- A clean user interface that's easier on the eyes than other job boards. This is achieved through a concise visual identity and an (continued on next page)advertising-free business model (employers buy service subscriptions).- A compatibility engine that gives candidates and companies suggestions for potential fits based on elements like company size, work style, and work environment in addition to the usual filters, such as location and industry.- A back-end management system that allows recruiters to organize candidates, track progress, send automatic email alerts to candidates about the progress of their applications, etc.- Crawler technology makes maintenance simple for companies; it links into the corporate careers site and pulls job posts into KODA to ensure opportunities are current and correct.A beta version of launched in May 2009 and focuses on the San Francisco Bay Area and New Orleans. In August, its founders (who began KODA with $2 million) received $1 million in funding from a private angel investor, and the site is generating significant buzz. "There are hundreds of companies using KODA already," wrote Alison Doyle, a job-search expert at , "it's an interesting mix of social and professional networking. already offers thousands of job leads in various cities around the country." was honored as a 2009 Innovator of the Year by New Orleans CityBusiness, which recognizes organizations that keep the city a center of creativity, operate in new ways, and develop unique products.HACKFWDAN EARL Y-STAGE INVESTMENT COMPANY DESIGNED TO FUEL TECH INNOVATIONIN EUROPEProject Date: 2010Lars Hinrichs is an Internet icon. XING, the company he co-founded in 2003, was the first Web 2.0 business to go public in Europe, and it now ranks among the world's most popular social-networking sites for professionals. In 2008, the Startups Initiative named Hinrichs as the most important Web entrepreneur in Germany. In 2009, the World Economic Forum recognized him as a Young Global Leader. The same year Hinrichs - who calls himself a "serial entrepreneur" - stepped down as CEO of XING to pursue new projects. He came up with the idea for an early-stage, pre-seed, evergreen (no fixed funding cycle) investment company for top technology talent across Europe. His vision: free the best developers and coders from their day jobs and help them build their own game-changing companies.Hinrichs asked IDEO to help turn his vision into reality. Together, they set out to design the business from scratch. IDEO immediately assigned a European-led global team to the project, which began with interviews of developers and coders throughout the U.S. and Europe, to understand what they wanted and needed most. Their input provided valuable insights into how to build a truly revolutionary company. Working with Hinrichs, IDEO designed the new company's business model, financial model, organizational structure, hiring strategy, tools to vet and build investments, brand strategy and identity, and events to cultivate a strong sense of community, encourage sharing and collaboration, and provide inspiration.The resulting company, HackFwd, launched on June 8, 2010, with a series of high profile European tech investors on board. The company accelerates the route to beta, profitability, and success by enabling individuals to focus on what they do best and providing support on tasks that do not drive actual value. "Putting geeks' needs ﬁrst informs everything we do - from our brand and how we organize ourselves to how equity is shared and what tools we offer," Hinrichs says.How it works: Developers, coders, and engineers submit business ideas through a referral network, which makes selections based on an individual's or team's passion and on whether the idea meets a clear consumer need. Ideas also must be pioneering, scalable, and ready to beta test within six months. If chosen, HackFwd roughly matches current salary of those behind the idea and brings them into a network of highly experienced digital entrepreneurs. These advisers review and offer input over a 12-month period. During that time, the individual or team launches a beta and receives critical user input to help determine how best to move the business(continued on next page)forward.HackFwd handles most of the administrative load, including legal set-up, payroll, and accounting, and helps each individual or team find talent when they are ready to scale up. This allows the "geeks" to focus on what they do best. In exchange for the financial support, creative and strategic advice, and the help of promoting and marketing the product, HackFwd takes 27 percent of the equity, the founders keep 70 percent, and advisers who assisted the team receive 3 percent in options. HackFwd already has four investments and plans to make a six more in 2010. The company expects to scale up its investment rate in 2011.With HackFwd, Hinrichs aims to revolutionize the European technology scene and put Europe on the global innovation map.See for further information.BRAND STRATEGYEXTENDING A FOOTWEAR ICON'S FASHION SENSE AND NATIONAL PRIDETO ACCESSORIESProject Date: 2009The Havaianas brand embodies what it means to be Brazilian: free-spirited, with a zest for life and an instinct for improbable combinations. The company first designed inexpensive cloth shoes for Brazilian coffee farmers in 1907; today it offers nearly 20 product lines worldwide and sells 162 million pairs of flip-flops a year. Scan the pathways and streets the world over, and you'll find Havaianas' iconic footwear on everyone from celebrities like Gisele Bundchen and Beyonce Knowles to everyday consumers who appreciate affordable, versatile shoes.With 60 years of consistent growth, Havaianas came to IDEO to help extend its brand into a new product category. The company wanted to offer a line of tote bags that shared the simple pleasure and boldness of its flip-flops, starting with ways in which people could carry their beloved Havaianas around.To imagine what these Havaiana accessories might look like, IDEO first sought to capture the essence of the brand by understanding its tight connection to Brazil's national identity. The team interviewed citizens from various demographic and socioeconomic backgrounds and found that they all had vibrant spirits and an appreciation for simplicity, authenticity, play, enjoyment, and physical comfort. (The team conducted additional research in Australia, India, London, and Paris.)The team also ran a series of "handstorms," or bag-making brainstorm sessions that led to hundreds of prototypes. Each was considered in relation to the brand vision for Havaianas. Working with a seamstress, the team tried various shapes and materials. In the end, they developed a consistent design language that echoed the look and feel of Havaianas flip-flops with textural and visual cues.The resulting line of modular bags was designed to meet the needs of Havaianas customers and its retailers. The bags' style is relaxed, individualistic, and uncomplicated. The colors, based on a palette curated from photographs taken throughout Brazil, capture the fun and the vibrancy of the country. The commercial rollout of the bags put great emphasis on staying true to Havaianas's Brazilian heritage. By introducing the bag in Brazil, Havaianas allowed its flip-flop fans at home to be the first to enjoy the product. Havaianas unveiled the bags at Sao Paulo's fashion week in June 2008. The totes rolled out elsewhere in 2009.GE CAULK SINGLESNEW WAYS TO FILL IN THE GAPS FOR AN EXPANDING MARKETProject Date: 2008In the US, do-it-yourselfers (DIY) spend 300 billion dollars a year looking for the next convenient and innovative home improvement solution. Recognizing an opportunity in the caulk market, Momentive Performance Materials (formerly GE Sealants and Adhesives) asked IDEO to explore new caulking experiences for professionals and DIYers alike. To capture the attention and imagination of these consumers, the program took on a holistic scope, entailing engineering, manufacturing, brand development, packaging, and communication design as way to reinvigorate the caulk market and pioneer new user experiences.The Momentive and IDEO team performed extensive user research in order to understand consumer needs and behaviors. Major opportunities were revealed to simplify the entire caulking experience, including selecting the appropriate product in store, learning how to correctly apply the product, and cleaning up after use. Iterative prototyping and user studies informed the final packaging design, style guides, and initial graphic package for the entire silicone and acrylic lines. A single-use caulk packet was at the center of the program, serving as the launching point for a new and differentiated brand direction.The resulting product is GE Caulk Singles, a one-time use packet that consists of a one-inch structural plastic sheet adhered to a vacuum-sealed foil pouch. To produce a round and consistent bead throughout application, IDEO engineers incorporated a cardboard backing to provide uniform and constant pressure to the pouch during use. Users simply tear, squeeze, and toss, using a one-handed application without the need for tools. The individual caulk doses provide the perfect amount for most common household jobs, minimizing waste and storage and making clean-up quick and efficient. Communication and graphic elements help simplify the interaction with intuitive, consistent, and direct messaging.The GE Caulk Singles launched in June 2008, and are available at Lowe's, Ace Hardware, True Value and Do-It-Best.CLEANWELL BUSINESS PLATFORMUSING DESIGN THINKING TO BUILD A BUSINESSProject Date: 2007In a culture that is at once increasingly vigilant about germs and the value of natural ingredients, an herb-based antimicrobial that is lab-proven to kill 99.9 percent of harmful germs would fulfill unmet user needs.Working with Dr. Larry Weiss, a scientist and physician with expertise in natural products, chemistry, and infection control, and a set of strategic capital partners, IDEO helped found the CleanWell brand of Ingenium products. In addition to helping to build CleanWell's operational and manufacturing capabilities, IDEO designed the packaging and graphics for the brand's line of sprays and skin towelettes.Ingenium, the first all-natural antimicrobial that meets FDA and EPA standards for germ-killing efficacy, is the effective ingredient in CleanWell hand sanitizer and hand wash, which is now available at Whole Foods Markets and select independent pharmacy, grocery, and specialty retailers across the United States. As a non-toxic, alcohol-free alternative to competing hand sanitizers, CleanWell stands out as a school- and child-friendly solution.BRAND ESSENCEDEFINING A BRAND ESSENCE TO INSPIRE AND GUIDE A COMPANY CUL TUREProject Date: 2007QVC sells merchandise twenty-four hours a day through its television and web channels, offering 1,700 items a week via TV alone. During its twenty-year history, QVC has shipped more than 1 billion packages to more than 30 million customers. After two decades of double-digit growth, QVC began to recognize newfound challenges and opportunities in the cable industry and in consumers' viewing habits. With the desire to ensure continued growth by becoming an even more compelling shopping destination, QVC approached IDEO to develop a clear and concise brand and offering.Working closely with several members of QVC's senior management team, IDEO focused its efforts on the internal and external aspects of QVC's brand to gain insight into potential growth opportunities. IDEO looked internally to understand the company's operations, heritage, culture, and its employees and their key values in order to arrive at a brand that rang true. IDEO also sought to understand how QVC's brand and offering were perceived among a broad cross-section of shoppers. This was particularly important to QVC's growth strategy, which requires balancing the needs of existing customers with those of potential new customers.The resulting brand essence brought clarity to QVC's management and merchandising teams and served as a catalyst for innovation within the company. Using IDEO's inspiration, QVC has since embarked on numerous design-inspired projects, including an ambitious plan for upgrading every aspect of the live broadcast, including sets, graphics, and host image. The newly articulated brand essence will also serve as a guiding principle for a redesigned website and graphic identity.According to Doug Rose, QVC's VP of merchandising brand development, "As a result of this work, we have a clear strategy and common language that the entire company can rally around."LIBERATING INTELLECTUAL PROPERTYDEVELOPING COMMERCIALL Y VIABLE PRODUCTS FOR A DRY, EASY-TO-USEANTI-FRICTION MATERIALProject Date: 2007GE knew its Advanced Materials division was on to something when a team of in-house chemists developed a boron-nitride-based powder that was surprisingly nice to the touch. But the compound, though clearly useful in the semiconductor processing industry and as an additive in the cosmetics industry, lacked a broader place in the marketplace. To move beyond having applications in two narrow niches, GE asked IDEO to look for new commercial applications for the familiar material.The substance is an industrial compound with unique properties-it's dry, exceptionally lubricious and odorless, clean, and easy to use. During the summer of 2006, IDEO studied the powder's attributes and the needs of people faced with all sorts offriction-plagued tasks, from DIY enthusiasts and carpenters to runners and masseuses. IDEO held structured brainstorms and asked how might the material's technical value be combined with humanistic attributes and moved, ultimately, into new markets.The brainstorming sessions generated a list of applications, and from there, the team identified and observed people associated with these applications. IDEO engineers, material scientists, and designers spoke with and watched people in a variety of fields. They also studied extreme users, including an ultra-distance marathoner who went on 100-mile runs. The team compared the existing solutions these people used with the boron-nitride option, and developed concepts using the boron-nitride solutions that better served their needs. Among the deliverables, IDEO supplied GE's Advanced Materials division with four so-called traveling salesmankits-hands-on prototypes that showed the most-promising potential products, partners, and strategies for reaching key markets.In September 2006, GE sold its Advanced Materials division, and two years later, Momentive Performance Materials teamed with ASICS to offer a new line of anti-friction products for athletes and runners, based on the boron-nitride-based substance (trademarked by Momentive as Aceba). Developed with the performance athlete in mind, ASICS' Chafe Free gels and powder (created using Aceba) reduce skin-to-skin and skin-to-clothing friction, helping improve performance and enhance the enjoyment of sport and physical activity."Conventional anti-friction products, including petroleum jelly and other alternatives, have tended to be greasy, messy, and difficult to apply," said Monica Ettamarna, Momentive's marketing manager for consumer products. "By contrast, our Aceba dry lubricant(continued on next page)allows Chafe Free to provide high, long-lasting lubricity and an exceptionally dry, non-sticky feel, coupled with easy, clean application."Within the three-day lead up to the New York City Marathon, ASICS sold 2,500 tubes of Chafe Free, the company's complete inventory for the event, and numerous athletes used the product in the race.As a specialty materials company, Momentive provides high-technology materials solutions to the silicones, quartz, and ceramics markets. The company, with worldwide operations, is continuing to look for additional partners for marketing other uses of the boron-nitride-based material.Founded in 1991, IDEO* is an innovation and designfirm that uses a human-centered, design-based approachto help organizations in the business, government, education, and social sectors innovate andgrow in three ways:Identify new ways to serve and support people byuncovering their latent needs, behaviors, and desires.Visualize new directions for companies and brands anddesign the offerings - products, services, spaces, media,and software - that bring innovation strategy to life.Enable organizations to change their cultures and build the capabilities required to sustain innovation.* Pronounced "eye-dee-oh"HIGHLIGHTS & ACCOLADESRanked #10 on Fast Company's list of the Top 25 Most Innovative Companies (2009)Ranked as one of the most innovative companies in the world by Boston Consulting Group (2005 - 2007, Business Week) Ranked #15 on Fortune's list of 100 most-favored employers by MBA students (2009, compiled by research firm Universum) Awarded the Smithsonian Cooper-Hewitt National Design Museum's National Design Award for Product Design (2001) Winner of more IDEA awards than any other design firm20 Red Dot awards; 19 iF Hanover awardsFeatured in a 1999 episode of ABC's Nightline, which followed an IDEO team as they redesigned the shopping cart in four days.PEOPLE550+ employees in thefollowing disciplines:BrandingBusiness designCommunication designElectrical engineeringEnvironments designFood scienceHealthcareHuman factorsIndustrial designInteraction designMechanical engineeringOrganizational designSoftware engineeringCONTACTSNew Business:---USAJohn Ravitchjravitch@415 615 5054EuropeSabine V oeglersabine@0049 89 3834 6754AsiaTina Chentchen@86 21 6132 5500Press Inquiries---Lindsey Zoueinlzouein@650 289 3582LOCATIONSSan FranciscoPalo AltoChicagoBostonNew York CityLondonMunichShanghaiEXPERTISE AREASBrandBusiness designDigital experiencesEducationEnergyEngineeringFinancial servicesFood & beverageHealthMobilityPlayPublic sectorSocial innovationOrganizational designCLIENTSIncluding:Acumen FundAT&T MobilityBank of AmericaBBV ACenters for Disease Controland PreventionConAgraEileen FisherEli LillyFord Motor CompanyGE Money BankThe Bill and MelindaGates FoundationGlaxoSmithKlineHavaianasHyundaiCardIntercellKaiser PermananteMarriottMayo ClinicMicrosoftNestleOxfam GBNokia(RED)Rockefeller FoundationSamsungSesame WorkshopTargetUS Department of Energy。

清末民初的文化与教育

新的教育制度建立，它规定初等教育、中等教育、高等教育等几个阶段。此后，全国各地逐渐建立起各级新式学堂。今天的教育的基本制度就是那时建立的。
陈旧的科举考试制度的大门关闭
二、废科举，兴学堂
《奏定学堂章程》（1904）——中国近代第一个以教育法令公布并在全国实行新的学制。
1905年，清政府宣布废止科举考试。
响最大的一份报纸。原全称《申江新报》，它从1872年4月30日创刊，直到1949年5月27日停刊，前后办了 77年，共出版25600号。初由英商安纳斯·美查（Ernest Major）等人集资创办。1909年为买办席裕福收买， 1912年转让给史量才，次年由史接办，先后创办《自由谈》等副刊，发表民主自由言论。1934年史量才被蒋介石派遣特务暗杀后，言论又趋保守。1937年7月改组为股份有限公司。同年12月15日，因日军检查新闻，自动停刊。抗战期间，曾在
中国最著名的综合性大学。前身是京师大学堂，创办于1898年，1912年改名为北京大学。
北大校牌
北大校牌上“北京大学”四字系毛泽东手书
北大校徽
北京大学的校徽是鲁迅先生设计的，是“北大”两个篆字
的上下排列，其中“北字”构成背对背的两个侧立的人像，而
“大字”构成了一个正面站立的人像。校徽突出一个办学理念，
即ห้องสมุดไป่ตู้
北
大
要“以人为本”。
北大老照片(1)
新北大
理科楼
图书馆
食堂与宿舍
静园
博雅塔
未名湖
计算机房
留学生活
读书
首脑来访
北大课堂
雕塑与故居
二、废科举、兴学堂
1、科举制的废除 2、新式学堂兴起

清末民初的文化与教育PPT课件6

感谢聆听，再见！
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1、成功呈概率分布，关键是你能不能坚持到成功开始呈现的那一刻。 2、得不到的东西永远总是最好的，失去的恋情总是让人难忘的，失去的人永远是刻骨铭心的。 3、后悔是一种耗费精神的情绪，后悔是比损失更大的损失，比错误更大的错误，所以不要后悔。 4、生命对某些人来说是美丽的，这些人的一生都为某个目标而奋斗。 5、生气是拿别人做错的事来惩罚自己。 6、如果我们想要更多的玫瑰花，就必须种植更多的玫瑰树。 7、做自己就可以了，何必在乎别人的看法。 8、相信所有的汗水与眼泪，最后会化成一篇山花烂漫。 9、忘掉失败，不过要牢记失败中的教训。 10、如果敌人让你生气，那说明你还没有胜他的把握。 11、一百次心动不如一次行动。 12、天下之事常成于困约，而败于奢靡。 13、人生短短数十载，最要紧是证明自己，不是讨好他人。 14、世上并没有用来鼓励工作努力的赏赐，所有的赏赐都只是被用来奖励工作成果的。 15、只要我们能梦想的，我们就能实现。 16、只要站起来比倒下去多一次就是成功。 17、诚心诚意，诚字的另一半就是成功。 18、我终于累了，好累，好累，于是我便爱上了寂静。 19、只有收获，才能检验耕耘的意义；只有贡献，方可衡量人生的价值。 20、赚钱之道很多，但是找不到赚钱的种子，便成不了事业家。 21、追求让人充实，分享让人快乐。 22、世界上那些最容易的事情中，拖延时间最不费力。 23、上帝助自助者。 24、凡事要三思，但比三思更重要的是三思而行。 25、如果你希望成功，以恒心为良友，以经验为参谋，以小心为兄弟，以希望为哨兵。 26、没有退路的时候，正是潜力发挥最大的时候。 27、没有糟糕的事情，只有糟糕的心情。 28、不为外撼，不以物移，而后可以任天下之大事。 29、打开你的手机，收到我的祝福，忘掉所有烦恼，你会幸福每秒，对着镜子笑笑，从此开心到老，想想明天美好，相信自己最好。 30、不屈不挠的奋斗是取得胜利的唯一道路。 31、生活中若没有朋友，就像生活中没有阳光一样。 32、任何业绩的质变，都来自于量变的积累。 33、空想会想出很多绝妙的主意，但却办不成任何事情。 34、不大可能的事也许今天实现，根本不可能的事也许明天会实现。 35、再长的路，一步步也能走完，再短的路，不迈开双脚也无法到达。 36、失败者任其失败，成功者创造成功。 37、世上没有绝望的处境，只有对处境绝望的人。 38、天助自助者，你要你就能。 39、我自信，故我成功；我行，我一定能行。 40、每个人都有潜在的能量，只是很容易：被习惯所掩盖，被时间所迷离，被惰性所消磨。

八年级历史清末民初的文化与教育课件

清末民初时期，一些知识分子认为教育可以救国，提出“教育救国”的口号，认为通过发展教育可以培养国民素质，提高国家竞争力。
爱国教育的兴起
在民族危机日益加深的背景下，爱国教育逐渐兴起，学校开始注重爱国主义教育和民族精神的培育，激发学生的爱国热情。
Part
04
清末民初的文化与教育影响
对现代中国的启示
辛亥革命后，民族资本主义得到一定的发展，但受到国内外多种因素的影响，发展并不顺利。
社会动荡
清朝末年，社会矛盾尖锐，农民起义和民族起义不断。
辛亥革命后，虽然建立了中华民国，但由于政治和经济的不稳定，社会动荡依然存在。
这一时期，各种思潮和运动兴起，如民主主义、科学主义、创新
01
02
03
绘画变革
清末民初时期，绘画艺术发生了重大变革，出现了 “岭南画派”、“海派” 等新的绘画流派。
音乐创作
音乐创作在清末民初也有了很大发展，出现了很多优秀的音乐家，如聂耳、冼星海等。
工艺美术
工艺美术在清末民初也有很大发展，出现了很多优秀的工艺品，如景泰蓝、紫砂壶等。
学术进步
女子教育的进步
女子学堂的创办
清末民初时期，一些开明人士开始认识到女子教育的重要性，创办了如上海女子学校、苏州景海女塾等女子学堂，开启了女子教育的先河。
女权运动的兴起
随着女子教育的发展，女权运动逐渐兴起，女性开始争取自己的权利和地位，进一步推动了女子教育的发展。
教育与救国思潮的结合
教育救国思想的提出
聚力和向心力。
03
推进教育改革
清末民初时期的教育改革为中国现代教育的发展奠定了基础。现代中国
应借鉴历史经验，深化教育体制改革，提高教育质量和水平，培养具有

Master-Trader

Contents Introduction: (4)Philosophy: (5)Trading strategies – an overview: (6)Introduction to direct access trading: (8)The US stock markets: (9)Bids and offers: (10)NASDAQ and level 2: (12)The New York Stock Exchange (NYSE): (14)NYSE stocks in the level 2 window: (15)NYSE stocks on “Island”: (16)The basics of Nasdaq order routing: (16)Short selling: (18)Basic rules for using technical analysis: (19)Market and sector analysis: (20)Types of charts: (22)Development of trends: (24)Moving averages: (26)Volume: (28)Breakouts: (29)The pivot setup: (31)Continuation patterns: (32)Moving average crossovers (36)Basic swing trading setups: (38)Flags and pennants: (40)Triangles: (42)The cup and handle: (44)Candlestick indicators: (45)Price resistance: (49)What makes stock prices move? (50)Price/Volume studies: (51)Momentum trading: (53)Gainers and dumpers: (54)Liquidity: (57)Spotting the “ax” on level 2: (59)Gaps and premarket trading: (60)Unusual prices: (63)Nasdaq order routing systems: (65)The Island ECN (ISLD): (67)Archipelago (ARCA): (68)Small order execution system (SOES): (69)Selectnet (SNET): (70)Instinet (INCA): (71)Trade Management: (72)Learning plan: (73)Paper trading: (75)Choosing brokers: (76)Commissions (77)Technical requirements/computer setup: (79)Graphics and multi monitor setup: (80)A typical trading day and pre market preparation: (83)Keys to success - psychological aspects: (88)Disclaimer: (95)Introduction:This book is designed to introduce you to the exciting world of active trading. Active trading means to actively participate in everyday price movements of the financial markets. Active trading enables you to actively manage risks and to participate from both rising and falling prices. The trades I am describing in this book can be from as short as a few seconds to as long as a few days. Many of the strategies can be applied to various timeframes. The difference between active traders and investors is that active traders trade the actual price movement versus investors who make their decisions based on the anticipation of future price movements. I tried to make this book as complete as possible. However, you will find as many strategies as traders. As you gain more experience you will realize that most strategies are based on the same basic principles and that there is really no holy grail out there.I have been trading and coaching for many years now. The need to be independent certainly was the biggest reason for me to enter the world of trading. In what other job do you have the freedom to work from anywhere in the world where you have access to the Internet? I started with investing but always felt that there has to be more to the stock market. That’s when I started watching quotes in real time and realized how big the profit potential must be if I could just cut out a small piece of the everyday movements. There are many obstacles to conquer though in order to get to a consistent success. A solid strategy, a neutral state of mind and rigid risk management are only some of the key traits needed to be successful.Whether you are planning to trade full time or just part time, this book will give you very valuable insight into the whole business. Even if you are just planning to invest you should read this book and take some of the basics of technical analysis into consideration when making your next decisions.Philosophy:Personally I don’t think trading needs to be complicated. Keeping it simple is the way to success. I have seen that with all of the worlds leading traders. They only use a few basic strategies in combination with simple tools and indicators.That does not mean trading is simple. There is great room for failure when it comes to staying neutral and to discipline.You don’t need to know everything. The key is to find a few solid strategies that work for YOU and master them. My goal is to help you on this search.I believe the most effective way to become successful as a trader is to learn directly from a pro who as already made his mistakes and been thru the struggle one faces when starting out.In my career as a coach I met many traders that were confused by all the tools they were given. Basically they had all the knowledge they needed, but no one told them how to apply itto real trading. This is why I started one-on-one coaching.For more information on coaching please see.Trading strategies – an overview:There are as many different trading strategies as there are traders. Generally they can be distinguished though by the time frame in which they take place. I suggest that every trader experiments with different strategies and then decide for himself what he is most comfortable with.A)Longer term strategies (from a day trader perspective)Investing: Investors buy shares of a certain company because they believe in its long-term growth perspective. They have little interest in most of the daily price movements and are looking to hold their shares for several years.Swingtrading:Swingtrading means to hold stocks anywhere from one to five days and sometimes more. Swingtraders try to take advantage of certain “key” situations in a stock price’s movement. Such a situation would be a buy after a pullback into solid support during a longer term uptrend. Swingtrading belongs to one of the easier to implement strategies and is excellent for people with small accounts.Overnight trading:B)Short term strategiesMomentum trading: A momentum trade usually lasts anywhere from 30 seconds to about 1 hour. Momentum trading is based on strong price movements and counter price movements often caused by news.Breakout trading: breakouts (breakdowns) do occur in any time frame.Popular charts for breakout traders are 5 minute and 15 minute charts. The holding period is anywhere from a few seconds (breakout scalp) up to the end of the day.Breakout trading means to buy stock after it has broken out above a certain price. Vice versa for shorts.Pullback trading: Pullback trading is the opposite of breakout trading. Pullback traders are looking for stock prices to pull back a significant enough amount (usually into support) in order for them to justify an entry (vice versa for shorts). Personally I am more of a breakout trader since I like the confirmation of the stock prices’ movement that I get thru the breakout; although pullback trading often has the smaller stops though. The holding period is usually a few seconds up to an hour.Scalping:Scalping describes “ultra short term” trading. Scalpers try to take advantage of very small price movements and sell their shares immediately when they have a big enough profit or the stock isn’t moving in their direction or goes against them.Cutting the spread: Cutting the spread can be seen as a scalping variety. Cutting the spread means to take advantage of the spread (the price difference between the bid and the ask price). It means to buy a stock on the bid side and to sell it immediately afterwards on the ask side for a small profit. Since the decimalization of the markets this type of trading has certainly become much more difficult because spreads have gotten much smaller, however I still see traders implementing this strategy pretty successfully.Please note that the strategies presented in this book are by no means the “holy grail”. Trading setups have to be monitored and adjusted continuously. I did try to cover all the major strategies though in order to give you a sound insight into how traders work.Introduction to direct access trading:Direct access trading has revolutionized trading in the late 90’s. Many traders are still not aware of the tremendous advantages it offers, especially for the active trader. Imagine being able to place an order with the push of one button and to get executed instantly. This is what direct access trading is all about.The traditional way to route orders was to call your broker, who would then send your order to his person on the exchange floor or to the market maker to actually execute your order. After that is done the whole process reverses in order confirm what happened with your order. If you are lucky this process will only take a few minutes, but in many cases it takes much longer. For some time now people have used online trading, which in most cases is not much different to the traditional way, with the exception that your order gets sent electronically to your broker who then processes it.With the introduction of direct access trading order execution has improved dramatically. You are now able to route your order directly to the exchange without any middlemen involved. Access to the market that was formerly only available to institutions is now available to everyone. You can decide which way your order is going to be routed and you can change or cancel it at any time in an instant.On your level 2 screen you can see all the competing bids and offers for any stock listed at the Nasdaq. Every market maker and every ECN is displayed in the level 2 window and you can directly trade with them. Think about how fast your voice travels over the phone? This is the speed you can use for routing your orders. It works solely electronically and there are no middleman involved.There are different order routes integrated into every direct access trading platform, which allow you to send orders to the various market participants.The US stock markets:The NASDAQ is a computerized exchange without an actual trading floor. Orders are executed thru a complex computer system. You will find 2 types of market participants on the NASDAQ, Market Makers (MM) and electronic communication networks (ECN’s). There are various different Market makers as well as ECN’s which all interact thru computer systems.The NYSE is a centralized exchange where shares are traded on an actual exchange floor. Every stock traded on the NYSE has it’s own “specialist” who is responsible for maintain-ing a fair and orderly market in that particular stock.On the NYSE only the specialist has insight into the order book, which holds all the orders for the stock he is responsible for. Let’s assume you are trying to buy XYZ for $15 but the best seller wants at least $15.25 for XYZ. In this case your order will be placed in the specialist’s order book on the bid side and will be executed once a seller is willing to sell you shares for your limit price. The information in the order book can be very valu-able since big buy or sell orders are points of support/resis-tance.Bids and offers:The 2 main forces in the markets are supply (bid) and demand (offer/ask). It is basically a very simple concept. But many new traders are irritated by itThere are two ways to trade stocks based on bids and offers: Passive:Passive buyingPassive buying means that you are trying to buy a stock at a price that is lower than the current best ask price. Therefore your order cannot be executed immediately (since you are not agreeing to the seller’s price) and gets displayed on the bid side of the level 2.Passive buying means to place a bid and to wait for a seller to sell you his stocks.Passive sellingWhen selling passively you are trying to sell a stock at a higher price than the current bid price. Your order won’t be executed immediately and gets displayed on level 2.Passive selling means to place an offer (ask) and to wait for a buyer to buy shares from you.There is no way to ensure that your order gets executed when trading passively, since there might be no one willing to agree to your price.Active:Active buyingActive buying means to buy shares from an existing seller who has an offer in the market. You are agreeing to someone else’s price offer.Active sellingActive selling means to sell shares to an existing buyer who has a bid in the market.When trading actively you are most likely to get your order filled immediately, unless someone else steps in front of you.Remember that you can only get filled for as many shares as the counter-part is willing to trade. Therefore you might get partial fills.NASDAQ and level 2:Level 2 is a quote screen that displays all the competing bids and offers. These bids and offers come from big institu-tions and banks as well as individual traders displaying their orders thru ECN’s. There are over 400 registered market partici-pants who are able to place bids and offers in every single stock listed on the NASDAQ. Level 2 trading literally revolutionized the markets. The NASDAQ stock exchange was the first to introduce level 2. Meanwhile there are a few international ex-changes following.Here is a look at a level 2 window that also has order entry implemented:The upper part of the window gives you some basic information about the stock, i.e. the current price, the highest price of the day, the low of the day and the total volume traded.The next part of the window is used for order entry:Here is the part with the actual level 2 quote information:The left column displays all the buy orders:The higher the price that people are willing to pay for the stock, the higher the entry in the left column. The price on top is called the “best bid”. Each different color displays another price level. There is no other meaning to these colors.The right column displays all of the sell orders:The lower the price that people are willing to sell their stocks for, the higher the entry in the right column. The price on top is called the “best ask.”The prices on top of the two columns are the best prices available at the moment. They are referred to as the “inside market.” These prices will be the ones you can find in regular level 1 quotations.Let’s take a little closer look at the ask side of our level 2 window:The first column (MMID) gives you information about the market participant. The second column (ask) tells you what price the participant is willing to sell the stock for. The third column dis-plays the size at which he or she is willing to sell. You have to multiply the number by 100, so 10 would mean that there are 1000 shares for sale. In the screen above, RSSF for example, is trying to sell 1000 shares at a price of $62.The New York Stock Exchange (NYSE):Every stock listed on the NYSE has it’s own specialist. He is responsible for maintaining a fair and orderly market in that particular stock. If you send your order to the NYSE via a direct access trading platform it will be send (via SuperDOT) directly to the specialist’s order book for execution. The special-ist is the only one who has access to the order book. Orders are executed strictly on a first come first serve basis.It is the specialist’s responsibility to maintain a fair and orderly market. One example of this would be a situation where there is a huge sell order coming into the market but there are almost no buyers - without the specialist’s help the stock price would dump irrationally. It is his responsibility to buy the stock in this situation and to keep the stock at a “fair” level. The specialist is therefore always the buyer of last resort.Every order on the NYSE has the chance to receive price improvement. For example if you are trying to buy XYZ for $100 and someone is entering a sell order with a limit of $99 you would end up buying the stock for $99.5.Since the NYSE is not fully computerized you will notice a differ-ence in the speed of execution versus Nasdaq orders. This applies to the cancellation of orders as well. Even though it is slower I usually never wait longer than a few seconds for my order to get executed; only if there a buyers/sellers at my price limit of course.NYSE stocks in the level 2 window:If you place an NYSE symbol into a level 2 box you might be confused since there is more than just the NYSE displayed. This is because most of the stocks listed on the NYSE are also traded on various regional exchanges. Even though the quotes you see are in a level 2 box they are all level 1 quota-tions since they only display the inside market (best bid and ask).Here is an example:I highlighted the NYSE quotes in this example. The NYSE quotes are almost always the most important since the major market participants use the NYSE for executions. Other market participants here include “BOST” (Boston stock exchange) or “PHIL” Philadelphia stock exchange.NYSE stocks on “Island”:Many of the ma j or NYSE and AMEX stocks are no longer only traded on the traditional exchanges. They are now being traded thru ECN’s as well, with Island “ISLD” being the most important. The basics of Nasdaq order routing:Placing trades on the Nasdaq is a little more complicated than doing so on the NYSE. There are different order routes available. Those are Selectnet, SOES (small order execution system) and ECN’s (electronic communication networks).Selectnet can be seen as the center of the Nasdaq market even though it is only the second choice at best for most active traders. Access to Selectnet allows you to send your order to every available market participant. It is also possible to place bid and offers via Selectnet.SOES was implemented as a system for non-professional traders and allows them to execute their orders against market makers. SOES only sends the order out to market makers, not ECN’s. It’s mandatory for market makers to fill orders sent to them thru SOES.ECN’s are electronic networks that allow traders to execute orders against other ECN’s as well as to place their own bids and offers. Trading thru ECN’s is the fastest order way available since there are no middlemen involved and the ECN’s computers are usually very very fast. My ECN orders usually get filled immediately if I am agreeing to someone else’s bid or offer.Order routing can get pretty complex since there are different rules and limitations for each route. Luckily there are intelligent order systems out there, which take a variety of order systems into account and do the work for you, making order routing pretty easy for the most part.I will explain order entry in more detail later in this book.Short selling:Short selling allows you to make money on a falling stock price.When selling short, you sell a stock that you don’t own (you borrow it from your broker) and try to buy it back (covering)for a lower price. For example you sell 100 shares of XYZ short for a price of $10 per share. This will ad $1000 to your account. No, the money does not actually get credited to your account since you are only borrowing from you broker. If you buy those 100 shares back for $9 per share that will mean you have to pay $900 for that transaction, leaving you with a $100 gain. When you are shorting a stock, your potential risk is unlimited since a stock can go up more than 100% but sink not more than 100%. Therefore I would stay away especially from small stocks (they often rise dramatically in price) when shorting!Short selling rulesShort selling is a little more complicated than regular buying because the short selling rule (up tick rule) prohibits you from selling into an already falling stock price and therefore making an entry more difficult. In order to short a stock the current bid and ask prices must be on an up tick, meaning they must be higher than the previous price. Your order entry software will automatically prevent you from violating this rule. You will usually find an arrow in the upper part of your level 2 window that tells you if the stock is on an up tick. Even if the stock is not on an up tick you will always be able to short it on the ask side. When there is a lot of selling pressure though, chances of getting a fill on the ask are slim.Furthermore the stock you are aiming to short has to be available for borrowing from your broker. Your broker will hold a list of stocks you can almost always borrow and has a short lookup tool. I have had very good experiences with the availability of stocks for shorting.Basic rules for using technical analysis:Multiple timeframesMost traders use technical analysis as their primary tool to find potential trades and to determine entry/exit points. Only momentum traders and scalpers might only look at the stock movement or the supply and demand they can see on the level 2 screen.When using technical analysis it is very important to get the bigger picture of the stock’s price movement. That’s why you should always have a look at multiple timeframes of charts before making a trade. Imagine a stock is looking ready to go up on the 5 min chart but is running into strong resistance on the daily chart. You don’t want to get caught buying it here but rather wait for it to break that resistance before entering a long position.I always try to look at least one intraday chart as well as the daily chart. Previous days highs and lows are always points that are every important. Other timeframes that I like to look at are 5min and 15min charts.The perfect setup shows the same “picture” on multiple time frames. Here is an example of a stock that is breaking down on the intraday chart as well as on the daily chart:Please see next page.Market and sector analysisThe overall market is most likely to determine how strong the stocks you are watching might move. Make sure to not trade against the overall market and know what to expect every day. I use the same tools and patterns for market analy-sis that I use for stock analysis. The most important thing for me to look at is the previous day’s range. The previous day’s low will serve as support to the downside and the high will serve as resistance to the upside. Besides analyzing the overall market you should also know what the individual sectors are doing to further increase your success rate. A good top down approach would be too look at the overall market first, then to determine what the general direction is most likely going to be and to look for sectors that reflect that direction the best, and finally filter out stocks out of that particular sector that provide interesting set-ups.Indicator analysisBesides the price patterns described in this book there are vari-ous technical indicators that you can use in conjunction with them. The simplest technical indicators are moving averages. Others include stochastic, money flow, rate of change etc. Gen-erally speaking, the more indicators that confirm your setup, the better. I only use moving averages and stochastic for my trad-ing. Technical indicators go along with everything described in this book; they should be seen as additional tools. However, some trades might only use certain indicators and make trades based on them. I will not describe all the technical indicators in detail since it would be too much to fit in here and most likely just be confusing. I would rather refer to the link section on my website for further reading on technical indicators.Setup pricesA setup price is a predetermined price where you are looking to enter a position. Make sure that setup prices get broken significantly before you enter your position. For example if I am looking at a buy above $50, I would wait for the stock to break that price by approximately 5 cents. This varies though, and depends a lot on the stock I am trading. The important thing is that there are trades being made ABOVE the setup price in order for the setup to be valid.Also make sure, that the stock actually trades above the setup price. This can be problem with low volume stocks where the inside market (best bid and ask) changes without any trades taking place.Types of charts:The most common way to display charts is the line chart fol-lowed by the bar chart. In the bar chart the vertical line marks the high and low, the left horizontal line marks the opening price and the right horizontal line marks the closing price. If you selected a 5 min chart, that means that each bar reflects the price movement of only 5 minutes. In a daily chart each bar/ candle displays one entire days movement.The type of chart used most by active traders is the candle-stick chart. This type of chart has been in use for over 100 years and has its origin in Japan. It is also referred to as a Japanese candlestick chart. The color of the candlestick itself tells us if there was an up - or downtrend in that par-ticular timeframe and makes reading them very easy. There are also numerous indicator based on the shape of the candlestick itself. I will talk about the most common ones later.The following candlesticks are open candlesticks, meaning that their opening price was lower than the closing price and therefore reflect an overall uptrend in the timeframe you selected. The color used here for an open candlestick is green; sometimes people will use white instead.If the opening price was higher than the closing price you get a closed candlestick that reflects a downtrend. The colors used are usually black or red.The vertical line on the top of the candlestick is always the high, no matter what color the candlestick has. The line on the bottom always marks the low. These lines are also called shadows (upper/lower) or tail. There might be no shadows at all if the opening price marks the high and the closing price the low or vice versa. The colored part is always referred to as “the body” of the candlestick.Development of trends:There are 3 trends a stock can move in:a) Uptrendb) Sideways trendc) Downward trenda) An uptrend is a series of price advances followed by price declines that don’t violate the prior low (higher highs and higher lows).In an uptrend the prior low serves as support and the last high serves as resistance. The best trade during an uptrend is of course a long trade.At some point after a rise in price the stock will be “tired” and has to “relax” a little to gain strength to make a move again. This is when a sideways trend (consolidation) develops.b) In a sideways trend highs and lows are approximately on the same level. The highs mark resistance and the lows serve as support.After a long sideways trend stocks often reverse the prior direc-tion and fall in to a downtrend (in case the prior trend was up).c) A downtrend is a series of price declines followed by price advances that don’t violate the prior highs (lower highs and lower lows). The prior high serves as resistance to the upside and the prior low serves as support to the downside. On the next page you will see a chart displaying all the trends.Trend lines and trend channels:Trend lines and trend channels are a very important part in technical analysis since they define the trend itself and show you important areas of support and resistance. I use them mostly for the longer-term analysis based on daily charts.In an uptrend a line is drawn below the “major” lows of the trend. The uptrending line shows you relevant support. The opposite is done in a downtrend; you draw a line above the “major” highs of the trend. As with many things in technical analysis it is much easier to see what I am talking about by looking at an example:Trend channels occur in stable trends when you can draw a second (parallel) trend line in addition to the one we talked about before. This time we will also draw a line above the highs of the up trend and vice versa for down trends. By drawing this line we have established a trend channel that not only shows us support, but also shows the most likely range the stock will be trading in, thus us very nice entry points at support (refer-ring to the core swing trading buy setup) and profit targets at resistance.Moving averages:Moving averages are probably the most widely followed and therefore most significant indicators. And yet, they are very simple to use.Moving averages have multiple functions. They serve as im-portant areas of support and resistance and give trade signals if a stocks’ price is crossing above or below them. If a stock trades above the moving average line it serves as support to the downside, if it trades below it will serve as resistance to the upside. An example would be the 200MA, which is often used by fund managers. A stock that is trading above its 200 day moving average is generally a good long position, as long as it holds that moving average.。

NOMENCLATURE

Abstract Providing reliable group communication is an ever recurring topic in distributed settings. In mobile ad hoc networks, this problem is even more signiﬁcant since all nodes act as peers, while it becomes more challenging due to highly dynamic and unpredictable topology changes. In order to overcome these difﬁculties, we deviate from the conventional point of view, i.e., we “ﬁght ﬁre with ﬁre”1 , by exploiting the nondeterministic nature of ad hoc networks. Inspired by the principles of gossip mechanisms and probabilistic quorum systems, we present in this paper P ILOT (ProbabilistIc Lightweight grOup communication sysTem) for ad hoc networks, a two layer system consisting of a set of protocols for reliable multicasting and data sharing in mobile ad hoc networks. The performance of P ILOT is predictable and controllable in terms of both reliability (fault tolerance) and efﬁciency (overhead). We

清末民初的正定古迹老照片

清末民初的正定古迹老照片作者：张永波来源：《文物春秋》2010年第02期【关键词】正定;清末民初;老照片;古代建筑【摘要】本文介绍了清末民初由日本人山本赞七郎拍摄、收录于《震旦旧迹图汇》里的一组正定文物古迹老照片,并对其拍摄者、拍摄年代及相关问题做了考述。

我做导游工作讲解古城正定近十年,业余时间喜欢收集老照片。

在与各地朋友的交流中得知,当年日本人曾在中国营造学社之前对正定的文物建筑等做过调查、拍摄,所以近几年一直留心这方面的资料。

曾托好友吴晓冬在日本留学的同学帮忙寻找日本学者常盘大定、关野贞合著的《支那文化史迹》、《支那佛教史迹》、《支那佛教史迹踏查记》系列书籍,后该同学在日本学校的图书馆找到了几张古塔的老照片,转发给我一看,正是正定四塔的旧影,分别是广惠寺华塔(图一)、天宁寺凌霄塔(图二)、临济寺澄灵塔(图三)和开元寺须弥塔(图四)。

从照片中看,天宁寺凌霄塔前面有殿堂,现已毁无存;与开元寺须弥塔并排的东侧为钟楼,后面为法船殿,后者也已无存。

这几张照片是由山本赞七郎(山本明)拍摄的,曾收录在《震旦旧迹图汇》里。

这可能是目前我所见到正定古城文物建筑类最早的照片了。

几个月以后,我又见到山本明拍摄的一组14张正定的老照片,除《震旦旧迹图汇》里收录的正定四塔外,另有3张从不同角度拍摄的华塔(图五)、凌霄塔(图六)、须弥塔(图七)的照片,以及隆兴寺摩尼殿的背坐观音(图八、九)和已经被毁的大悲阁内宋代壁塑的照片(图一○～一三),十分珍贵!让我惊讶的是,其中一张隆兴寺摩尼殿背坐观音的照片(图八)竟然和我曾经在北京鲁迅纪念馆见到鲁迅书桌上的那张是同一版,这才使我反应过来,1923年鲁迅先生就是在北平山本照相馆买到的那张背坐观音照片,而山本照相馆的老板就是山本赞七郎。

这些老照片目前珍藏在东京大学东洋文化研究所。

为了弄清这组照片的准确拍摄年代,我查阅了有关信息和资料。

《震旦旧迹图汇》的作者是山本明,出版于日本昭和八年(1924年),总共有2编6册:第1编“云冈石窟”3册,即第1册文字说明及第2、3册图版;第2编“北京及其附近旧迹”3册,包括第4册文字说明及第5、6册的图版,均为汉、英、日文对照。

AN3419 System Integrity for XE

Freescale Semiconductor Application NoteDocument Number: AN3419Rev. 0, Draft A, 02/2007Contents©Freescale Semiconductor, Inc., 2007. All rights reserved.1IntroductionThe system designer is responsible for the correct operation of a microcontroller-based system. If an operational breakdown causes bodily harm, the system integrity will likely be the most important aspect of the design. In other cases, cost or functionality may be higher considerations. Many aspects of an embedded system’s operation can influence its operational integrity.This application note examines how you can use the on-chip capabilities of the S12XE to address memory integrity, operational correctness, and correct software behavior. The S12XE MCU family introduces features that can improve operational robustness and allow early detection and prevention of potential hazards in the system.1Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12Memory Integrity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12.1ECC System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22.2Margin Checking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33Correct Software Behavior . . . . . . . . . . . . . . . . . . . . . . . 53.1CPU Permissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53.2MPU Descriptors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73.3MPU Protection Response and Recovery. . . . . . . . . . . 113.4Operational Correctness . . . . . . . . . . . . . . . . . . . . . . . . 124Correct RAM contents. . . . . . . . . . . . . . . . . . . . . . . . . . 134.1Errors in Embedded Systems . . . . . . . . . . . . . . . . . . . . 134.2XGATE Co-Processor . . . . . . . . . . . . . . . . . . . . . . . . . . 134.3Functional Correctness . . . . . . . . . . . . . . . . . . . . . . . . . 144.4Timeliness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164.5Error Recovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185Summary of Techniques . . . . . . . . . . . . . . . . . . . . . . . . 216References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21System Integrity Techniques for the S12XEby:Steve McAslanMCD Applications, East KilbrideMemory Integrity2Memory IntegrityThe application software provides most of the functionality of the system. It is the aspect of an embedded system with the greatest influence on system integrity.In most S12X applications, the CPU software and important system data are stored in the on-chip flash. To help maintain the integrity of the on-chip flash, the S12XE includes an automatic Error Correction Coding (ECC) system in hardware and a margin-checking system that checks how completely a flash cell is programmed or erased.2.1ECC SystemThe ECC system, always enabled, detects and corrects some errors that may arise in the flash. It is applied to the P-flash in groups of four consecutive words known as phrases. Each phrase is assigned a syndrome value that allows error correction of one error and detection of two errors across the phrase.Each time its contents are read, the flash module checks the consistency of the data and the syndrome. If a single bit has changed, the error is corrected. This error correction is transparent to you and does not disturb the normal read process of the P-flash. The flash module also sets a single-bit error flag. You can choose to receive an interrupt if this occurs.If while reading the flash the module detects that two bits are in error, it sets a double-bit fault flag. Because these errors cannot be corrected, the value read from memory will contain an incorrect value. To avoid this problem, the user software may enable an interrupt when the module finds a double-bit error. If the error word contains an instruction opcode, the interrupt prevents execution of the incorrect opcode and allows the CPU to take corrective action. The flash error detection and correction process is shown in Figure1.System Integrity Techniques for the S12XE, Rev. 0, Draft AMemory IntegritySystem Integrity Techniques for the S12XE, Rev. 0, Draft AFigure 1. Flash ECC ProcessThe syndrome is automatically created each time the software programs a phrase in P-flash. The P-flashprogramming operation is shown in Figure 2.0111001011001010110110111111100010000001110111000010111010111000Write phrase to flash P-Flash P-Flash Memory controllerprograms data and syndrome into flash 010111010111001011001010110110111111100010000001110111000010111010111000Set ECC error flag if requiredSingle bit errorcorrected automaticallyRead wordfrom flashSet ECC error flag(with optional interrupt)Double bit error detectedP-Flash 1000000111011100Set ECC error flag(with optional interrupt)1000000110011110100000011001110000101110101110000111001011001010110110111111100001011101P-Flash100000011001111000101110101110000111001011001010110110111111100001011101Read wordfrom flashSystem Integrity Techniques for the S12XE, Rev. 0, Draft AMemory IntegrityFigure 2. P-flash Programming SequenceThe on-chip D-flash also has an automatic ECC system, but each word in D-flash is protected by its own syndrome. The module detects single- and double-bit errors and these errors can cause interrupts if necessary.ECC errors can be detected only if the software reads the flash. Performing periodic block reads of flash helps find these errors, especially in infrequently accessed flash. Because ECC is always performed on an eight-byte phrase, reading a single 16-bit word that spans the last byte of one phrase and the first byte of the next phrase performs ECC checks on all bits of the two phrases.If an ECC error occurs, you can choose a response dependent on the nature of the error. For example, a system with a single-bit error may continue operating, while a system with a double-bit error may perform a graceful power-down.2.2Margin CheckingThe S12XE allows a detailed examination of the actual programmed or erased state of each flash cell. This is achieved by setting the logic margin to a different value than that of normal operation.The MCU determines the binary value of a flash cell by sensing the current through the cell. In normal operation, the MCU uses a measurement level that allows for the maximum discrimination between normally programmed or erased values. The charge values vary across a narrow distribution for a correctly programmed or erased cell. If the cell is marginally programmed or erased, the actual cell charge margin0111001011001010110110111111100010000001110111000010111010111000Write phrase to flashP-FlashP-Flash100000011101110000101110101110000111001011001010110110111111100001011101Memory controller programs data and syndrome into flash Set ECC error flag if requiredMemory IntegritySystem Integrity Techniques for the S12XE, Rev. 0, Draft Amay be smaller than expected (see Figure 3). This decreased margin could be further reduced by normal variations of the MCU caused by voltage or temperature fluctuations, or by the gradual change in charge on the cell over the course of many years.Figure 3. Effect of Marginally Programmed CellsThe MCU provides two additional measurement levels that allow for a more accurate examination of the cell’s state. One of the two additional levels is closer to the programmed state (see Figure 4). This allows the user software to detect a cell with a smaller-than-expected margin, a margin that can be used to provide advanced warning of an incorrect reading. The erased margin level works in exactly the same way but towards the erased state. The alternative margin reads may be used in diagnostic routines that read the flash and ensure there are no ECC errors.Flash cell currentR e a d a s l o g i c 1R e a d a s l o g i c 0Reduced marginReduced marginNormal distribution of programmed cellsNormal distribution of erased cellsWeakly erased cellWeakly programmed cellNormal read pointSystem Integrity Techniques for the S12XE, Rev. 0, Draft ACorrect Software BehaviorFigure 4. Adjusted Margin Level3Correct Software BehaviorIn real world systems, it is generally impossible to prove that application software is error-free. This is especially true when multiple software components from several suppliers are integrated into a single device. Individual components may operate correctly on their own, but incorrectly when used together. Potential errors include modification of the private variables of another task, incorrectly calling another task, and changing the configuration of on-chip peripherals.The S12XE provides a protection scheme that prevents software components from inappropriatelyaccessing memory and peripherals. The scheme consists of restricted permissions on CPU operation and access descriptors in the memory protection unit (MPU).3.1CPU PermissionsIn an embedded system the CPU executes software that initializes and executes the application while handling interrupt events. During initialization, the CPU needs to modify RAM contents and configure peripheral registers and interrupts. During execution of the application, the CPU should not modify orR e a d a s l o g i c 1R e a d a s l o g i c 0Normal distribution of programmed cellsNormal distribution of erased cellsFlash cell currentNormal read pointAdjusted read pointWeakly programmed cell reads as erased with adjusted marginCorrect Software BehaviorSystem Integrity Techniques for the S12XE, Rev. 0, Draft Aaccess unused parts of the MCU. The CPU needs to modify the stack and clear flags when responding to interrupts. In general, the CPU’s required capabilities will vary according to the type of software being run.The S12XE addresses these requirements by providing three operating states with different levels of permissions (see Table 1). The state is controlled by the U bit in the CCR and the SVSEN bit in the MPUSEL register.3.1.1Supervisor State and Protected Supervisor StateAt reset, the MCU sets the CPU to Supervisor state and sets the MPU so that there is a descriptor for all memory. In Supervisor state, the CPU can access any resource on- or off-chip without restriction (as with CPUs in other S12 and S12X devices), making it an ideal state for initialization purposes and for use with kernal and interrupt tasks.The CPU can enter Protected Supervisor state by setting the SVSEN bit in the MPUSEL register. This prevents the CPU from accessing areas of memory that have not been specified by the MPU; but the CPU can change this bit at any time and is not subject to other limitations. This state may be useful for providing protection in kernel or interrupt tasks.3.1.2User StateWhen the CPU sets the U bit, it enters User state. This state limits the access rights of the CPU as well as its ability to change its state, the U bit, the I bit or X bit (in the CCR) via any opcode, execute the STOP or WAI opcodes, or access areas of memory not specified by the MPU. The only way to return to Supervisor state from User state is to receive an interrupt. User state is used during the execution of standard application code, such as user tasks. After the CPU is in the appropriate protected state, the MPU controls the access permissions to resources on- or off-chip.An interrupt causes the CPU to return to Supervisor state. In this state, the CPU can perform activities necessary for an interrupt such as modifying interrupt flags and copying data from peripherals. When the CPU returns from interrupt, the U-bit returns to its pre-interrupt condition. This is achieved by stacking the U bit as part of the CCR. It is not possible to force a return to Supervisor state by executing an RTI while in User state.Table 1. CPU Operating StatesStateUSVSENRestrictionsSupervisor 0X NoneProtected Supervisor 01Descriptor ranges onlyUser1X Descriptor ranges only; No interrupt control;No low-power capability (STOP or WAIT)System Integrity Techniques for the S12XE, Rev. 0, Draft ACorrect Software Behavior3.2MPU DescriptorsThe MPU provides eight descriptors, each of which defines an accessible memory range. Each descriptor applies to one or more bus masters. On the S12XE, the bus masters are the CPU (in Supervisor state and User state) and the XGATE.A descriptor consists of a memory range (defined by a start and an end address) that applies across the entire global memory space with a resolution of eight bytes. It also includes one or more bus masters, a control to make the range read-only, and a control that allows the bus master to execute code within the range. Without these additional controls, every address within the range would be readable, executable, and writable to the relevant masters. See Figure 5 for the content of the descriptors.Figure 5. MPU DescriptorsIn Supervisor state, the CPU configures the descriptors required by the next User state task. Aftercompleting the configuration, the CPU executes the User state task. The CPU typically jumps to the start of the task by modifying the return address on the stack and executing an RTI. This approach is shown in Figure 6. The CPU can alternatively set the U-bit and jump directly to the User state task.M0M1M2M3Protected Supervisor state (CPU)User state (CPU)XGATENot used on S12XEP100WP NEX Write protectedNo executionM0M1M2M3Start address [22:3]WP NEX 00End address [22:3]Correct Software BehaviorSystem Integrity Techniques for the S12XE, Rev. 0, Draft AFigure 6. Typical Flow When Using MPUSoftware integrators must pay attention to the configuration of descriptors. For a User state task, this requires a task code descriptor and a task data descriptor. In many cases, however, this is not enough. WhenD (A:B)X Y PCCCRU=1CPU StateActivityUser taskD (A:B)X Y PC = ISRCCRU=0InterruptI=1D (A:B)X YPC = ISRCCR U=0I=1M0M1M2M3Start address [22:3]WPNEXEnd address [22:3]M1M2M3Start address [22:3]WP NEX 00End address [22:3]M0M1M2M3Start address [22:3]WPNEXEnd address [22:3]Configure descriptors Configure PC on stackRTID (A:B)X YPC = User Task 2CCRU=1I=0User task 2M0Correct Software Behaviorcoding in C, the linker may add library calls to implement certain functions, so there must also be a descriptor that fulfills that requirement. There may also be task-specific constants stored in emulated EEPROM or in flash that also require a descriptor.There are no restrictions on the CPU when in Supervisor state. All memory is accessible so no specific requirements for descriptors exist. But for a Protected Supervisor state task, the CPU must have access to the task code, variables, interrupt vectors, stack and register space, and library code and constants.The system integrator must also configure the correct type of access. The NEX-bit in each descriptor prevents the bus master executing code from within the descriptor range. This bit must be cleared for code spaces like application and library code. The WP-bit prevents the bus master from writing to the memory range. This is useful when sharing global variables in RAM or for allowing tasks access to the current input status of ports (but preventing them from changing the port configuration). See Table2 for the combinations that these bits create.Table2. Combinations of PermissionsWP NEX Access Example of use00Read, write, execute Executing code from RAM01Read, write Stack and data regions10Read, execute Code in flash11Read Data in flashIt is possible to combine descriptors by overlapping their memory ranges. The overlapped region accumulates the restrictions of the source descriptors. In Figure7,descriptor 0 applies to CPU User state from address $40_0000 to $41_BFFF with read-only and execute permissions (typically a flash code configuration). Descriptor 1 also applies to CPU User state, but goes from address $40_8000 to $41_FFFF with read/write and no execute permissions (typically a RAM data configuration). Because the descriptors overlap, an overlapped range is created from $40_8000 to $41_BFFF. This accumulates the restrictions from the two descriptors and so is a read-only and no execute region (typically flash data configuration). Figure8 shows an example of the use of descriptors in a simple application that contains a scheduler, two CPU tasks, and an XGATE thread.System Integrity Techniques for the S12XE, Rev. 0, Draft ACorrect Software BehaviorSystem Integrity Techniques for the S12XE, Rev. 0, Draft AFigure 7. Effect of Overlapping DescriptorsM0M1M2M3Start address [22:3]WPNEXEnd address [22:3]1100_0000_0000_0000_0000_00100100_0001_1011_1111_1111_1PD0M0M1M2M3Start address [22:3]WPNEXEnd address [22:3]1100_0000_1000_0000_0000_01000100_0001_1111_1111_1111_1PD1M0M1M2M3Start address [22:3]WP NEX 00End address [22:3]010100_0000_1000_0000_0000_01100_0001_1011_1111_1111_1Overlapped range1Correct Software BehaviorFigure8. Example of a Simple System Using the MPUAs seen in the figure, the tasks can call the scheduler via an interrupt opcode provided by the S12XE called SYS that is not inhibited by the I-bit in the CCR. In this configuration, the scheduler is considered trusted code and can run in Supervisor state.3.3MPU Protection Response and RecoveryIf the MPU detects that a bus master has accessed resources outside of permissions defined by the descriptors, it causes an interrupt that prevents intended access. If the CPU violates the memory protection, it receives a non-maskable MPU access interrupt.The CPU interrupt sets the AEF-bit in the MPU flag register (MPUFLG) and loads the address where the violation occurred into the MPUSTAT[0:2] registers. The MPUFLG register also contains information about write or execute violations in the WPF and NEXF bits, respectively.System Integrity Techniques for the S12XE, Rev. 0, Draft ACorrect Software Behavior The XGATE is stopped if it violates the protection configuration, and an XGATE software error interrupt flag is set so that the CPU can manage the error. In this case, the AEF bit is not set and the MPUSTAT[0:2] registers are not loaded with the violation address. The CPU can read the XGATE core registers (including PC and CCR) directly from its register space to determine the XGATE violation condition.When defining memory protection, you must consider an error recovery strategy. You may need to place the application in a known good state or perform a graceful shutdown. You may be able to shut down the task and give a warning to seek assistance if the error occurs during a non-critical task. Ultimately, the error-recovery strategy depends on the application requirements.In Protected Supervisor state, CPU access to MCU resources is controlled by descriptors. These must be configured carefully to allow operation in normal and error conditions. The MCU helps in cases of CPU violation by automatically creating an explicit descriptor for the register space within the MCU. This means that the CPU recovery routine can make adjustments to the register space when in Protected Supervisor state even if such access is not normally allowed. Descriptors for the error interrupt vector and the error recovery ISR must be included to allow the CPU to process the error interrupt correctly.If the interrupt vector is not included in a descriptor range, then the original application violation leads to an interrupt that causes another interrupt when it attempts to fetch the vector and so on. This also applies to the ISR code and to any other resources the ISR requires.Problems can arise because the CPU and XGATE perform pre-fetches of opcodes to maximize the performance of the MCU. Because the MPU cannot distinguish between pre-fetch opcodes that execute and those that do not, several extra bytes must be added beyond the end of the memory range. The CPU reference manual contains information about its queue and the microcontroller reference manual contains information for the XGATE.3.4Operational CorrectnessThe MPU protects against software that accesses invalid resources. However, even if a function has not accessed an invalid resource, it may continue to operate incorrectly. Errors causing incorrect operation or slow responses in the system can arise in variable contents or stacks because of unusual combinations of system events. You can address this type of error by providing diagnostics that can detect or correct undesired behavior that occurs once the system is in use. Often provided at reset, these diagnostics perform detailed checks on the status of the system before the operation begins. Hardware interlocks like the COP watchdog or ECC protection on the S12XE are two of the diagnostics available when the application is running.You can also add functional checks to running systems. In traditional single-core MCUs, this capacity is limited by the additional load the checks place on the CPU as well as the fact that they are running on the processor whose performance they are checking. With these limitations in mind, you can see why the dual core S12X family brings specific advantages for embedded diagnostics.System Integrity Techniques for the S12XE, Rev. 0, Draft ACorrect RAM contents4Correct RAM contents4.1Errors in Embedded SystemsThe operational times for many embedded systems extend to weeks or even longer between resets. These systems are at risk of accumulating small errors that eventually disrupt normal operation. These errors affect volatile memory and are caused by memory leaks, stack overflow, array index overrun, and invalid pointer values. The number of errors can be reduced in embedded systems, though, because dynamic memory may be allocated on the stack and so this can be classified as a stack overrun. Similarly, it is possible to combine array overruns and invalid variable values because the overrun is an issue only if it damages another variable.It is possible for the MPU to detect these errors if the software stack and variables have some natural separation in memory. This makes it easier to separate protection descriptors. It is common practice, however, to group a task’s RAM requirements under a single descriptor.4.2XGATE Co-ProcessorBecause the XGATE co-processor shares access to RAM, it can validate its contents and identify some of the errors listed above. In Figure9, the XGATE verifies the correct extent of the stack by examining the contents of a buffer zone.@interrupt void XG_StackCheck(void){char i;/* Check that each word in the stack buffer zone isat the default value */for (i=0;i<stackMax;i++)if (stackBuffer[i] != stackDefault)/* Call response function to handle */XG_StackFail(i);}Figure9. Stack Overflow TestingFigure10shows how a secondary processor can validate variable contents and detect errors on the primary core. The code declaration shows an array and a far (24-bit) pointer placed beside each other in memory. There is no range checking on the received message length, and a message longer than four bytes causes an overflow into the pointer. Because the pointer is used rarely and because a five byte message is rarely received, the error can go undetected for an extended time.The XGATE code checks the pointer value and sends an interrupt to the CPU to signal that the value is incorrect. It can also restore the pointer to a known good value, can ensure that peripheral configuration registers have the correct values, and can verify that the CPU is operating at a functional level.System Integrity Techniques for the S12XE, Rev. 0, Draft ACorrect RAM contentsSystem Integrity Techniques for the S12XE, Rev. 0, Draft AFigure 10. Variable Validation4.3Functional CorrectnessA dual core system provides opportunities for functional correctness that do not exist in single-core systems, such as the double-checking of algorithms, dual-key authorization, and challenge-response authentication.You should ensure that the outcome of an algorithm is correct in all circumstances for certain applications, particularly in environments where the microcontroller affects the active or passive safety of a system. Figure 11 shows an example in which both cores compute the outcome of an algorithm but must agree on it before it is used. Run different versions of the algorithm on each core to ensure independent outcomes. In some cases, an approximation of the correct outcome may be sufficient as a margin check. Even if identical high-level code is used for both algorithms, the machine-dependent implementation iscompletely different because of the architectural differences between the CPU and XGATE on the S12XE. This provides an added measure of protection against algorithmic or compiler errors.CPU Codeunsigned char message[4];far char* menu;interrupt void SCIRx(){char byte;/Assume no receive errors byte = SCI->scisr1.byte;byte = SCI->scidrl.byte;if (index == 0){index = 1;message[0] = byte } else {message[index] = byte;index++;}if (message[0] == index)index = 0;}XGATE Checking Codeunion u_farptr {unsigned long l;unsigned int i[2];}farptr;interrupt void Checker(void){// Check far pointers are validfarptr = menu;if (farptr.i[0] !=menupage)_SIF(SW1);}System Integrity Techniques for the S12XE, Rev. 0, Draft ACorrect RAM contentsFigure 11. Example of Algorithm CheckingIn the approach illustrated above, a single core determines the validity of the algorithm output. If an error occurs during the operation of this hardware or software, it is conceivable that this decision may not be correctly evaluated. In this case, a dual-key approach may offer additional integrity. Seen in Figure 12, this approach allows both cores to present their calculated results to an independent checkpoint such that both can agree on the conclusion before the specified activity occurs.Calculations agree?ActionReviewNoYesCPUXGATEExecute full algorithmExecute alternative algorithmOR Execute partial algorithmExecute full algorithmCorrect RAM contentsSystem Integrity Techniques for the S12XE, Rev. 0, Draft AFigure 12. Dual-Key OutputYou can use a variation of the cryptographic challenge-response approach to assess the accuracy of the primary core’s activity. The secondary core can challenge the primary core to provide information on its status, and then check that the response was provided in a correct and timely manner. One useful challenge might be to return its program counter, stack the pointer values and any other useful internal registers, perform a checksum on a fixed area of memory, and verify the interrupt vector locations. The returned information would be checked against the values known to or calculated by the secondary core.Even if a system is behaving correctly, there may be an issue with its response time.4.4TimelinessThe overall loading of the processor influences the timeliness of a system response. A system designer must ensure that the highest priority tasks can directly access the CPU while also preventing lower priorityExecute full algorithmCalculation positive?Action 1CompleteNoYesCPUXGATEExecute full,Action 2NoYesCalculation positive?Completepartial, or alternative algorithmMCUAction 1Action 2ConfirmHardware check。

清末民初书画家名单

清末民初书画家名单清末民初，是中国书画艺术历史上一个特殊而浓烈的时期。

在这个时期，书画艺术的风格多种多样，画家、书法家也层出不穷，形成了一个独特的时代风貌。

在这些杰出的书画家中，有常常被提及的，也有默默无闻却对艺术史留下深刻印象的。

本文将根据不同的类别，列举一些清末民初著名的书画家名单。

山水画：1. 张大千（1899-1983）张大千是二十世纪中国著名的文化名人，他的山水画独具特色，有着浓郁的东方色彩和西方艺术的影响。

他擅于运用传统山水画法，但又不拘泥于传统，大胆创新。

2. 于右任（1879-1964）于右任不仅是一位政治家和文学家，同时也是一位画家。

他的山水画给人以深邃、沉稳的感觉，笔触细腻而又具有浓厚的历史文化氛围。

花鸟画：1. 齐白石（1864-1957）齐白石是中国现代书画界的代表人物之一，花鸟画是他擅长的领域。

他的花鸟画风格清新自然、富有生命力，尤其擅长描绘老鱼、老虾、老鸟等寿命较长的动物。

2. 陈师曾（1880-1971）陈师曾是清末民初最具代表性的画家之一，他的花鸟画观察入微、造型传神，追求自然传神、始终贯彻真实主义的原则。

人物画：1. 吴昌硕（1844-1927）吴昌硕是中国近代著名的人物画大师之一，他擅长人物像的刻画、形神兼备，准确地表现了人物的气质和精神，是中国文化艺术的宝贵遗产。

2. 许鼎（1902-1992）许鼎是著名的画家、美术教育家，擅长绘画人物画。

他的画作描绘了各种不同阶层的人物，造型鲜明，个性丰富。

尤其以民族英雄、后妃女性为主的人物画，富有美感和时代气息。

书法：1. 王羲之（303-361）王羲之是中国历史上最伟大的书法家之一，他的书法以草书、隶书、篆书著称。

他的书法多种多样、风格独特，给中国书法艺术史留下了深刻的印象。

2. 欧阳询（557-641）欧阳询是唐代书法家、书法理论家，其书法风格优美大方、宽阔流畅，被誉为“书中之圣”。

以上只是人们熟知的部分书画家，还有很多未被广为人知的优秀艺术家。

晚清1899老照片还原百年前的中国社会

晚清1899老照片还原百年前的中国社会清朝已经离我们远去，许多往事已然模糊。

让我们来重拾那份记忆，看看这一个真实的晚清社会。

以下照片为当时法国驻清朝领事方苏雅所拍，地点在大多在云南。

1899年的昆明城内的集市晚清时期昆明茶馆一景昆明街头清朝人装束，看他穿着打扮应该是富裕之人。

1899年，云南红河，被吊死的囚犯。

补碗匠。

这个职业现在没有了，现在谁家碗破了还会去补呢。

但那时候如果碗破了便扔掉，那时土豪的做法，一般人家都会“缝缝补补又三年”。

三人抬轿，中间那一个人主要起辅助性的作用。

云南的地方戏，滇戏。

富人的太太们，坐着的太太应该是原配，站着的是二房。

广西提督苏元春一身戎装打扮，看来他与法国领事方苏雅私人关系不错，要不然方苏雅不会为他拍照。

清末的庙会也是非常热闹的义和团运动，八国联军侵华……一系列不稳定因素使得晚清时候民不聊生，乞丐遍地走。

更夫的行头茶马古道，的确是山高路远坑深！茶马古道上的背夫们看到方苏雅这位外国人皆充满敬畏，纷纷给其让路，而此刻方苏雅按下了快门。

1899年，昆明，游街示众的囚犯（后面带伽者）云南少数民族崇拜勇士，所以摔跤这项运动在此已延续了千年。

摄于1899年，云南河口县。

清末时候的送葬队伍看西洋景的小孩。

西洋景又作“西洋镜”，是一种清末民间娱乐装置，用若干幅可以左右推动的画片，观众从透镜看放大的画面。

画片多是西洋风情，故叫西洋景。

云南讲武堂兵士训练中总督的排场。

云南总督骑着高头大马行走在昆明的大街。

1900年义和团运动，方苏雅为躲避义和团，曾逃回法国在越南的殖民地。

这就是以上照片的拍摄者、法国领事方苏雅。

他身上穿的不是戏袍，而是越南皇帝的龙袍。

其实当时越南是法国的殖民地，即使没有龙袍，这些法国官僚也是越南的土皇帝！。

tof-confidence-cvpr2011

Time-of-Flight Data with ConﬁdenceJozef Doboˇs ,Leto Peel †,Tim Weyrich,Gabriel J BrostowAdvanced Technology Centre,BAE Systems,Bristol,UKUniversity College LondonAbstractTime-of-Flight cameras provide high-frame-rate depth measurements within a limited range of distances.These readings can be extremely noisy and display unique errors,for instance,where scenes contain depth discontinuities or materials with low infrared reﬂectivity.Previous works have treated the amplitude of each Time-of-Flight sample as a measure of conﬁdence.In this paper,we demonstrate the shortcomings of this common lone heuristic,and propose an improved per-pixel conﬁdence measure using a Random Forest regressor trained with real-world ing an industrial laser scanner for ground truth acquisition,we evaluate our technique on data from two different Time-of-Flight cameras 1.We argue that an improved conﬁdence measure leads to superior reconstructions in subsequent steps of traditional scan processing pipelines.At the same time,data with conﬁdence reduces the need for point cloud smoothing and median ﬁltering.1.IntroductionTime-of-Flight (ToF)cameras have been successfully used for a variety of applications such as Simultaneous Localisation and Mapping (SLAM)[22,25],3D recon-struction of static scenes [10,34],and object tracking and scene analysis [27,33].In contrast to stereo vision and triangulation-based scanners,the ToF camera operates from a single viewpoint and does not rely on matching of cor-responding features,which greatly increases its robustness in the presence of traditionally difﬁcult scene materials and internal pared to ToF laser scanners,ToF cameras deliver a lower-resolution 2D depth image at much higher framerates and at competitive costs of deployment.With costs expected to dramatically decrease in the near fu-ture,ToF cameras are likely to become a commodity item.Despite the advantages of this new technology,it suffers from problems such as low signal-to-noise ratio,multiple light reﬂections and scattering.All these affect the recorded1For code,data,further results and other supplementary material pleasesee /pubs/tofconfidence/(a)3D Point Cloud (b)Intensities(c)AmplitudesFigure 1.Flying pixels.(a)Color-coded ToF depth map (red indi-cates close geometry,blue indicates distant geometry).Erroneous depth readings are caused by materials with unsuitable reﬂectivity and large depth discontinuities (see bottom ellipse).These pixels collect modulated light from both the foreground and the back-ground and give a resulting distance which is somewhere between the two.The top ellipse shows pixels ﬂying towards the camera,but this is a different phenomenon,caused by highly specular re-ﬂections from the top of the monitor.(b)intensity and (c)ampli-tude images.depth values and hence produce erroneous 3D points,whose reliability depends on various scene parameters.Previous works [21,34]rely on the amplitude values (Fig.1(c))as an indicator of conﬁdence for point removal and ﬁltering.In agreement with Kolb et al .[16],our experiments reveal that simply thresholding low-amplitude values is insufﬁcient to remove inaccurate pixels—valid points could be lost before ﬂying pixels and other anomalies as depicted in Fig.1(a)are removed.We observe that basing a conﬁdence measure on the im-age amplitude alone disregards valuable additional informa-tion.Explicitly incorporating additional cues,such as depth and local orientation,however,may be challenging,due to the difﬁculty of developing a descriptive error model that covers their interplay.Instead,we propose to use RandomForests[5]to train a regressor to infer the reliability of a depth sample from these input quantities.Random Forests have proven useful as a supervised learning technique that performs feature selection,and are even being used for real-time3D classiﬁcation tasks[32].Analogously to Mac Aodha et al.[19]and Carreira and Sminchisescu[6],we obtain a per-point conﬁdence assign-ment,which could be exploited in other dataﬁltering and 3D reconstruction systems[23].We qualitatively and quantitatively evaluate the robust-ness of our approach with two different ToF cameras,com-paring it against baseline approaches of depth sampleﬁlter-ing.We show that our machine-learning-based approach outperforms previous methods,leading to a conﬁdence measure of high discriminative power.2.BackgroundOptical range cameras[17,30]measure the time ofﬂight of near-infrared(NIR)light emitted by the camera’s active illumination and reﬂected back from the scene.The illu-mination is modulated either by a sinusoidal or pseudo-noise signal,depending on the camera model.The phase shift of the received demodulated signals conveys the time between emission and detection,indicating how far the light has traveled and thus indirectly measuring the depth of the scene.Due to the periodicity in the modulation signal,these devices have a limited working range,usu-ally only up to7.5m.Distances beyond this point are recorded modulo of the maximum depth,known as wrap-around error[24].ToF cameras output images of dis-tance and grayscale intensity(Fig.1(b)).With knowledge of the camera intrinsics,the distance readings can be con-verted to a3D point cloud(Fig.1(a)).Some ToF sensors also record an amplitude value at each pixel,which indi-cates the amount of modulated light reaching the sensor and disregards external illumination and incoherent stray light(Fig.1(c)).Similarly to intensity,this amplitude is inﬂuenced by scene depth,surface materials and orienta-tion,and lens vignetting[15,22].Low amplitude suggests that a pixel has not received enough light to produce an accurate depth measurement,although saturation of high-intensity responses may also lead to invalid readings,de-spite a high amplitude[13].If the amplitude values are not provided by the camera,intensity images have been used instead[10,21].2.1.Depth Corrections and CalibrationWhen dealing with measurement devices,it is impor-tant to properly understand the causes of potential errors and calibrate for them.Given the optical properties of ToF cameras,a standard intrinsic camera calibration sufﬁciently describes the focal length,principal focal point and lens dis-tortion[14,18].To further improve the unreliable depth readings caused by systematic errors,Lindner and Kolb[18] estimate a higher-order polynomial function which takes into account global as well as per-pixel inaccuracies.As-suming such initial depth calibration,Beder et al.[2]fur-ther estimate the camera’s3D pose using a single image of a checkerboard.Unfortunately,as Boehm and Pattin-son[3]recently showed,even a careful calibration is not sufﬁcient for real-life scenarios.Despite calibration,they report errors in pose estimates that are an order of magni-tude larger than the error modeled by the calibration.These errors are scene-dependent and cannot be overcome by cali-bration[3,20].Fuchs and May[12]and May et al.[21,22], on the other hand,ﬁnd known camera positions using a robotic arm to achieve high-precision results for both the initial calibration andﬁnal3D reconstruction,using vari-ations of the Iterative Closest Point(ICP)[26]algorithm. They demonstrate an overall error accumulation which has to be globally relaxed when closing a loop.In contrast,Schiller et al.[28]use up to80images from each camera from their rigidly mounted multi-recording setup,improving on previous results of single ToF cam-era calibration methods.Similarly,Kim et al.[15]combine several video and ToF cameras,while taking into account the scene-dependent effects.They introduce ratios of nor-malized amplitude values,which are median and Gaussian ﬁltered.Random noise,however,is assumed to be negligi-ble.2.2.3D Reconstruction and FilteringFeulner et al.[10]register consecutive ToF frames by detecting binary edge presence in the intensity images and aligning their corresponding3D coordinates by maximiz-ing the uncentered correlation coefﬁcient.In contrast, Fuchs and May[12]ﬁlter points at depth discontinuities, as these exhibit the largest distance error.Swadzba et al.[34]present a full acquisition pipeline that relies on several preprocessing steps to improve depth accuracy:a varying distance-adaptive medianﬁlter is applied to the intensity, amplitude,and depth images,and points with low ampli-tude are thresholded.Subsequently,a custom-made neigh-borhood consistencyﬁlter detects and removesﬂying pixels at edge locations in the distance image.Nevertheless,all of the above produce only binary classiﬁcation of erroneous depth readings.Fusing high resolution color images with ToF data pro-vides enhanced range maps,as shown by Yang et al.[36]. In their work,a bilateralﬁlter aggregates probabilities of depth estimates based on color afﬁnity,iteratively reducing the level of up-sampled blur that occurs due to interpola-tion in discontinuous areas.Our main inspiration,however, comes from the work of Schuon et al.[29].Their Lidar-Boost method combines several noisy ToF frames into a sin-gle,high-resolution depth image.By assigning zero conﬁ-Figure2.Pipeline for computing a conﬁdence value for each pixel in a ToF depth-scan.The ofﬂine supervised training stage builds a Random Forest for regression by correlating extracted features (see§3.4)with depth errors.These errors were computed by com-paring the raw depths from the ToF camera against the“ideal”geometry of the training scene.dence to low amplitude pixels(i.e.thresholding),the results are further improved.This technique was later built upon by Cui et al.[8]to achieve state-of-the-art reconstruction results.In their work,one point cloud is randomly selected as a reference model,and remaining frames from the se-quence are aligned with it.At each point,a multi-variate Gaussian withﬁxed uniform covariance is centered,and the maximum likelihood estimate is found through Expectation Maximization(EM).Unfortunately,all sources of system-atic error(such as surface orientation and reﬂectance,cam-era’s integration time,or depth)are neglected,and the only assumption is an increasing bias growing from the image center.In contrast,our approach derives conﬁdence measures across both systematic and random errors.As a single-frame method operating directly on the camera output,our method is further amenable to combination with many of the approaches mentioned above,providing an improved prior for depth sample conﬁdences.3.Learning ToF ConﬁdenceEach model of ToF camera has its own set of charac-teristics and inaccuracies.To make meaningful predictions about the conﬁdence of a camera’s output,we propose that these characteristics can be learned by comparing data from the camera with ground truth distance.The objective is to assign a conﬁdence value in the range[0,1]to each point produced by a ToF camera,where1signiﬁes a distance reading believed to be completely accurate while0signiﬁes a reading that should be ignored.This assignment of conﬁ-(a)room scene(b)stairs scene Figure3.Intensity images of training scenes from the laser scan-ner.dence is performed with a trained Random Forest operating in regression mode.Regression is preferable here to clas-siﬁcation because conﬁdence is continuous,and different applications will stipulate their own expectations of accu-racy.Other supervised learning systems such as SVMs[7] could also be used.Random Forests were chosen because they automatically perform feature selection and they re-quire no cross-validation due to the use of out-of-bag es-timates during training[5].We acquire training data of the form{(x i,y i)}N i=1.Each feature vector x i is encoded from a single point recorded by the ToF camera,and the corresponding y i is the target value,or conﬁdence.These conﬁdences are computed from the difference between the ToF’s reported distance and the ground truth distance for each reading.When the disparity between these two val-ues is low,the correct label is close to1,whereas for pixels where the camera’s reported distance is signiﬁcantly in er-ror,the label value is close to0.Once the Random Forest has been trained,we can compute a prediction y∗for an un-seen x∗,allowing a conﬁdence map to be created for each new image(see Fig.2).3.1.Data CaptureTo compute ground truth label values for each pixel,a true distance imageˆD must be generated.This image is the same resolution as the output from the ToF camera,and each pixel contains the true distance,i.e.the reading that the ToF would have produced were it completely accurate.The viewpoint of this perfect depth image must match the actual location of the ToF camera to be meaningful.In practice, we obtained ground truth byﬁxing a ToF camera close to a high-end,Time-of-Flight laser scanner,so that the view frustra overlap as much as possible(see Fig.4).In princi-ple,its measurements may be subject to similar errors as the ToF camera.The scanner’s precision and resolution are sig-niﬁcantly higher than the camera’s[3,8,11],which makes it sufﬁcient for ground-truth in our case.3.2.CalibrationThe ToF camera intrinsics were computed using images of a checkerboard and the Bouguet toolbox[4].The inten-sity images I have properties similar to a standard grayscale camera,so no modiﬁcations to the calibration code are re-quired.The intrinsic parameters included focal lengths in x and y,the principal point,and a5-parameter lens distortion model.The extrinsics of the scanner relative to the ToF camera were computed by placing targets in the scene.Pa-per photogrammetry targets(see Fig.3)with printed cir-cles were attached to availableﬂat surfaces throughout the scene,spanning the range of depths and spread across the jointﬁeld of view of both systems.The2D location of el-lipse centers was determined with sub-pixel accuracy in I, and in an equivalent intensity image from the laser scan-ner.The2D target centers in I and the3D locations of the target centers determined by the laser scanner were used to compute an estimate of the extrinsics which minimized the squared reprojection error.Due to the low resolution of current ToF cameras,the extrinsics calibration based on target detection may only be accurate enough to initialise registration.Experiments with targets at the back and sides of the capture volume showed that background geometry reprojected accurately but fore-ground objects were offset sideways inˆD compared to the ToF recorded ing unweighted rigid ICP,the initial registration was reﬁned by aligning the laser point cloud with the ToF point cloud.A streamlined one-shot registra-tion technique could be useful if large quantities of training data were acquired in the future,possibly leading to even better conﬁdence estimates.puting Ground Truth DepthAfter calibration,a projection of the laser geometry into the ToF camera is computed for comparison against the ToF’s own depth image.The laser scan has much higher res-olution than the ToF,(a typical scan was1251×1055com-pared to200×200for the ToF,while covering roughly the sameﬁeld of view),but still consists of points not surfaces. As a result,many laser scanned points at various depths fall within the point spread function of a ToF pixel,as shown in Fig.4.One could choose the closest,i.e.frontmost of the points forˆD i,j,butﬂying pixels occasionally occur in laser data as well and could provide erroneous depths.Neither the minimum nor the mean of these points’depths would be appropriate(Fig.4).All laser points which project into the sensor region of pixel(i,j)are used to computeˆD i,j. Weﬁt a Mixture of3Gaussians using EM to the depth val-ues of these points.ˆD i,j is set to the minimum of the means of theﬁtted Gaussians.Finally,due to parallax and occlu-sion,the ToF camera may be sampling parts of the surface geometry that were unseen by the laser scanner.By placingFigure4.Overhead view of a scene acquired by the laser scanner (red)and ToF camera(blue).We project the laser geometry into the ToF camera(green)without considering occlusions.The inset depth histogram shows how laser points from distinct locations can project into a single ToF pixel,but only points from A are valid. The area visible to the ToF between D and E is not covered by the laser scanner and so no training data is acquired here.the ToF camera very close to the laser scanner,we minimize the areas of the ToF image for which no ground truth exists.Each depth value inˆD must be converted to a training la-bel y∈[0,1].Intuitively,the difference betweenˆD and the ToF’s D is normalized byˆD toﬁnd the relative depth er-ror.Relative depth error is used because we assume that it is unrealistic to expect ToF cameras to have uniform absolute accuracy at all depths[3,11,15].The relative depth error is passed through the arctan sigmoidal function,reﬂected, and rescaled so that a relative distance error of0becomes 1,and a hypothetical inﬁnite distance error becomes0.The ﬁnal conﬁdence label y isy ij=1−2πarctanαabs(ˆD i,j−D i,j)ˆDi,j,(1)where the parameterαcontrols how quickly y tends to0as relative distance error tends to inﬁnity.3.4.Feature ExtractionA feature vector x i,j is computed for each pixel(i,j)in the ToF image.To allow the Random Forest to generalize the properties of pixels with high and low conﬁdence,three types of data are used to construct the feature vector.Note that Random Forests calculate the importance of each entry in the feature vector as part of the training process.This feature selection property makes it possible to use a large number of features that may be correlated with conﬁdence, and have the training process automatically determine which ones were in fact useful.The categories of features are:local features from a single point(i,j),spatial features calculated from a neighbourhood surrounding(i,j),and global features calculated from the entire frame.Local Features.The primary feature values at each ToF pixel(i,j)consist of the intensity I i,j,signal amplitude A i,j if available,and distance D i,j.These elements are included to allow learning of different physical phenomena, e.g.pixels not receiving enough light to compute an accu-rate distance where the scene has low reﬂectivity in NIR light spectrum,or receiving too much light so that pixels saturate.Distance is included because,in accordance with the inverse square law,surfaces a larger distance away are likely to have a higher error magnitude.The radial distance of each pixel is also included in the feature vector to allow the possibility of learning different error characteristics near to the edge of an image.Spatial Features.Filters are used to incorporate local spatial information about the scene.Neglecting perspec-tive projection of the camera,weﬁnd approximations to the normal angle in x and y.A Laplacianﬁlter is convolved with both the distance and intensity images.Thisﬁlter is commonly used for edge detection,so it is included in the feature vector to allow the forest to learn the relationship between error and depth discontinuities.Our initial obser-vations indicated thatﬂying pixels at depth discontinuities were among the most noticeable artifacts in ToF data,but that not all sharp depth changes were incorrect.As well as the3×3Laplacian kernel,5×5and7×7versions are used to incorporate information at different scales.Gaborﬁlters[9]describe a family of kernels,which differ from the Laplacian in that they can be set to a particular orientation.Whereas a large response from the Laplacian indicates the presence of an edge,Gaborﬁlters at different orientationsθproduce a response which can determine the direction of the edge.Filters at0◦,45◦,90◦and135◦were computed over the image.Due to the more complex form of the Gabor wavelet,a13×13ﬁlter was used over the whole image.Global Features.Inspired by Kim et al.[15]we allow the forest to learn the relationship between some feature at a particular point and the distribution of feature values across the whole image.For every feature listed above except ra-dial distance,the minimum,maximum,mean,and median values across the whole frame were included.Because of this,each feature already mentioned contributes5elements to the feature vector,giving aﬁnal dimensionality of91(18 features×5,plus radial distortion).Ourﬁnal feature vector takes advantage of many of the cues used deterministically in previous works to probabilistically assess the quality of ToF data.4.ExperimentsTo evaluate the conﬁdence assignment method,scenes were captured with a PMD CamCube2.0ToF camera.This camera has a resolution of200×200px,and provides both amplitude and intensity images.One scene of a stair-case(stairs Fig.3(b))and one scene of an ofﬁce(room Fig.3(a))were captured,both with a number of clutter items placed at different depths in the scene.The objects were chosen to include a wide variety of surface reﬂectance prop-erties,including plastic,wood,textiles,polystyrene,and brushed metal.The object placement was intended to create multiple depth discontinuities of varying magnitudes,and to span a variety of different errors on which to train the model.For these scenes,a laser scanner was used to ac-quire ground truth,so the scenes could be used for training and quantitative experiments.Additional scenes were cap-tured without ground truth(see supplementary material for full details)to be used for qualitative evaluation,so that our algorithm was tested on a total of4different scenes.4.1.Quantitative ResultsTo generate quantitative results,a forest was trained on the stairs scene and tested on the room scene.125trees were generated with a maximum depth of25,forest accu-racy of0.0001and the labels were generated withα=7.It is standard practice when dealing with noisy point cloud data to delete points which are deemed to have the highest error,so that further processing must cope only with more reliable data.Previous work such as Schuon et al.[29]used amplitude as a per-point quality indicator,delet-ing points with low values.The probabilistic conﬁdence output from our algorithm can be thresholded in the same way to produce an alternative binary classiﬁer for points.During the review process a“geometric”baseline algo-rithm was suggested,which classiﬁed each point by com-puting the distance to the closest other point in each scan, so a small distance would indicate a reliable point.We in-clude comparisons of our method to both this method and the amplitude baseline.To quantify the relative improvement of our method over simply using the amplitude or examining geometric proper-ties of the data,we present ROC curves on room in Fig.5. After discarding points for which no laser geometry orﬁl-ter output was available,31k points were left for classiﬁca-tion.The ground truth in this binary classiﬁcation task was computed by thresholding relative distance error from(1). Fig.5(a)shows the classiﬁers’accuracy at correctly iden-tifying which points are within4%of the true depth,and Fig.5(b)shows the same for25%.ROC curves are gener-ated by sweeping each metric from minimum to maximum value,where the metrics are amplitude,conﬁdence poste-rior probability,and inverse distance to sample point.The average ROC Area Under Curve(AUC)calculated across the error tolerance range{1...25}%(see Table1) shows our technique outperforming both the amplitude and geometric baselines on average,and in fact in each of theseThreshold (relative error,%)Amp AUC Geom AUC Conf AUC 40.6390.5990.729250.6850.7220.757{1...25}(average)0.6540.6610.736Table 1.AUC of ROC curves for Amplitude (Amp),Geometric (Geom)and Conﬁdence (Conf)algorithms.00.20.40.60.81False Positive RateT r u e P o s i t i v e R a t e(a)00.20.40.60.81False Positive RateT r u e P o s i t i v e R a t e(b)Figure 5.ROC parison of our method against ampli-tude and geometric ﬁltering on:(a)less than 4%error with 21,290actual positives and 9,931actual negatives,(b)less than 25%er-ror with 29,102actual positives and 2,119actual positives.4%error corresponds to an error margin of 16cm on a surface 4meters away.25cases,the AUC for conﬁdence thresholding is largest.Further graphs for these settings can be found in the sup-plementary material.Our technique no longer has a clear advantage over the baselines for tolerances above 25%,but very few bad points remain.Over 93%of the data already falls within this tolerance.4.2.Qualitative ResultsIt is informative to examine which points are deleted by each ToF assessment algorithm.We compare ampli-tude thresholding,the thresholding technique of Swadzba et al .[34],and conﬁdence thresholding,referred to as meth-ods A,B,and C (Fig.6).Method B performs amplitude thresholding after smoothing the amplitude image,and ad-ditionally discards points based on edge detection.It is un-clear how to vary both the amplitude and edge parameters to draw a fair ROC.As our technique is only being used to threshold points and not actually change their positions,we omit their distance smoothing step.Using the recommended parameters,method B removes 27%of the 35,344points in room .Thresholds for meth-ods A and C were adjusted to remove the same number of points for direct comparison,although we would use a lower threshold on method C to improve results.As the Fig.6insets show,methods A and B are not as comprehen-sive in their removal of ﬂying pixels at depth discontinuities.Please see the supplementary material to better inspect the3D data.4.3.Feature ImportanceI m p o r t a n c eFigure 7.The 25most important Random Forest features after training on stairs and room data.The Random Forest training process calculates the rele-vance of each entry of the feature vector,which allows us to analyze which features carry the most information as shown in Fig.7.A full graph is included in the supplementary ma-terial.Distance is the most important factor,followed by in-tensity,the Laplacian of the distance image at two different scales,and ﬁnally the amplitude.For this analysis,the for-est has been trained on both stairs and room ,which al-lows the variation in the global features to contribute to the conﬁdence model.It is shown that the global features based on amplitude contain the most discriminative content.Note that when training on only one scene,all the global features are constant across the whole training set and so have the same level of importance.4.4.Application to Different Model ToFTo evaluate the effectiveness of our technique on a differ-ent model of ToF camera,we also collected training data for the Mesa SwissRanger SR-3100.This camera does not pro-vide an amplitude image,only intensity and distance,so the feature vector was correspondingly smaller.Fig.8shows a photo of a test scene which included large depth disconti-nuities and many ﬂying pixels.The thresholded point cloud once more shows successful elimination of ﬂying pixels and other forms of outliers.5.Discussion and Future WorkWe have presented a method to assign per-point conﬁ-dences to ToF scans using a supervised learning approach.Both qualitative and quantitative results show marked im-provement on contemporary methods for the removal of in-accurate points.The algorithm is particularly good at de-(a)Amplitude (methodA)(b)Swadzba et al .[34](methodB)(c)Conﬁdence (method C)Figure 6.Points removed by a method are shown in red.The green box in (c)shows an artifact of our method,see §5.(a)(b)(c)Figure 8.Cluttered desk environment captured by Mesa SwissRanger 3100.(a)photo of the scene,(b)and (c)are respective frontal and side views of the obtained point cloud where points with conﬁdence below 0.6are shown in red.tecting ﬂying pixels.Demonstrations using different cam-era models suggest that our conﬁdence assignment method is hardware agnostic.We anticipate methods such as Schuon et al .[29]could be improved with our conﬁdences.ICP is a common algorithm for merging point clouds,and could be augmented by applying conﬁdence weighting to each matching pair of points.Our conﬁdence measure could potentially be adapated to support the semantic cues used for SfM in Bao and Savarese [1].Similarly,the initial mesh models reconstructed using multi-view stereo and then re-ﬁned based on shading in Wu et al .[35]could instead start from our ToF data,and be more malleable where the conﬁ-dence was low.As with all supervised learning methods,for a high level of performance,a representative training set is required.Despite the limited size and variety in our training set,the technique has proven successful on unseen test data.Some artifacts remain,such as the incorrectly deleted points in the green square in Fig.6(c).We surmise that adding data from a wider range of scenes,including bright nearby ob-jects,would further improve generalisation ability and per-formance.The registration problems detailed in §3.2couldbe solved either with higher-resolution ToF cameras or a better distribution of targets within the scene.Applying a non-linear calibration for the systematic depth [3,8,11]could improve the raw distance readings which would then be used to compute feature vectors.An area we have not yet explored is the possibility of using a variant of the standard Random Forest or indeed a different Machine Learning algorithm altogether.Recent developments such as Adaptive Random Forests [31]have the potential to increase the speed of the conﬁdence assign-ment,which is currently around 5seconds for a 200×200frame.AcknowledgementsWe are grateful to Jan Boehm and Stuart Robson,for sharing their laser scanner and expertise,and to Marc Pollefeys for lending us the SR-3100.Thanks to Maciej Gryka,Oisin Mac Aodha,and Fr´e d´e ric Besse who helped during data collection,and to Jim Rehg for valuable discussions.We would also like to thank the reviewers for their feedback and suggestions.The student authors were sup-ported by the UK EPSRC-funded Eng.Doctorate Centre in Virtual Environments,Imaging and Visualisation (EP/G037159/1).。

清末民初服饰变迁

清末民初服饰变化影响
• 清末民初，随着社会的大变革与大发展，服饰变化随之而来。中国的服饰开始呈现多元化，男女服饰样式繁多，中西服装交相辉映，一时间百花齐放，在中国服饰史上形成了一个重要的转折时期。这对社会制度与经济文化产生了深远影响，使中国服装走向西式化道路。
ห้องสมุดไป่ตู้
谢
谢
• 长袍：袍长过膝，领子是圆的，带大襟有扣襻，袖子肥瘦适中，下摆留有开禊。
主题秀之穿在身上的历史
• 无袖外褂：坎肩，它是由汉族穿的“半臂”逐渐演变而来，纽扣较多
主题秀之穿在身上的历史
• 清末女子：旗装，偏襟大衫 • 讲究装饰的妇女，每每在长袍的衣襟、领口、袖边处镶上绦子花边，作为美的点缀。
主题秀：
清末男子服饰变化：
• 长装是中国传统装式 • 满族长袍，汉族上衣下裙或上衣下裤 • 晚清男子经典装束为长袍马褂，有长发辫马褂穿在长袍外边，这样便于骑射。马褂又分为半袖、短袖和无袖几种，根据官品等级和家庭状况分不同的布料制成，等级最高的是黄马褂。
主题秀之穿在身上的历史
• 马褂：衣长及脐，袖长及肘，四面开衩；袖口装有剪袖
清末民初服饰变化特点
• 从保守到开放，更注重个性解放长久以来，服装一直被看作遮羞保暖御寒的工具，很少注重美感，加之礼教的束缚更不用说本身的形体特征。清末民初，受新思潮的影响，人们思想观念逐渐文明开化，服饰开始产生新的变化，这种变化主要体现在逐渐开放和注重个性解放上。男子脱掉老气的长袍马褂，穿上笔挺的西装或制服。西服更能显示出男性的阳刚之美，意气风发，紧凑灵活；旗袍洋装则更好的勾勒出除了女性的自然美。
清末民初服饰变化特点
• 造型款式上从平面走向立体
• 穿着方式从传统中式走向中西结合辛亥革命理论和其推崇的社会制度模式都是直接源于欧美国家，效仿西方成为一时风气。当时，许多大城市兴起一股西化思潮，服饰崇洋风气便受到其推动。1912年初，民国政府制定新礼服方案，一些人倾向以西服为定式，这对服饰的趋洋风气不无推波助澜。

清末民初大事记

背景：1840年6月——1842年8月【道光二十年庚子——道光二十二年壬寅】鸦片战争，8月29日中英《南京条约》【道光、林则徐、关天培、维多利亚女王（1837——1901）】1844年7月3日中美《望厦条约》、10月24日中法《黄埔条约》【道光二十四年甲辰道光、耆英、维多利亚女王】1851年——1855年1850年夏太平天国运动兴起，1851年1月11金田起义，1853年3月19日太平军攻克江宁（南京），随后开始北伐（5月8日）【失败】、西征（6月3日）。

1854年美国与琉球政府签订开放那霸港口的条约。

1855年初翼王石达开破湘军，复陷武昌。

【咸丰元年辛亥——咸丰五年乙卯咸丰、曾国藩（湘军）、钦差大臣向荣琦善、洪秀全、杨秀清、萧朝贵、冯云山、韦昌辉、石达开、林凤祥李开芳（北伐）】【国1856年——1860年【咸丰六年丙辰——咸丰十年庚申】内】1856年4月5日太平军再克扬州，攻破“江北大营”，6月太平军攻破“江南大营“，解天京三年之围；9月太平军内乱——天京事变，三王（东王杨秀清、北王韦昌辉、燕王秦日纲）被杀，一王（翼王石达开）远走转战（1857年5月）1858年清军重建江北、江南大营，4月洪仁轩辗转至天京被封为干王。

1860年2月-5月李秀成第二次攻破江南大营，9月太平军部署西征。

【对外】第二次鸦片战争（1856年10月-1860年11月）1857年12月英人攻占广州，掳两广总督叶名琛。

1858年4月英、法、俄、美四国公使率舰队至大沽口外，5月20日英法联军炮轰大沽，26日沿白河北上至天津城郊，扬言进攻北京。

6月13日清政府与英、法、俄签订《天津条约》，与美国则是18日，签约后英法联军撤退南下。

1859年6月25日英法联军再度进攻大沽，于58年英法联军撤兵后接替大沽防务的科尔沁亲王僧格林沁指挥得当击退英法联军。

1860年4月英法联军占领舟山，5-6月英人占大连湾，法人占烟台，封锁渤海湾，得窥大沽口。

高考历史清末民初知识点

高考历史清末民初知识点清末民初是中国历史上一个重要的时期，这段时期对于中国近代史的发展具有重要的影响。

考生在备战高考历史科目时，要熟悉清末民初的知识点，掌握该时期的重要事件、人物和历史背景。

本文将对清末民初的知识点进行详细讲解。

一、清朝的衰落在清朝晚期，由于内外因素的影响，许多社会问题逐渐显现出来，导致了清朝的衰落。

其中，甲午战争和八国联军侵华战争是中国在近代历史上遭受的两场重大战争，给清朝带来了巨大的冲击。

此外，义和团运动和戊戌变法也是清朝衰落的重要原因。

二、辛亥革命与中华民国的建立辛亥革命是清末民初最重要的政治事件之一。

辛亥革命的爆发标志着封建君主专制制度的结束，中华民国在南京宣告成立，结束了中国两千多年的封建帝制。

辛亥革命后，中国出现了各个地方军阀割据、政治动荡的情况。

三、战争与外国列强的干涉在清朝的末年，中国遭受了外国列强的入侵和干涉。

在甲午战争和八国联军侵华战争中，清朝无力对抗外敌，丧失了对外的主权。

随着列强势力的扩张，中国沦为列强的半殖民地。

四、改革与变革的尝试戊戌变法是中国近代历史上的一次改革尝试，标志着近代中国社会觉醒的开始。

戊戌变法提出了一系列变革措施，如改革教育、建立宪政制度等，但由于保守势力的阻碍和内外因素的影响，变法未能成功。

五、新文化运动与思想解放新文化运动是一场以反对封建文化为核心的思想解放运动。

新文化运动倡导科学、民主、自由，对中国传统文化进行了批判，推动了思想解放和现代化进程。

六、五四运动与中国共产党的崛起五四运动是中国近代史上具有重要影响的一次爱国运动。

五四运动促进了中国社会的思想解放和文化进步，为中国共产党的崛起提供了条件。

七、清末民初社会变革清末民初是中国社会变革的重要时期。

在这一时期，中国社会发生了巨大的变革，如中国农民解放运动、妇女解放运动等，为中国近代化进程奠定了基础。

八、清末民初文化艺术的发展清末民初是中国文化艺术发展的重要时期。

在这段时期，中国文人墨客涌现，文化艺术在政治、社会等方面均取得了突破性的进展。

#老上海影集#之静安寺路（2）

#老上海影集#之静安寺路（2）老上海影集#之静安寺路（2）11:愚园在静安寺路西首，赫德路（今常德路）8号，建于1890年，原是私家花园，也向公众开放，但收门票两角。

园内有假山、亭阁，是中西合璧的园林(图1、2)。

园林五易其主，曾以“夜公园”招来各路豪客。

1917年废弃。

当常州人刘葆良拥有愚园时，因没什么西洋名物可与张园竞争，曾蓄养猩猩、孔雀等动物吸引游客，但收效甚微。

后来，园主人不顾租界当局有关公共场所必须晚上12点前谢客关门的规定，园门洞开，开“夜公园”之先例。

“夜公园”曾一度在沪西流行，被道学先生斥为沪上陋风之最。

愚园路因愚园而得名。

图3为愚园内茶室。

12:1908年，静安寺路开通了有轨电车，马路蜕变成道路，从此加快了西区繁荣的进程。

图1:有轨电车驶过静安寺门前；图2:人力车在静安寺路上飞奔。

13:这两张是清末民初静安寺门前街景老照片，很是罕见。

现分享诸位朋友，让大家领略一下当年静安寺门前悠然惬意的安静。

14:英美租界工部局于清光绪二十四年(1898年)在静安寺对面（南京西路1649号）辟建静安寺公墓(又称外国坟山)。

公墓北靠南京西路，南接延安中路，东近常德路，西邻华山路，因静安古寺而得名。

1954年改建成静安公园。

图1:静安寺公墓大理石亭；图2:静安寺公墓的小教堂，1978年被拆除。

15:1929年，原戈登路（今江宁路）兼营舞厅“大华饭店”停业，以致上海西区没有一个与“贵族区”相称的娱乐场。

1931年，华商顾联承投资70万两白银，购置静安寺西北处(愚园路218号，图1)地皮建百乐门舞厅，次年建成。

百乐门全称是“百乐门大饭店舞厅”。

“百乐门”是英文“Ｐａｒａｍｏｕｎｔ”译音，英文原义是“卓越”和“千变万化”的意思。

因建筑结构别致，舞厅装饰考究，当时有“远东第一乐府”之称。

百乐门由中国建筑师杨锡镠设计，陆根记营造厂承建。

建筑风格与大光明、大上海电影院有相似之处。

解放后，原舞厅主体建筑改为红都戏院，其他附属建筑则改建为商场。

1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。