Phylogenetic inferences in Prunus (Rosaceae) using chloroplast ndhF and nuclear ribosomal ITS

合集下载

人工智能领域中英文专有名词汇总

名词解释中英文对比<using_information_sources> social networks 社会网络abductive reasoning 溯因推理action recognition(行为识别)active learning(主动学习)adaptive systems 自适应系统adverse drugs reactions(药物不良反应)algorithm design and analysis(算法设计与分析) algorithm(算法)artificial intelligence 人工智能association rule(关联规则)attribute value taxonomy 属性分类规范automomous agent 自动代理automomous systems 自动系统background knowledge 背景知识bayes methods(贝叶斯方法)bayesian inference(贝叶斯推断)bayesian methods(bayes 方法)belief propagation(置信传播)better understanding 内涵理解big data 大数据big data(大数据)biological network(生物网络)biological sciences(生物科学)biomedical domain 生物医学领域biomedical research(生物医学研究)biomedical text(生物医学文本)boltzmann machine(玻尔兹曼机)bootstrapping method 拔靴法case based reasoning 实例推理causual models 因果模型citation matching (引文匹配)classification (分类)classification algorithms(分类算法)clistering algorithms 聚类算法cloud computing(云计算)cluster-based retrieval (聚类检索)clustering (聚类)clustering algorithms(聚类算法)clustering 聚类cognitive science 认知科学collaborative filtering (协同过滤)collaborative filtering(协同过滤)collabrative ontology development 联合本体开发collabrative ontology engineering 联合本体工程commonsense knowledge 常识communication networks(通讯网络)community detection(社区发现)complex data(复杂数据)complex dynamical networks(复杂动态网络)complex network(复杂网络)complex network(复杂网络)computational biology 计算生物学computational biology(计算生物学)computational complexity(计算复杂性) computational intelligence 智能计算computational modeling(计算模型)computer animation(计算机动画)computer networks(计算机网络)computer science 计算机科学concept clustering 概念聚类concept formation 概念形成concept learning 概念学习concept map 概念图concept model 概念模型concept modelling 概念模型conceptual model 概念模型conditional random field(条件随机场模型) conjunctive quries 合取查询constrained least squares (约束最小二乘) convex programming(凸规划)convolutional neural networks(卷积神经网络) customer relationship management(客户关系管理) data analysis(数据分析)data analysis(数据分析)data center(数据中心)data clustering (数据聚类)data compression(数据压缩)data envelopment analysis (数据包络分析)data fusion 数据融合data generation(数据生成)data handling(数据处理)data hierarchy (数据层次)data integration(数据整合)data integrity 数据完整性data intensive computing(数据密集型计算)data management 数据管理data management(数据管理)data management(数据管理)data miningdata mining 数据挖掘data model 数据模型data models(数据模型)data partitioning 数据划分data point(数据点)data privacy(数据隐私)data security(数据安全)data stream(数据流)data streams(数据流)data structure( 数据结构)data structure(数据结构)data visualisation(数据可视化)data visualization 数据可视化data visualization(数据可视化)data warehouse(数据仓库)data warehouses(数据仓库)data warehousing(数据仓库)database management systems(数据库管理系统)database management(数据库管理)date interlinking 日期互联date linking 日期链接Decision analysis(决策分析)decision maker 决策者decision making (决策)decision models 决策模型decision models 决策模型decision rule 决策规则decision support system 决策支持系统decision support systems (决策支持系统) decision tree(决策树)decission tree 决策树deep belief network(深度信念网络)deep learning(深度学习)defult reasoning 默认推理density estimation(密度估计)design methodology 设计方法论dimension reduction(降维) dimensionality reduction(降维)directed graph(有向图)disaster management 灾害管理disastrous event(灾难性事件)discovery(知识发现)dissimilarity (相异性)distributed databases 分布式数据库distributed databases(分布式数据库) distributed query 分布式查询document clustering (文档聚类)domain experts 领域专家domain knowledge 领域知识domain specific language 领域专用语言dynamic databases(动态数据库)dynamic logic 动态逻辑dynamic network(动态网络)dynamic system(动态系统)earth mover's distance(EMD 距离) education 教育efficient algorithm(有效算法)electric commerce 电子商务electronic health records(电子健康档案) entity disambiguation 实体消歧entity recognition 实体识别entity recognition(实体识别)entity resolution 实体解析event detection 事件检测event detection(事件检测)event extraction 事件抽取event identificaton 事件识别exhaustive indexing 完整索引expert system 专家系统expert systems(专家系统)explanation based learning 解释学习factor graph(因子图)feature extraction 特征提取feature extraction(特征提取)feature extraction(特征提取)feature selection (特征选择)feature selection 特征选择feature selection(特征选择)feature space 特征空间first order logic 一阶逻辑formal logic 形式逻辑formal meaning prepresentation 形式意义表示formal semantics 形式语义formal specification 形式描述frame based system 框为本的系统frequent itemsets(频繁项目集)frequent pattern(频繁模式)fuzzy clustering (模糊聚类)fuzzy clustering (模糊聚类)fuzzy clustering (模糊聚类)fuzzy data mining(模糊数据挖掘)fuzzy logic 模糊逻辑fuzzy set theory(模糊集合论)fuzzy set(模糊集)fuzzy sets 模糊集合fuzzy systems 模糊系统gaussian processes(高斯过程)gene expression data 基因表达数据gene expression(基因表达)generative model(生成模型)generative model(生成模型)genetic algorithm 遗传算法genome wide association study(全基因组关联分析) graph classification(图分类)graph classification(图分类)graph clustering(图聚类)graph data(图数据)graph data(图形数据)graph database 图数据库graph database(图数据库)graph mining(图挖掘)graph mining(图挖掘)graph partitioning 图划分graph query 图查询graph structure(图结构)graph theory(图论)graph theory(图论)graph theory(图论)graph theroy 图论graph visualization(图形可视化)graphical user interface 图形用户界面graphical user interfaces(图形用户界面)health care 卫生保健health care(卫生保健)heterogeneous data source 异构数据源heterogeneous data(异构数据)heterogeneous database 异构数据库heterogeneous information network(异构信息网络) heterogeneous network(异构网络)heterogenous ontology 异构本体heuristic rule 启发式规则hidden markov model(隐马尔可夫模型)hidden markov model(隐马尔可夫模型)hidden markov models(隐马尔可夫模型) hierarchical clustering (层次聚类) homogeneous network(同构网络)human centered computing 人机交互技术human computer interaction 人机交互human interaction 人机交互human robot interaction 人机交互image classification(图像分类)image clustering (图像聚类)image mining( 图像挖掘)image reconstruction(图像重建)image retrieval (图像检索)image segmentation(图像分割)inconsistent ontology 本体不一致incremental learning(增量学习)inductive learning (归纳学习)inference mechanisms 推理机制inference mechanisms(推理机制)inference rule 推理规则information cascades(信息追随)information diffusion(信息扩散)information extraction 信息提取information filtering(信息过滤)information filtering(信息过滤)information integration(信息集成)information network analysis(信息网络分析) information network mining(信息网络挖掘) information network(信息网络)information processing 信息处理information processing 信息处理information resource management (信息资源管理) information retrieval models(信息检索模型) information retrieval 信息检索information retrieval(信息检索)information retrieval(信息检索)information science 情报科学information sources 信息源information system( 信息系统)information system(信息系统)information technology(信息技术)information visualization(信息可视化)instance matching 实例匹配intelligent assistant 智能辅助intelligent systems 智能系统interaction network(交互网络)interactive visualization(交互式可视化)kernel function(核函数)kernel operator (核算子)keyword search(关键字检索)knowledege reuse 知识再利用knowledgeknowledgeknowledge acquisitionknowledge base 知识库knowledge based system 知识系统knowledge building 知识建构knowledge capture 知识获取knowledge construction 知识建构knowledge discovery(知识发现)knowledge extraction 知识提取knowledge fusion 知识融合knowledge integrationknowledge management systems 知识管理系统knowledge management 知识管理knowledge management(知识管理)knowledge model 知识模型knowledge reasoningknowledge representationknowledge representation(知识表达) knowledge sharing 知识共享knowledge storageknowledge technology 知识技术knowledge verification 知识验证language model(语言模型)language modeling approach(语言模型方法) large graph(大图)large graph(大图)learning(无监督学习)life science 生命科学linear programming(线性规划)link analysis (链接分析)link prediction(链接预测)link prediction(链接预测)link prediction(链接预测)linked data(关联数据)location based service(基于位置的服务) loclation based services(基于位置的服务) logic programming 逻辑编程logical implication 逻辑蕴涵logistic regression(logistic 回归)machine learning 机器学习machine translation(机器翻译)management system(管理系统)management( 知识管理)manifold learning(流形学习)markov chains 马尔可夫链markov processes(马尔可夫过程)matching function 匹配函数matrix decomposition(矩阵分解)matrix decomposition(矩阵分解)maximum likelihood estimation(最大似然估计)medical research(医学研究)mixture of gaussians(混合高斯模型)mobile computing(移动计算)multi agnet systems 多智能体系统multiagent systems 多智能体系统multimedia 多媒体natural language processing 自然语言处理natural language processing(自然语言处理) nearest neighbor (近邻)network analysis( 网络分析)network analysis(网络分析)network analysis(网络分析)network formation(组网)network structure(网络结构)network theory(网络理论)network topology(网络拓扑)network visualization(网络可视化)neural network(神经网络)neural networks (神经网络)neural networks(神经网络)nonlinear dynamics(非线性动力学)nonmonotonic reasoning 非单调推理nonnegative matrix factorization (非负矩阵分解) nonnegative matrix factorization(非负矩阵分解) object detection(目标检测)object oriented 面向对象object recognition(目标识别)object recognition(目标识别)online community(网络社区)online social network(在线社交网络)online social networks(在线社交网络)ontology alignment 本体映射ontology development 本体开发ontology engineering 本体工程ontology evolution 本体演化ontology extraction 本体抽取ontology interoperablity 互用性本体ontology language 本体语言ontology mapping 本体映射ontology matching 本体匹配ontology versioning 本体版本ontology 本体论open government data 政府公开数据opinion analysis(舆情分析)opinion mining(意见挖掘)opinion mining(意见挖掘)outlier detection(孤立点检测)parallel processing(并行处理)patient care(病人医疗护理)pattern classification(模式分类)pattern matching(模式匹配)pattern mining(模式挖掘)pattern recognition 模式识别pattern recognition(模式识别)pattern recognition(模式识别)personal data(个人数据)prediction algorithms(预测算法)predictive model 预测模型predictive models(预测模型)privacy preservation(隐私保护)probabilistic logic(概率逻辑)probabilistic logic(概率逻辑)probabilistic model(概率模型)probabilistic model(概率模型)probability distribution(概率分布)probability distribution(概率分布)project management(项目管理)pruning technique(修剪技术)quality management 质量管理query expansion(查询扩展)query language 查询语言query language(查询语言)query processing(查询处理)query rewrite 查询重写question answering system 问答系统random forest(随机森林)random graph(随机图)random processes(随机过程)random walk(随机游走)range query(范围查询)RDF database 资源描述框架数据库RDF query 资源描述框架查询RDF repository 资源描述框架存储库RDF storge 资源描述框架存储real time(实时)recommender system(推荐系统)recommender system(推荐系统)recommender systems 推荐系统recommender systems(推荐系统)record linkage 记录链接recurrent neural network(递归神经网络) regression(回归)reinforcement learning 强化学习reinforcement learning(强化学习)relation extraction 关系抽取relational database 关系数据库relational learning 关系学习relevance feedback (相关反馈)resource description framework 资源描述框架restricted boltzmann machines(受限玻尔兹曼机) retrieval models(检索模型)rough set theroy 粗糙集理论rough set 粗糙集rule based system 基于规则系统rule based 基于规则rule induction (规则归纳)rule learning (规则学习)rule learning 规则学习schema mapping 模式映射schema matching 模式匹配scientific domain 科学域search problems(搜索问题)semantic (web) technology 语义技术semantic analysis 语义分析semantic annotation 语义标注semantic computing 语义计算semantic integration 语义集成semantic interpretation 语义解释semantic model 语义模型semantic network 语义网络semantic relatedness 语义相关性semantic relation learning 语义关系学习semantic search 语义检索semantic similarity 语义相似度semantic similarity(语义相似度)semantic web rule language 语义网规则语言semantic web 语义网semantic web(语义网)semantic workflow 语义工作流semi supervised learning(半监督学习)sensor data(传感器数据)sensor networks(传感器网络)sentiment analysis(情感分析)sentiment analysis(情感分析)sequential pattern(序列模式)service oriented architecture 面向服务的体系结构shortest path(最短路径)similar kernel function(相似核函数)similarity measure(相似性度量)similarity relationship (相似关系)similarity search(相似搜索)similarity(相似性)situation aware 情境感知social behavior(社交行为)social influence(社会影响)social interaction(社交互动)social interaction(社交互动)social learning(社会学习)social life networks(社交生活网络)social machine 社交机器social media(社交媒体)social media(社交媒体)social media(社交媒体)social network analysis 社会网络分析social network analysis(社交网络分析)social network(社交网络)social network(社交网络)social science(社会科学)social tagging system(社交标签系统)social tagging(社交标签)social web(社交网页)sparse coding(稀疏编码)sparse matrices(稀疏矩阵)sparse representation(稀疏表示)spatial database(空间数据库)spatial reasoning 空间推理statistical analysis(统计分析)statistical model 统计模型string matching(串匹配)structural risk minimization (结构风险最小化) structured data 结构化数据subgraph matching 子图匹配subspace clustering(子空间聚类)supervised learning( 有support vector machine 支持向量机support vector machines(支持向量机)system dynamics(系统动力学)tag recommendation(标签推荐)taxonmy induction 感应规范temporal logic 时态逻辑temporal reasoning 时序推理text analysis(文本分析)text anaylsis 文本分析text classification (文本分类)text data(文本数据)text mining technique(文本挖掘技术)text mining 文本挖掘text mining(文本挖掘)text summarization(文本摘要)thesaurus alignment 同义对齐time frequency analysis(时频分析)time series analysis( 时time series data(时间序列数据)time series data(时间序列数据)time series(时间序列)topic model(主题模型)topic modeling(主题模型)transfer learning 迁移学习triple store 三元组存储uncertainty reasoning 不精确推理undirected graph(无向图)unified modeling language 统一建模语言unsupervisedupper bound(上界)user behavior(用户行为)user generated content(用户生成内容)utility mining(效用挖掘)visual analytics(可视化分析)visual content(视觉内容)visual representation(视觉表征)visualisation(可视化)visualization technique(可视化技术) visualization tool(可视化工具)web 2.0(网络2.0)web forum(web 论坛)web mining(网络挖掘)web of data 数据网web ontology lanuage 网络本体语言web pages(web 页面)web resource 网络资源web science 万维科学web search (网络检索)web usage mining(web 使用挖掘)wireless networks 无线网络world knowledge 世界知识world wide web 万维网world wide web(万维网)xml database 可扩展标志语言数据库附录 2 Data Mining 知识图谱（共包含二级节点15 个，三级节点93 个）间序列分析)监督学习)领域二级分类三级分类。

琥珀推测的内容和推测的依据英语作文

琥珀推测的内容和推测的依据英语作文全文共3篇示例，供读者参考篇1The Amber Hypothesis: Unraveling the Mystery of the K-T Mass ExtinctionAs a student fascinated by the wonders of our planet's history, I have been captivated by the enigmatic tale of the Cretaceous-Tertiary (K-T) mass extinction event, which occurred approximately 66 million years ago. This cataclysmic episode wiped out a staggering 75% of all plant and animal species on Earth, including the mighty non-avian dinosaurs that had dominated the terrestrial realm for over 160 million years. Among the myriad theories proposed to explain this profound upheaval, one stands out as a beacon of scientific ingenuity and perseverance: the Amber Hypothesis.The Amber Hypothesis, first proposed by paleontologists David Grimaldi and Doreen Daly in the early 1990s, offers a compelling narrative rooted in the intricate analysis of fossilized tree resin, better known as amber. This remarkable substance, renowned for its ability to preserve ancient life forms withexquisite detail, has proven to be an invaluable time capsule, unlocking secrets from the distant past.At the heart of the Amber Hypothesis lies a tantalizing clue embedded within the amber deposits found in various locations worldwide, including the renowned fossil beds of Messel, Germany, and the Cerrejón region of Colombia. These amber specimens, dating back to the late Cretaceous period, contain an astonishing array of encapsulated organisms, from minute insects and arachnids to plant remnants and even microscopic fungi.One of the most striking observations made by Grimaldi and Daly was the presence of certain types of fungi within the amber samples. These fungi exhibited characteristics strikingly similar to those of modern-day species that are known to thrive in environments rich in decaying organic matter and possessing a remarkable ability to digest complex plant materials. This remarkable discovery led the researchers to postulate a profound shift in the global ecosystem during the late Cretaceous, fueled by an unprecedented proliferation of these specialized fungi.According to the Amber Hypothesis, the rapid diversification and dominance of these fungal species may have been triggeredby a series of cataclysmic events, such as widespread volcanic activity, asteroid impacts, or a combination of both. These disturbances could have caused massive environmental changes, including alterations in atmospheric composition, climate patterns, and the destruction of vast swaths of vegetation.As the once-lush forests and plant life succumbed to these catastrophic events, a wealth of decaying organic matter would have been left in their wake. This abundance of dead and decomposing plant material provided the perfect breeding ground for the specialized fungi encapsulated in the amber. With their unique ability to break down complex plant tissues, these fungi could have proliferated rapidly, consuming the available biomass at an unprecedented rate.The implications of this fungal dominance were far-reaching and potentially devastating. As the fungi relentlessly consumed the decaying vegetation, they would have released vast quantities of carbon dioxide and other greenhouse gases into the atmosphere, exacerbating the already-perturbed climate conditions. This feedback loop could have triggered a cascading effect, leading to further environmental disruptions and ultimately contributing to the mass extinction of numerous species unable to adapt to the rapidly changing conditions.The Amber Hypothesis elegantly ties together various strands of evidence, including geological data, paleobotanical records, and the remarkable fossil record preserved within amber itself. The presence of these specialized fungi in late Cretaceous amber deposits from disparate locations around the globe lends credence to the hypothesis, suggesting a global phenomenon rather than a localized event.Moreover, the hypothesis aligns with other lines of evidence, such as the presence of soot and ash layers in the geological record, indicative of widespread wildfires and volcanic activity during the late Cretaceous. These events could have served as the catalyst for the initial environmental disturbances that set the stage for the fungal proliferation described in the Amber Hypothesis.While the Amber Hypothesis has gained significant traction within the scientific community, it is important to acknowledge that it remains a hypothesis, subject to ongoing research, scrutiny, and potential refinement or revision. Science thrives on the continuous interplay between observation, hypothesis formulation, and rigorous testing, as new evidence emerges and our understanding of the natural world deepens.Nonetheless, the Amber Hypothesis stands as a testament to the ingenuity and perseverance of scientists in unraveling the mysteries of Earth's past. By harnessing the remarkable preservation capabilities of amber and combining insights from diverse fields, researchers like Grimaldi and Daly have shed light on a pivotal moment in our planet's history, offering a compelling narrative that resonates with both scientific rigor and human curiosity.As a student captivated by the wonders of paleontology and Earth's history, I find the Amber Hypothesis a fascinating and thought-provoking concept. It reminds us that the intricate tapestry of life on our planet is woven with threads of complexity, interconnectedness, and resilience, even in the face of cataclysmic events. The K-T mass extinction may have spelled the end for the non-avian dinosaurs, but it also paved the way for the rise of new species and the eventual emergence of our own species, Homo sapiens.The Amber Hypothesis serves as a poignant reminder of the dynamic nature of our planet and the delicate balance that sustains life. As we navigate the challenges of our modern era, with environmental concerns and the looming threat of climate change, the lessons gleaned from Earth's past mass extinctionsbecome increasingly relevant. By studying and understanding these pivotal events, we may gain valuable insights into the intricate web of life and the resilience required to navigate the ever-changing landscapes of our planet's future.篇2The Amber Hypothesis: Unraveling the Mystery of Ancient Life Preserved in Golden ResinAs a student fascinated by the marvels of the natural world, I find the amber hypothesis to be one of the most intriguing and captivating theories in the study of paleontology. This hypothesis delves into the extraordinary phenomenon of ancient life forms being preserved in exquisite detail within the confines of fossilized tree resin, offering an unprecedented glimpse into the past.The Content of the Amber HypothesisAt its core, the amber hypothesis proposes that the diverse array of organisms encapsulated within ancient amber provides a remarkable window into the ecosystems and life forms that existed millions of years ago. This fossilized tree resin, formed through the polymerization of sticky plant secretions, has provento be an unparalleled time capsule, preserving intricate details of the creatures it ensnared with astonishing fidelity.The preserved specimens found within amber range from minuscule insects and arachnids to plant matter, feathers, and even small vertebrates. These inclusions offer an extraordinary glimpse into the biodiversity of the distant past, shedding light on the morphological characteristics, behaviors, and evolutionary relationships of these ancient denizens.One of the most remarkable aspects of the amber hypothesis is its ability to capture and preserve organisms in their natural state, frozen in time with exquisite detail. The clarity of these inclusions is truly breathtaking, allowing researchers to study the intricate structures, colors, and even the internal anatomy of these ancient life forms with unprecedented precision.The Basis of the Amber HypothesisThe foundation of the amber hypothesis rests on a multitude of scientific disciplines, each contributing vital pieces to the puzzle of understanding these ancient fossils. Let us delve into the primary pillars that support this captivating theory:Geological Evidence: The formation of amber is a complex process that requires specific environmental conditions. By studying the geological context in which amber deposits are found, researchers can gain insights into the ancient landscapes, climates, and plant communities that existed during that era. This information is crucial for reconstructing the ecosystems in which the entrapped organisms once thrived.Chemical Analysis: Advancements in analytical techniques, such as infrared spectroscopy and gas chromatography-mass spectrometry, have enabled scientists to precisely determine the chemical composition of amber. This analysis not only aids in authenticating the age and origin of the specimens but also reveals valuable information about the environmental conditions prevalent during the resin's formation.Paleobotanical Studies: The study of fossilized plant remains found in association with amber deposits provides invaluable insights into the ancient flora that produced the resin. By examining the anatomical features and taxonomic affinities of these plant fossils, researchers can reconstruct the ancient ecosystems and understand the ecological relationships between the entrapped organisms and their botanical hosts.Comparative Morphology: One of the cornerstones of the amber hypothesis is the meticulous examination and comparison of the morphological features of the preserved organisms. By studying the intricate details of their body structures, researchers can draw inferences about their evolutionary relationships, behaviors, and adaptations to their ancient environments.Molecular Analysis: Recent advancements in molecular techniques have opened up new avenues for exploring the genetic makeup of ancient organisms preserved in amber. Through the extraction and analysis of ancient DNA and biomolecules, scientists can unravel evolutionary histories, reconstruct phylogenetic relationships, and gain insights into the genetic diversity of past ecosystems.The Significance of the Amber HypothesisThe amber hypothesis has profoundly impacted our understanding of the Earth's biological and ecological history. By providing an unprecedented window into the past, these fossilized inclusions have challenged long-held assumptions and reshaped our perspectives on the evolution of life on our planet.One of the most remarkable contributions of the amber hypothesis is the discovery of entirely new lineages of organisms that were previously unknown to science. These remarkable findshave expanded our knowledge of biodiversity and highlighted the vast array of life forms that once inhabited our planet, many of which have no modern-day counterparts.Furthermore, the study of amber inclusions has shed light on ancient ecological interactions, revealing intricate relationships between predators and prey, pollinators and plants, and even symbiotic associations that existed millions of years ago. These insights have provided a deeper appreciation for the complexity and interconnectedness of ancient ecosystems, reminding us of the delicate balance that sustains life on Earth.Beyond its scientific significance, the amber hypothesis has captured the imagination of the public, inspiring countless works of art, literature, and popular culture. The idea of ancient life forms being perfectly preserved in golden resin has captivated audiences worldwide, sparking curiosity and a sense of wonder about the mysteries of our planet's past.As a student, I am constantly in awe of the incredible discoveries and revelations that the amber hypothesis has brought forth. Each new inclusion, each intricate detail preserved in exquisite detail, reminds me of the vastness of our planet's history and the incredible diversity of life that has graced its surface over eons.The amber hypothesis stands as a testament to the relentless pursuit of knowledge and the human spirit's insatiable curiosity. It serves as a reminder that the secrets of our past are not lost forever, but rather, encapsulated and waiting to be uncovered, offering us a glimpse into the wonders that once flourished on our ancient Earth.篇3The Amber Hypothesis: Exploring the Origins of Avian EvolutionAs a student of paleontology, I have always been fascinated by the evolutionary journey of life on Earth. Among the myriad mysteries that captivate my curiosity, the origin of birds and their ascent from the realm of feathered dinosaurs remains a topic of profound interest and ongoing debate. In this essay, I delve into the Amber Hypothesis, a compelling theory that sheds light on the pivotal transition from the Age of Reptiles to the emergence of avian marvels that grace our skies today.The Amber Hypothesis, proposed by paleontologists David Dilcher and Günter Bechly in the early 2000s, postulates that the evolutionary leap from dinosaurs to birds occurred not solely through the fossilized remains entombed in sedimentary rockbut also through the remarkable preservation of organisms trapped in ancient amber. This hypothesis challenges thelong-held notion that our understanding of avian origins is derived exclusively from the fossil record of skeletal remains.At the heart of the Amber Hypothesis lies the extraordinary discovery of exquisitely preserved feathers encased in fossilized tree resin, or amber, dating back to the Late Cretaceous period, approximately 100 million years ago. These feathered inclusions, found predominantly in amber deposits from Canada, France, Spain, and Myanmar, offer an unprecedented glimpse into the intricate structures and morphologies of ancient feathers, shedding invaluable light on the evolutionary trajectory of feather development.One of the most remarkable findings that lend credence to the Amber Hypothesis is the discovery of a remarkable feather specimen from the Late Cretaceous amber deposits of Canada. This feather, referred to as "Kernel," exhibits a striking resemblance to the flight feathers of modern birds, with a highly intricate branching pattern and a well-developed rachis (central shaft). The exceptional preservation of this feather has allowed researchers to conduct detailed analyses, revealing insights intothe aerodynamic properties and flight capabilities of the organism that bore it.Further bolstering the Amber Hypothesis are the discoveries of diverse feather types encased in amber, ranging from downy feathers to highly specialized flight feathers. These findings suggest a remarkable diversity of feathered organisms during the Late Cretaceous period, potentially representing various evolutionary stages in the transition from non-avian dinosaurs to the earliest birds.The Amber Hypothesis also finds support in the discovery of amber-preserved plant remains that offer clues about the environmental conditions prevalent during the Late Cretaceous. These fossils provide invaluable insights into the ecosystems in which feathered organisms thrived, shedding light on the ecological pressures and selective forces that may have driven the evolution of feathers and, ultimately, the emergence of powered flight.However, it is important to note that the Amber Hypothesis is not without its critics and challenges. Some paleontologists argue that the amber inclusions, while remarkable, represent only a tiny fraction of the overall fossil record and may not accurately reflect the full scope of avian evolution. Additionally,the interpretation of feather morphologies and their implications for flight capabilities is a subject of ongoing debate, with varying perspectives on the aerodynamic properties and evolutionary stages represented by these ancient feathers.Despite these challenges, the Amber Hypothesis has gained significant traction within the scientific community, prompting further exploration and analysis of amber-preserved specimens. Researchers continue to unearth new amber deposits and employ cutting-edge techniques, such as synchrotron radiation imaging and molecular analyses, to glean deeper insights into the intricate structures and chemical compositions of these ancient feathers.Moreover, the Amber Hypothesis has sparked a renewed interest in interdisciplinary collaboration, bridging the realms of paleontology, evolutionary biology, and amber research. By combining expertise from diverse fields, scientists aim to piece together a more comprehensive understanding of the intricate evolutionary puzzle that unfolded during the Late Cretaceous period, ultimately leading to the emergence of modern birds.As a student immersed in the study of paleontology, I find the Amber Hypothesis captivating and thought-provoking. It challenges us to reevaluate our assumptions and broaden ourperspectives on the origins of avian evolution. While the fossil record remains a vital source of information, the Amber Hypothesis reminds us that nature's secrets are not always confined to the sedimentary layers of rock but can also be preserved in the amber time capsules of the ancient past.The journey to unravel the mysteries of avian evolution is ongoing, and the Amber Hypothesis represents a significant step forward in our understanding. As we continue to explore and analyze the remarkable feathered inclusions encased in amber, we inch closer to unveiling the intricate tapestry of life that has woven the vibrant diversity of birds we witness today. It is a humbling reminder of the immense complexity and resilience of life on our planet, and a testament to the enduring pursuit of knowledge that drives scientific inquiry.。

小波熵在植物电信号识别中的应用研究

２特征向量提取
２１试验数据．
具有良的时频局部化能力，好将小波变换与信息熵相结合可以得到信号的小波熵。假如把小波变换后的系数矩阵看成一个概率分布序列，计算得到该矩阵的熵值就反映了信号概率分布的有序程度，概率分布越接近无序，熵值越大Ｊ其。定义１假设为信号（信号或各层小波变换系原数）为该信号在某组正交基上的第ｉ系数，，项用表示小波熵，Ｈ是一个由每个正交基上的系数则的某种变换叠加起来的值，：即
效果。同时，用小波能量熵对结果进行分析。结果表明，利小波熵比小波系数能量作为特征对植物
电信号的识别更有效。关键词：植物电信号；小波熵；特征提取；Ｐ神经网络Ｂ中图分类号：Ｐ０．Ｔ３１６文献标志码：Ａ
目前，科学家们已从验证植物电波信号的存在
到探知植物电信号的特性和传递机理，发展到当前对植物电信号模型的建立及相关的生理效应研究』。然而，作为一种十分微弱并且复杂的随机信号，物电信号的基本特征和频谱分布尚不明晰，植国内外学者的相关研究也不够深入。今年来，誉为被
收稿日期：０１１—３２１ — ２０
产生信号的高频细节分量，低通滤波器产生信号的低频近似分量。每分解一次信号的采样频率降低一Ｊ倍，近似分量还可以通过高通滤波和低通滤波进一步

桃YABBY转录因子家族的生物信息学分析

山东农业大学学报(自然科学版),2020,51(6):992-997VOL.51NO.62020 Journal of Shandong Agricultural University(Natural Science Edition)doi:10.3969/j.issn.1000-2324.2020.06.002桃YABBY转录因子家族的生物信息学分析韩继红1,刘金凤2,刘慧敏2*1.濮阳市市容环境卫生管理处,河南濮阳4570002.中国林业科学研究院经济林研究开发中心,国家林业和草原局泡桐研究开发中心,经济林种质创新与利用国家林业和草原局重点实验室,河南郑州450003摘要:为了解桃中YABBY家族的成员和特征，本研究基于桃(Prunus persica)最新基因组组装注释数据，首次通过生物信息学手段系统鉴定并分析了YABBY基因家族特征，包括染色体定位、蛋白特征、基因结构、进化关系、保守结构域、启动子区域顺式元件。

结果显示：桃中含有6个YABBY基因，分布在5条染色体上，其推导蛋白序列较短，分子量较小。

家族成员含有4~6个内含子，结构较为复杂。

桃中YABBY基因含有植物YABBY家族所有5个亚类，即CRC、INO、FIL/YAB3、YAB2和YAB5，且均含有保守的锌指结构和YABBY结构域。

除了核心启动元件，桃YABBY基因还含有大量光应答、激素响应等顺式作用元件，表明该家族基因的表达可能会受到环境、激素等的影响。

本研究初步研究了桃基因组中YABBY基因家族的特征，为后期该基因家族在桃中的功能验证提供了理论支撑。

关键词:桃;YABBY;进化分析;转录因子中图法分类号:S662.1文献标识码:A文章编号:1000-2324(2020)06-0992-06 Bioinformatic Analyses of YABBY Transcription Factors in Prunus persicaHAN Ji-hong1,LIU Jin-feng2,LIU Hui-min2*1.Puyang City Appearance and Environmental Sanitation Management Office,Puyang,457000,China2.Non-timber Forest Research and Development Center,Chinese Academy of Forestry;Paulownia Research&Development Center of China,National Forestry and Grassland Administration;Key Laboratory of Non-timber Forest Germplasm Enhancement&Utilization of National Forestry and Grassland Administration,Zhengzhou450003,ChinaAbstract:To understand the members and characteristics of YABBY family in peach,based on the latest genome assembly annotation data of Prunus persica,we systematically identified and analyzed YABBY gene family characteristics,including chromosomal localization,protein characteristics,gene structure,evolutionary relationships,conserved domains,promoter region cis-elements by bioinformatics means for the first time.The results show that P.persica contains6YABBY s, distributed on5chromosomes,its derived protein sequence is short,and molecular weight is small.The YABBY s gene family members contain4~6introns and the structure is relatively complex.The YABBY s in peach contain all five subclasses of plant YABBY family,namely,CRC、INO、FIL/YAB3、YAB2and YAB5,and all contain conserved zinc finger structure and YABBY domain.In addition to the core initiation elements,peach YABBY s also contains a large number of light response,hormone response and other cis-acting elements,indicating that the expression of the family gene may be affected by the environment, hormones,etc.A preliminary study on the characteristics of YABBY gene family in peach genome provides theoretical support for the functional verification of the gene family in P.persica in the later stage.Keywords:Prunus persica;YABBY;phylogenetic analysis;transcription factorYABBY是种子植物特有的一个小基因家族[1]，该基因家族编码的转录因子包括两个保守的结构域：N末端一个锌指结构和C末端一个YABBY结构（螺旋-环-螺旋）[2,3]。

ros点云的预处理方法

ros点云的预处理方法在机器人操作系统（ROS）中，点云数据是一种重要的传感器输入，广泛应用于机器人的环境感知、三维重建等领域。

然而，原始点云数据往往存在噪声、缺失和不均匀分布等问题，这些因素会影响后续处理的精度和效率。

因此，对点云进行预处理至关重要。

本文将详细介绍几种常见的ROS点云预处理方法。

一、滤波去噪滤波去噪是点云预处理的基础步骤，旨在去除原始点云数据中的噪声和异常点。

以下为几种常用的滤波方法：1.高斯滤波：对点云进行高斯滤波，可以平滑噪声，但可能会模糊边缘信息。

2.中值滤波：中值滤波对异常值有很好的抑制作用，适用于去除椒盐噪声。

3.双边滤波：双边滤波可以在去除噪声的同时保持边缘信息，是一种较为常用的滤波方法。

4.Voxel Grid滤波：将点云划分为体素网格，对每个体素内的点进行平均处理，可以降低数据量，提高处理速度。

二、点云补全由于传感器限制和遮挡等原因，原始点云数据往往存在缺失。

点云补全方法可以填补这些缺失，提高数据质量。

1.稀疏重建：利用稀疏重建方法（如ICP算法）对缺失区域进行补全。

2.基于深度学习的点云补全：利用深度学习方法（如PointNet、PointNet++等）对缺失区域进行预测和补全。

三、点云降采样点云数据量通常较大，为了提高处理速度和减少计算量，可以采用降采样方法。

1.最远点采样：选择距离最近的点作为采样点，可以保持点云的几何特征。

2.随机采样：随机选择一定比例的点作为采样点，简单易实现，但可能丢失部分几何信息。

四、点云配准点云配准是将多个点云合并为一个完整点云的过程，可以扩大点云覆盖范围，提高场景理解能力。

1.ICP算法：迭代最近点算法，通过迭代计算两个点云之间的变换矩阵，实现点云配准。

2.基于特征的点云配准：提取点云特征（如点、线、面等），利用特征匹配实现点云配准。

总结：通过对ROS点云进行预处理，可以有效地提高数据质量，为后续处理提供可靠的基础。

改进DeepLabV3＋网络的指针轨迹图像识别

第４３卷第１期２０２４年２月沈㊀阳㊀理㊀工㊀大㊀学㊀学㊀报ＪｏｕｒｎａｌｏｆＳｈｅｎｙａｎｇＬｉｇｏｎｇＵｎｉｖｅｒｓｉｔｙＶｏｌ４３Ｎｏ１Ｆｅｂ２０２４收稿日期:２０２３－０４－１８基金项目:辽宁省教育厅科学研究经费项目(ＬＧ２０２０１４)作者简介:袁帅(１９９９ )ꎬ男ꎬ硕士研究生ꎻ蒋强(１９７４ )ꎬ通信作者ꎬ男ꎬ教授ꎬ研究方向为智能控制理论及算法ꎮ文章编号:１００３－１２５１(２０２４)０１－００５０－０５改进ＤｅｅｐＬａｂＶ３＋网络的指针轨迹图像识别袁㊀帅１ꎬ蒋㊀强１ꎬ饶㊀兵２(１.沈阳理工大学自动化与电气工程学院ꎬ沈阳１１０１５９ꎻ２.沈阳天眼智云智能技术研究院有限公司ꎬ沈阳１１０１７９)摘㊀要:指针式机械记录仪通常用于记录精密设备运输过程中的震动轨迹图像ꎬ为了更好地监测运输过程中车辆颠簸对仪器设备的影响ꎬ提出一种改进ＤｅｅｐＬａｂＶ３＋网络的指针轨迹图像语义分割方法ꎮ首先将骨干网络替换为ＭｏｂｉｌｅＮｅｔＶ３ꎬ实现模型的轻量化ꎻ然后将解码器中４倍上采样替换为２次２倍上采样ꎬ增强图像中像素的连续性ꎬ使预测结果更接近原始图像ꎮ在自制数据集上进行对比实验ꎬ结果表明:改进ＤｅｅｐＬａｂＶ３＋网络的平均交并比(ＭＩｏＵ)达到８５.８４％ꎬ比原始ＤｅｅｐＬａｂＶ３＋网络提高了３.５７％ꎬ单位时间内检测图片数量(ＦＰＳ)提高了３.５８ｓ－１ꎻ改进ＤｅｅｐＬａｂＶ３＋网络在识别精度和速度上具有明显的优势ꎬ可为精密仪器检测提供数据支持ꎮ关㊀键㊀词:改进ＤｅｅｐＬａｂＶ３＋ꎻ语义分割ꎻ轨迹图像识别ꎻ轻量化中图分类号:ＴＰ３９１.４１文献标志码:ＡＤＯＩ:１０.３９６９/ｊ.ｉｓｓｎ.１００３－１２５１.２０２４.０１.００８ＰｏｉｎｔｅｒＴｒａｊｅｃｔｏｒｙＲｅｃｏｇｎｉｔｉｏｎｗｉｔｈＩｍｐｒｏｖｅｄＤｅｅｐＬａｂＶ３＋ＮｅｔｗｏｒｋＹＵＡＮＳｈｕａｉ１ꎬＪＩＡＮＧＱｉａｎｇ１ꎬＲＡＯＢｉｎｇ２(１.ＳｈｅｎｙａｎｇＬｉｇｏｎｇＵｎｉｖｅｒｓｉｔｙꎬＳｈｅｎｙａｎｇ１１０１５９ꎬＣｈｉｎａꎻ２.ＳｈｅｎｙａｎｇＳｋｙＥｙｅＩｎｔｅｌｌｉｇｅｎｔＣｌｏｕｄＴｅｃｈｎｏｌｏｇｙＲｅｓｅａｒｃｈＩｎｓｔｉｔｕｔｅＣｏ.ꎬＬｔｄ.ꎬＳｈｅｎｙａｎｇ１１０１７９ꎬＣｈｉｎａ)Ａｂｓｔｒａｃｔ:Ｐｏｉｎｔｅｒ￣ｔｙｐｅｍｅｃｈａｎｉｃａｌｒｅｃｏｒｄｅｒｓａｒｅｕｓｕａｌｌｙｕｓｅｄｔｏｒｅｃｏｒｄｉｍａｇｅｓｏｆｖｉｂｒａｔｉｏｎｔｒａｊｅｃｔｏ￣ｒｉｅｓｄｕｒｉｎｇｔｈｅｔｒａｎｓｐｏｒｔａｔｉｏｎｏｆｐｒｅｃｉｓｉｏｎｅｑｕｉｐｍｅｎｔ.Ｉｎｏｒｄｅｒｔｏｂｅｔｔｅｒｍｏｎｉｔｏｒｔｈｅｉｍｐａｃｔｏｆｖｅｈｉ￣ｃｌｅｂｕｍｐｓｏｎｉｎｓｔｒｕｍｅｎｔｓａｎｄｅｑｕｉｐｍｅｎｔｄｕｒｉｎｇｔｒａｎｓｐｏｒｔａｔｉｏｎꎬａｓｅｍａｎｔｉｃｓｅｇｍｅｎｔａｔｉｏｎｍｅｔｈｏｄｏｆｐｏｉｎｔｅｒｔｒａｊｅｃｔｏｒｙｉｍａｇｅｓｗｉｔｈｉｍｐｒｏｖｅｄＤｅｅｐＬａｂＶ３＋ｎｅｔｗｏｒｋｉｓｐｒｏｐｏｓｅｄ.ＦｉｒｓｔｌｙꎬｔｈｅｂａｃｋｂｏｎｅｎｅｔｗｏｒｋｉｓｒｅｐｌａｃｅｄｗｉｔｈＭｏｂｉｌｅＮｅｔＶ３ｔｏｒｅａｌｉｚｅｔｈｅｌｉｇｈｔｗｅｉｇｈｔｏｆｔｈｅｍｏｄｅｌ.Ｔｈｅｎｔｈｅ４￣ｆｏｌｄｕｐ￣ｓａｍｐｌｉｎｇｉｎｔｈｅｄｅｃｏｄｅｒｉｓｒｅｐｌａｃｅｄｗｉｔｈ２ｔｉｍｅｓ２￣ｆｏｌｄｕｐｓａｍｐｌｉｎｇｔｏｅｎｈａｎｃｅｔｈｅｃｏｎｔｉｎｕｉｔｙｏｆｐｉｘ￣ｅｌｓｉｎｔｈｅｉｍａｇｅꎬｗｈｉｃｈｍａｋｅｓｔｈｅｐｒｅｄｉｃｔｅｄｒｅｓｕｌｔｓｃｌｏｓｅｒｔｏｔｈｅｏｒｉｇｉｎａｌｉｍａｇｅ.Ｔｈｅｒｅｓｕｌｔｓｓｈｏｗｔｈａｔｔｈｅａｖｅｒａｇｅｉｎｔｅｒｓｅｃｔｉｏｎｒａｔｉｏ(ＭＩｏＵ)ｏｆｔｈｅｉｍｐｒｏｖｅｄＤｅｅｐＬａｂＶ３＋ｎｅｔｗｏｒｋｒｅａｃｈｅｓ８５.８４％ꎬｗｈｉｃｈｉｓ３.５７％ｈｉｇｈｅｒｔｈａｎｔｈａｔｏｆｔｈｅｏｒｉｇｉｎａｌＤｅｅｐＬａｂＶ３＋ｎｅｔｗｏｒｋꎬａｎｄｔｈｅｎｕｍｂｅｒｏｆｄｅｔｅｃｔｅｄｉｍａｇｅｓｐｅｒｕｎｉｔｔｉｍｅ(ＦＰＳ)ｉｎｃｒｅａｓｅｓｂｙ３.５８ｓ－１.ＴｈｅｉｍｐｒｏｖｅｄＤｅｅｐＬａｂＶ３＋ｎｅｔｗｏｒｋｈａｓｏｂｖｉ￣ｏｕｓａｄｖａｎｔａｇｅｓｉｎｒｅｃｏｇｎｉｔｉｏｎａｃｃｕｒａｃｙａｎｄｓｐｅｅｄꎬｗｈｉｃｈｃａｎｐｒｏｖｉｄｅｄａｔａｓｕｐｐｏｒｔｆｏｒｐｒｅｃｉｓｉｏｎｉｎ￣ｓｔｒｕｍｅｎｔｄｅｔｅｃｔｉｏｎ.Ｋｅｙｗｏｒｄｓ:ｉｍｐｒｏｖｅｄＤｅｅｐＬａｂＶ３＋ꎻｓｅｍａｎｔｉｃｓｅｇｍｅｎｔａｔｉｏｎꎻｔｒａｊｅｃｔｏｒｙｉｍａｇｅｒｅｃｏｇｎｉｔｉｏｎꎻｌｉｇｈｔ￣ｗｅｉｇｈｔ㊀㊀在精密仪器运输过程中ꎬ通常采用指针式机械记录仪记录震动轨迹图像数据ꎬ若采用人工检定方法ꎬ检定结果不精准ꎬ工作量巨大且效率很低ꎮ通过机器学习相关技术ꎬ可实现高效率高精度自动检测指针式机械记录仪震动轨迹图像ꎬ极大地减少人工工作量[１]ꎮ２０１４年开始ꎬ谷歌团队推出并发展了具有良好分割性能的ＤｅｅｐＬａｂ系列分割网络ꎬ其中Ｄｅｅ￣ｐＬａｂＶ３＋网络[２]在语义分割领域表现更为突出ꎬ准确度更高ꎮ以ＤｅｅｐＬａｂＶ３＋为基础ꎬ学者们进行了深入研究ꎬ将其应用于诸多领域ꎮ２０１９年ꎬＬｉｕ等[３]通过添加更多的跳跃连接和卷积层来设计ＤｅｅｐＬａｂＶ３＋解码器ꎬ改善了遥感图像中建筑物轮廓的检测结果ꎬ但对细微边界分割效果不理想ꎮ２０２２年ꎬＺｈａｎｇ等[４]在ＤｅｅｐＬａｂＶ３＋中加入一种基于边缘信息的损失函数ꎬ提高了网络对舌边的分离效果ꎬ但对错误分类处理能力不高ꎮ同年ꎬ刘慧等[５]使用轻量化ＭｏｂｉｌｅＮｅｔＶ２作为Ｄｅｅｐ￣ＬａｂＶ３＋骨干网络ꎬ减少了模型参数ꎬ并使用Ｒｅ￣ＬＵ６激活函数减少部署在移动端设备上的精度损失ꎬ但对小像素目标识别效果较差ꎮ２０２３年ꎬ周迅等[６]在ＤｅｅｐＬａｂＶ３＋中使用三点注意力模块提高了对坝面裂缝像素的提取能力ꎬ但存在漏检情况ꎮ本文将ＤｅｅｐＬａｂＶ３＋网络模型应用于指针式机械记录仪轨迹图像的识别ꎮ使用轻量化网络ＭｏｂｉｌｅＮｅｔＶ３替换原骨干网络Ｘｃｅｐｔｉｏｎꎬ实现模型轻量化ꎻ在解码器中使用２个连续的２倍上采样替换原网络中的４倍上采样ꎬ将提取到的特征图逐步放大ꎬ使得还原出的边界更加细化ꎮ１㊀ＤｅｅｐＬａｂＶ３＋网络及其改进１.１㊀ＤｅｅｐＬａｂＶ３＋网络介绍ＤｅｅｐＬａｂＶ３＋网络使用编码器－解码器(Ｅｎ￣ｃｏｄｅｒ￣Ｄｅｃｏｄｅｒ)结构[７－８]ꎬ在提升分割效果的同时关注边界信息ꎮ模型采用Ｘｃｅｐｔｉｏｎ作为骨干网络ꎬ使用空洞空间卷积金字塔池化(ａｔｒｏｕｓｓｐａｔｉａｌｐｙｒａｍｉｄｐｏｏｌｉｎｇꎬＡＳＰＰ)融合特征图多尺度信息ꎬ并将深度卷积和逐点卷积[９]应用于ＡＳＰＰ和Ｅｎ￣ｃｏｄｅｒ模块中ꎬ使网络训练速度更快ꎮＤｅｅｐＬａｂＶ３＋网络结构如图１所示ꎮ图中:Ｃｏｎｖ表示卷积ꎻｒａｔｅ表示膨胀率ꎻＵｐｓａｍｐｌｅ表示上采样ꎻＤＣＮＮ表示深度卷积神经网络ꎻＡｔｒｏｕｓＣｏｎｖ表示空洞卷积ꎻＰｏｏｌｉｎｇ表示池化ꎻＬｏｗ￣ｌｅｖｅｌＦｅａ￣ｔｕｒｅ表示低级特征ꎻＣｏｎｃａｔ表示数据拼接ꎮＤｅｅｐＬａｂＶ３＋网络通过Ｅｎｃｏｄｅｒ结构得到两部分图片特征ꎬ在Ｄｅｃｏｄｅｒ中使用卷积调整通道ꎬ融合两部分特征ꎬ再使用线性插值上采样使得输出层和原图片尺寸一致ꎬ获得预测结果[１０－１１]ꎮ１.２㊀改进的ＤｅｅｐｌａｂＶ３＋网络１.２.１㊀改进ＤｅｅｐＬａｂＶ３＋骨干网络由于本文检测对象是设备运输过程中的指针震动轨迹ꎬ为满足实时检测和移动检测的要求ꎬ需将训练好的模型部署在移动端硬件平台上ꎮ因此ꎬ在进行图像特征提取时要尽量保证全局信息的准确性ꎬ同时简化参数和计算量ꎬ保证识别效率ꎮ为解决上述问题ꎬ可采用轻量化网络模型ꎮＤｅｅｐＬａｂＶ３＋骨干网络为Ｘｃｅｐｔｉｏｎꎬ该网络结构比较复杂ꎬ参数量较多ꎬ消耗大量的显存ꎮ本文对ＤｅｅｐＬａｂＶ３＋结构的骨干网络实现轻量化ꎬ采用ＭｏｂｉｌｅＮｅｔＶ３代替Ｘｃｅｐｔｉｏｎ[１２]ꎬ轻量化网络Ｍｏ￣ｂｉｌｅＮｅｔＶ３的参数量较少ꎬ更易于部署到移动设备上ꎮＭｏｂｉｌｅｎｅｔＶ３的瓶颈(ｂｎｅｃｋ)结构如图２所示ꎮ图中:ＮＬ表示非线性激活函数ꎻＰｏｏｌ表示平均池化ꎻＤｗｉｓｅ表示深度可分离卷积ꎻＦＣ表示全连接层ꎻＲｅＬＵ㊁ｈａｒｄ￣σ表示激活函数ꎻ 表示乘法操作ꎮＭｏｂｉｌＮｅｔＶ３在ＭｏｂｉｌＮｅｔＶ２的结构基础上增加了注意力机制(ｓｑｕｅｅｚｅ￣ａｎｄ￣ｅｘｃｉｔａｔｉｏｎꎬＳＥ)ꎬ并使用ｈ￣ｓｗｉｓｈ激活函数替换ｓｗｉｓｈ函数ꎬ相比于ＭｏｂｉｌＮｅｔＶ２ꎬＭｏｂｉｌＮｅｔＶ３模型更加轻量化ꎬ降低了计算成本ꎬ识别精度更高ꎬ计算速度更快ꎮ１.２.２㊀改进ＤｅｅｐＬａｂＶ３＋模型解码器设计经Ｅｎｃｏｄｅｒ得到的特征图由大量的像素矩阵构成ꎬ各像素之间均存在密切的联系ꎬＤｅｅｐＬａｂＶ３＋网络中Ｄｅｃｏｄｅｒ将传入的特征图直接使用一次双线性４倍上采样恢复目标边界信息ꎬ会使图像的像素不连续ꎬ导致网络预测边界不精确[１３]ꎮ本文将传入Ｄｅｃｏｄｅｒ中的特征信息先进行一次２倍上采样ꎬ还原边界信息ꎬ然后再进行一次２倍上采样ꎬ即使用２个连续的２倍上采样替换原始ＤｅｅｐＬａｂＶ３＋网络中的４倍上采样ꎬ增强图像中像素的连续性ꎬ还原出的边界信息更接近原始标注图像ꎬ从而获得更清晰的目标边界ꎮ改进的ＤｅｅｐＬａｂＶ３＋网络模型在Ｄｅｃｏｄｅｒ中仅添加了１次上采样操作ꎬ相比ＤｅｅｐＬａｂＶ３＋网络模型ꎬ参数增加极少ꎮ改进前后的Ｄｅｃｏｄｅｒ部分如图３所示ꎮ１５第１期㊀㊀㊀袁㊀帅等:改进ＤｅｅｐＬａｂＶ３＋网络的指针轨迹图像识别图１㊀ＤｅｅｐＬａｂＶ３＋网络结构Ｆｉｇ.１㊀ＤｅｅｐＬａｂＶ３＋ｎｅｔｗｏｒｋｓｔｒｕｃｔｕｒｅ图２㊀ＭｏｂｉｌｅＮｅｔＶ３网络的ｂｎｅｃｋ结构Ｆｉｇ.２㊀ＢｎｅｃｋｓｔｒｕｃｔｕｒｅｏｆＭｏｂｉｌｅＮｅｔＶ３ｎｅｔｗｏｒｋ图３㊀改进前后的Ｄｅｃｏｄｅｒ部分Ｆｉｇ.３㊀Ｄｅｃｏｄｅｒｐａｒｔｂｅｆｏｒｅａｎｄａｆｔｅｒｉｍｐｒｏｖｅｍｅｎｔ㊀㊀改进后的ＤｅｅｐＬａｂＶ３＋网络结构如图４所示ꎮ２㊀实验与分析２.１㊀实验环境本文基于百度飞桨(ＰａｄｄｌｅＰａｄｄｌｅ)深度学习框架进行实验ꎮ模型训练过程中使用随机梯度下降法ꎬ最大训练轮次为１００００轮ꎮ单次训练样本数设置为２ꎬ初始学习率设置为０.０１ꎬ之后通过多项式衰减策略减少学习率ꎮ㊀㊀实验数据来自设备运输过程中采集的指针式机械记录仪震动轨迹图像ꎬ以此自制数据集ꎬ该数据集包含精细标注影像８７３张ꎬ标注内容包括轨迹和背景ꎮ由于总体样本数据较少ꎬ采用图像反转㊁水平和垂直镜像等处理方法进行数据增强ꎬ丰富实验数据集ꎮ为解决网络模型在不同工作场景的算法适用性问题ꎬ使用图像分割套件ＰａｄｄｌｅＳｅｇ中的预训练模型ꎬ加快模型训练速度并保证特征提取效果ꎬ提高模型对指针轨迹图像识别的准确性和泛化性ꎮ２.２㊀评价指标本文选用平均交并比(ＭＩｏＵ)作为实验结果评价指标ꎬＭＩｏＵ是语义分割领域的标准度量指标ꎮ分别对每个类计算交并比(ＩｏＵ)ꎬ再对所有类别的ＩｏＵ求均值ꎬ即为ＭＩｏＵꎮ其计算式为ＭＩｏＵ＝１ｋ＋１ðｋｉ＝０ｐｉｉðｋｊ＝０ｐｉｊ＋ðｋｊ＝０ｐｊｉ－ｐｉｉ(１)式中:ｋ＋１表示图像中所有分割类别数目ꎻｐｉｊ表示标签为ｉ被预测为ｊ的样本数量ꎻｐｊｉ表示标签为ｊ被预测为ｉ的样本数量ꎻｐｉｉ表示标签为ｉ被预测为ｉ的样本数量ꎮ２５沈㊀阳㊀理㊀工㊀大㊀学㊀学㊀报㊀㊀第４３卷图４㊀改进后的ＤｅｅｐＬａｂＶ３＋网络结构Ｆｉｇ.４㊀ＩｍｐｒｏｖｅｄＤｅｅｐＬａｂＶ３＋ＮｅｔｗｏｒｋＳｔｒｕｃｔｕｒｅ㊀㊀采用单位时间内模型检测图片的数量(ＦＰＳ)作为另一个评价指标ꎬ该评价指标越大ꎬ表示检测的速度越快ꎬ其值为待检测的图片总数与模型预测所需的时间之比ꎮ２.３㊀实验对比对改进前后的ＤｅｅｐＬａｂＶ３＋模型进行训练测试ꎬ两组实验均使用ＰａｄｄｌｅＳｅｇ套件中自带的预训练模型ꎬ训练过程中其他条件相同ꎬ得到的可视化参数如图５和图６所示ꎮ图５㊀原始ＤｅｅｐＬａｂＶ３＋网络训练参数Ｆｉｇ.５㊀ＯｒｉｇｉｎａｌＤｅｅｐＬａｂＶ３＋ｎｅｔｗｏｒｋｔｒａｉｎｉｎｇｐａｒａｍｅｔｅｒｓ㊀㊀由图５和图６可见ꎬ改进ＤｅｅｐＬａｂＶ３＋网络的ＭＩｏＵ更高ꎬ模型识别指针轨迹图像的能力更好ꎬ损失函数能够更快地收敛且下降程度更大ꎬ模图６㊀改进ＤｅｅｐＬａｂＶ３＋网络训练参数Ｆｉｇ.６㊀ＩｍｐｒｏｖｅｄＤｅｅｐＬａｂＶ３＋ｎｅｔｗｏｒｋｔｒａｉｎｉｎｇｐａｒａｍｅｔｅｒｓ型的性能更优ꎮ训练完成后ꎬ将测试集中图片统一调整为相同的分辨率ꎬ分别使用网络改进前后两组实验中的最好模型进行预测ꎬ以图片１和图片２为例ꎬ两个图片的预测结果如图７和图８所示ꎮ由图７和图８可见ꎬ本文提出的改进Ｄｅｅｐ￣ＬａｂＶ３＋网络模型可以更好地获取轨迹语义信息ꎬ漏检和误检的比例更小ꎬ对轨迹特征提取能力更强ꎬ能够更好地完成轨迹图像识别任务ꎮ㊀㊀模型预测参数及模型大小如表１所示ꎮ㊀㊀由表１可知ꎬ本文提出的改进ＤｅｅｐＬａｂＶ３＋网络模型检测结果更精细㊁检测速度更快㊁模型体３５第１期㊀㊀㊀袁㊀帅等:改进ＤｅｅｐＬａｂＶ３＋网络的指针轨迹图像识别图７㊀对比实验预测结果(图片１)Ｆｉｇ.７㊀Ｃｏｍｐａｒｉｓｏｎｓｏｆｅｘｐｅｒｉｍｅｎｔａｌｐｒｅｄｉｃｔｉｏｎｒｅｓｕｌｔｓ(ｉｍａｇｅ１)图８㊀对比实验预测结果(图片２)Ｆｉｇ.８㊀Ｃｏｍｐａｒｉｓｏｎｓｏｆｅｘｐｅｒｉｍｅｎｔａｌｐｒｅｄｉｃｔｉｏｎｒｅｓｕｌｔｓ(ｉｍａｇｅ２)表１㊀模型参数对比Ｔａｂｌｅ１㊀Ｃｏｍｐａｒｉｓｏｎｓｏｆｍｏｄｅｌｐａｒａｍｅｔｅｒｓ模型ＭＩｏＵ/％ＦＰＳ/ｓ－１模型大小/ＭＢＤｅｅｐＬａｂＶ３＋８２.８８２.７３１７６.９１改进ＤｅｅｐＬａｂＶ３＋８５.８４６.３１４７.２７积更小ꎬ体现了该算法的可行性和优越性ꎮ３㊀结论为解决运输过程中车辆颠簸是否导致仪器设备损坏的检测问题ꎬ以ＤｅｅｐＬａｂＶ３＋作为语义分割模型ꎬ将其骨干网络替换为轻量化网络Ｍｏｂｉｌｅ￣ＮｅｔＶ３ꎬ减少了参数量ꎬ解决了模型部署到移动端硬件平台的问题ꎻ针对Ｄｅｃｏｄｅｒ结构中４倍上采样操作导致图像中的边界像素不连续㊁丢失某些重要像素信息问题ꎬ采用２次２倍上采样增强图像中像素的连续性ꎬ还原出的边界信息更接近原始标注图像ꎬ获得了更清晰的目标边界ꎮ改进后的ＤｅｅｐＬａｂＶ３＋模型体积减少了１２９.６４ＭＢꎻ检测结果更精细ꎬＭＩｏＵ达到８５.８４％ꎬ提高了３.５７％ꎻ检测速度更快ꎬＦＰＳ提升了３.５８ｓ－１ꎮ本文提出的改进ＤｅｅｐＬａｂＶ３＋网络降低了模型的参数量㊁加快了检测速度㊁提高了模型对轨迹图像识别的泛化能力ꎬ更宜于实际应用ꎮ参考文献(Ｒｅｆｅｒｅｎｃｅｓ):[１]㊀王宇.运动轨迹检测识别技术研究[Ｄ].沈阳:东北大学ꎬ２０１５.[２]㊀ＣＨＥＮＬＣꎬＺＨＵＹＫꎬＰＡＰＡＮＤＲＥＯＵＧꎬｅｔａｌ.Ｅｎｃｏｄｅｒ￣ｄｅ￣ｃｏｄｅｒｗｉｔｈａｔｒｏｕｓｓｅｐａｒａｂｌｅｃｏｎｖｏｌｕｔｉｏｎｆｏｒｓｅｍａｎｔｉｃｉｍａｇｅｓｅｇｍｅｎｔａｔｉｏｎ[Ｃ]//ＰｒｏｃｅｅｄｉｎｇｓｏｆｔｈｅＥｕｒｏｐｅａｎＣｏｎｆｅｒｅｎｃｅｏｎＣｏｍｐｕｔｅｒＶｉｓｉｏｎ.ＭｕｎｉｃｈꎬＧｅｒｍａｎｙ:Ｓｐｒｉｎｇｅｒꎬ２０１８:８０１－８１８.[３]㊀ＬＩＵＪＱꎬＷＡＮＧＺＬꎬＣＨＥＮＧＫＸ.ＡｎｉｍｐｒｏｖｅｄａｌｇｏｒｉｔｈｍｆｏｒｓｅｍａｎｔｉｃｓｅｇｍｅｎｔａｔｉｏｎｏｆｒｅｍｏｔｅｓｅｎｓｉｎｇｉｍａｇｅｓｂａｓｅｄｏｎＤｅｅｐＬａｂｖ３＋[Ｃ]//Ｐｒｏｃｅｅｄｉｎｇｓｏｆｔｈｅ５ｔｈＩｎｔｅｒｎａｔｉｏｎａｌＣｏｎ￣ｆｅｒｅｎｃｅｏｎＣｏｍｍｕｎｉｃａｔｉｏｎａｎｄＩｎｆｏｒｍａｔｉｏｎＰｒｏｃｅｓｓｉｎｇ.ＮｅｗＹｏｒｋꎬＵＳＡ:ＡＣＭꎬ２０１９:１２４－１２８.[４]㊀ＺＨＡＮＧＸＦꎬＢＩＡＮＨＮꎬＣＡＩＹＨꎬｅｔａｌ.ＡｎｉｍｐｒｏｖｅｄｔｏｎｇｕｅｉｍａｇｅｓｅｇｍｅｎｔａｔｉｏｎａｌｇｏｒｉｔｈｍｂａｓｅｄｏｎＤｅｅｐｌａｂｖ３＋ｆｒａｍｅｗｏｒｋ[Ｊ].ＩＥＴＩｍａｇｅＰｒｏｃｅｓｓｉｎｇꎬ２０２２ꎬ１６(５):１４７３－１４８５.[５]㊀刘慧ꎬ姜建滨ꎬ沈跃ꎬ等.基于改进ＤｅｅｐＬａｂＶ３＋的果园场景多类别分割方法[Ｊ].农业机械学报ꎬ２０２２ꎬ５３(１１):２５５－２６１.㊀㊀ＬＩＵＨꎬＪＩＡＮＧＪＢꎬＳＨＥＮＹꎬｅｔａｌ.Ｍｕｌｔｉ￣ｃａｔｅｇｏｒｙｓｅｇｍｅｎｔａ￣ｔｉｏｎｏｆｏｒｃｈａｒｄｓｃｅｎｅｂａｓｅｄｏｎｉｍｐｒｏｖｅｄＤｅｅｐＬａｂＶ３＋[Ｊ].ＴｒａｎｓａｃｔｉｏｎｓｏｆｔｈｅＣｈｉｎｅｓｅＳｏｃｉｅｔｙｆｏｒＡｇｒｉｃｕｌｔｕｒａｌＭａｃｈｉｎｅｒ￣ｙꎬ２０２２ꎬ５３(１１):２５５－２６１.(ｉｎＣｈｉｎｅｓｅ)[６]㊀周迅ꎬ李永龙ꎬ周颖玥ꎬ等.基于改进ＤｅｅｐＬａｂＶ３＋网络的坝面裂缝检测方法[Ｊ/ＯＬ].清华大学学报(自然科学版)ꎬ２０２３:１－１１[２０２３－０４－０７].ｈｔｔｐｓ://ｋｎｓ.ｃｎｋｉ.ｎｅｔ/ｋｃｍｓ/ｄｅｔａｉｌ//１１.２２２３.Ｎ.２０２３０２０８.１５１２.００１.ｈｔｍｌ.ＤＯＩ:１０.１６５１１/ｊ.ｃｎｋｉ.ｑｈｄｘｘｂ.２０２３.２６.００６.㊀㊀ＺＨＯＵＸꎬＬＩＹＬꎬＺＨＯＵＹＹꎬｅｔａｌ.Ｄａｍｓｕｒｆａｃｅｃｒａｃｋｄｅｔｅｃ￣ｔｉｏｎｍｅｔｈｏｄｂａｓｅｄｏｎｉｍｐｒｏｖｅｄＤｅｅｐＬａｂ￣Ｖ３＋ｎｅｔｗｏｒｋ[Ｊ/ＯＬ].ＪｏｕｒｎａｌｏｆＴｓｉｎｇｈｕａＵｎｉｖｅｒｓｉｔｙ(ＳｃｉｅｎｃｅａｎｄＴｅｃｈｎｏｌｏ￣ｇｙ)ꎬ２０２３:１－１１[２０２３－０４－０７].ｈｔｔｐｓ://ｋｎｓ.ｃｎｋｉ.ｎｅｔ/ｋｃｍｓ/ｄｅｔａｉｌ//１１.２２２３.Ｎ.２０２３０２０８.１５１２.００１.ｈｔｍｌ.ＤＯＩ:１０.１６５１１/ｊ.ｃｎｋｉ.ｑｈｄｘｘｂ.２０２３.２６.００６.(ｉｎＣｈｉｎｅｓｅ)[７]㊀周倩倩ꎬ刘汉林ꎬ陈维锋ꎬ等.基于Ｄｅｅｐｌａｂｖ３＋的排水管道缺陷检测与语义分割[Ｊ].中国给水排水ꎬ２０２２ꎬ３８(１３):２２－２７.㊀㊀ＺＨＯＵＱＱꎬＬＩＵＨＬꎬＣＨＥＮＷＦꎬｅｔａｌ.ＤｒａｉｎａｇｅｐｉｐｅｌｉｎｅｄｅｆｅｃｔｓｄｅｔｅｃｔｉｏｎａｎｄｓｅｍａｎｔｉｃｓｅｇｍｅｎｔａｔｉｏｎｂａｓｅｄｏｎＤｅｅｐ￣ｌａｂｖ３＋[Ｊ].ＣｈｉｎａＷａｔｅｒ＆Ｗａｓｔｅｗａｔｅｒꎬ２０２２ꎬ３８(１３):２２－２７.(ｉｎＣｈｉｎｅｓｅ)[８]㊀赵祥ꎬ王涛ꎬ张艳ꎬ等.基于改进ＤｅｅｐＬａｂｖ３＋孪生网络的遥感影像变化检测方法[Ｊ].地球信息科学学报ꎬ２０２２ꎬ２４(８):１６０４－１６１６.㊀㊀ＺＨＡＯＸꎬＷＡＮＧＴꎬＺＨＡＮＧＹꎬｅｔａｌ.ＲｅｍｏｔｅｓｅｎｓｉｎｇｉｍａｇｅｃｈａｎｇｅｄｅｔｅｃｔｉｏｎｂａｓｅｄｏｎｉｍｐｒｏｖｅｄＤｅｅｐＬａｂｖ３＋Ｓｉａｍｅｓｅｎｅｔ￣ｗｏｒｋ[Ｊ].ＪｏｕｒｎａｌｏｆＧｅｏ￣ｉｎｆｏｒｍａｔｉｏｎＳｃｉｅｎｃｅꎬ２０２２ꎬ２４(８):１６０４－１６１６.(ｉｎＣｈｉｎｅｓｅ)[９]㊀马冬梅ꎬ李鹏辉ꎬ黄欣悦ꎬ等.改进ＤｅｅｐＬａｂＶ３＋的高效语义分割[Ｊ].计算机工程与科学ꎬ２０２２ꎬ４４(４):７３７－７４５.㊀㊀ＭＡＤＭꎬＬＩＰＨꎬＨＵＡＮＧＸＹꎬｅｔａｌ.Ｅｆｆｉｃｉｅｎｔｓｅｍａｎｔｉｃｓｅｇ￣ｍｅｎｔａｔｉｏｎｂａｓｅｄｏｎｉｍｐｒｏｖｅｄＤｅｅｐＬａｂＶ３＋[Ｊ].ＣｏｍｐｕｔｅｒＥｎｇｉｎｅｅｒｉｎｇａｎｄＳｃｉｅｎｃｅꎬ２０２２ꎬ４４(４):７３７－７４５.(ｉｎＣｈｉ￣ｎｅｓｅ)[１０]姚燕ꎬ胡立坤ꎬ郭军.基于改进ＤｅｅｐＬａｂｖ３＋网络的轻量级语义分割算法[Ｊ].激光与光电子学进展ꎬ２０２２ꎬ５９(４):２００－２０７.㊀㊀ＹＡＯＹꎬＨＵＬＫꎬＧＵＯＪ.Ｉｍｐｒｏｖｅｄｌｉｇｈｔｗｅｉｇｈｔｓｅｍａｎｔｉｃｓｅｇ￣ｍｅｎｔａｔｉｏｎａｌｇｏｒｉｔｈｍｂａｓｅｄｏｎＤｅｅｐＬａｂｖ３＋ｎｅｔｗｏｒｋ[Ｊ].Ｌａ￣ｓｅｒ＆ＯｐｔｏｅｌｅｃｔｒｏｎｉｃｓＰｒｏｇｒｅｓｓꎬ２０２２ꎬ５９(４):２００－２０７.(ｉｎＣｈｉｎｅｓｅ)(下转第６０页)较于原网络ꎬ本文改进网络重建图像的细节更加清晰ꎬＰＳＮＲ提升了０.５３ｄＢꎬＳＳＩＭ提升了０.０２ꎮ本文算法为红外图像去模糊任务提供了一种良好的解决方案ꎮ参考文献(Ｒｅｆｅｒｅｎｃｅｓ):[１]㊀钱阳.红外热成像技术在警用执法领域的应用[Ｊ].中国安防ꎬ２０２３(Ｚ１):８３－８８.㊀㊀ＱＩＡＮＹ.Ａｐｐｌｉｃａｔｉｏｎｏｆｉｎｆｒａｒｅｄｔｈｅｒｍａｌｉｍａｇｉｎｇｔｅｃｈｎｏｌｏｇｙｉｎｔｈｅｆｉｅｌｄｏｆｐｏｌｉｃｅｌａｗｅｎｆｏｒｃｅｍｅｎｔ[Ｊ].ＣｈｉｎａＳｅｃｕｒｉｔｙ＆Ｐｒｏｔｅｃｔｉｏｎꎬ２０２３(Ｚ１):８３－８８.(ｉｎＣｈｉｎｅｓｅ) [２]㊀赵国顺.红外热成像技术的应用延伸探析[Ｊ].中国安防ꎬ２０２３(Ｚ１):９４－９７.㊀㊀ＺＨＡＯＧＳ.Ａｎａｌｙｓｉｓｏｎｔｈｅａｐｐｌｉｃａｔｉｏｎｅｘｔｅｎｓｉｏｎｏｆｉｎｆｒａｒｅｄｔｈｅｒｍａｌｉｍａｇｉｎｇｔｅｃｈｎｏｌｏｇｙ[Ｊ].ＣｈｉｎａＳｅｃｕｒｉｔｙ＆Ｐｒｏｔｅｃｔｉｏｎꎬ２０２３(Ｚ１):９４－９７.(ｉｎＣｈｉｎｅｓｅ)[３]㊀郝建坤ꎬ黄玮ꎬ刘军ꎬ等.空间变化ＰＳＦ非盲去卷积图像复原法综述[Ｊ].中国光学ꎬ２０１６ꎬ９(１):４１－５０.㊀㊀ＨＡＯＪＫꎬＨＵＡＮＧＷꎬＬＩＵＪꎬｅｔａｌ.Ｒｅｖｉｅｗｏｆｎｏｎ￣ｂｌｉｎｄｄｅ￣ｃｏｎｖｏｌｕｔｉｏｎｉｍａｇｅｒｅｓｔｏｒａｔｉｏｎｂａｓｅｄｏｎｓｐａｔｉａｌｌｙ￣ｖａｒｙｉｎｇＰＳＦ[Ｊ].ＣｈｉｎｅｓｅＯｐｔｉｃｓꎬ２０１６ꎬ９(１):４１－５０.(ｉｎＣｈｉｎｅｓｅ) [４]㊀华夏ꎬ张悦芮ꎬ刘广ꎬ等.视觉灵敏性滤波和自适应正则化图像盲去模糊[Ｊ].华中科技大学学报(自然科学版)ꎬ２０２１ꎬ４９(１１):６４－７０.㊀㊀ＨＵＡＸꎬＺＨＡＮＧＹＲꎬＬＩＵＧꎬｅｔａｌ.Ｂｌｉｎｄｉｍａｇｅｄｅｂｌｕｒｒｉｎｇｂａｓｅｄｏｎｖｉｓｕａｌｓｅｎｓｉｔｉｖｉｔｙｆｉｌｔｅｒｉｎｇａｎｄａｄａｐｔｉｖｅｒｅｇｕｌａｒｉｚａｔｉｏｎ[Ｊ].ＪｏｕｒｎａｌｏｆＨｕａｚｈｏｎｇＵｎｉｖｅｒｓｉｔｙｏｆＳｃｉｅｎｃｅａｎｄＴｅｃｈｎｏｌ￣ｏｇｙ(ＮａｔｕｒａｌＳｃｉｅｎｃｅＥｄｉｔｉｏｎ)ꎬ２０２１ꎬ４９(１１):６４－７０.(ｉｎＣｈｉｎｅｓｅ)[５]㊀ＳＵＪＷꎬＸＵＢＹꎬＹＩＮＨＪ.Ａｓｕｒｖｅｙｏｆｄｅｅｐｌｅａｒｎｉｎｇａｐｐｒｏａ￣ｃｈｅｓｔｏｉｍａｇｅｒｅｓｔｏｒａｔｉｏｎ[Ｊ].Ｎｅｕｒｏｃｏｍｐｕｔｉｎｇꎬ２０２２ꎬ４８７:４６－６５.[６]㊀ＺＨＡＮＧＫＨꎬＲＥＮＷＱꎬＬＵＯＷＨꎬｅｔａｌ.Ｄｅｅｐｉｍａｇｅｄｅ￣ｂｌｕｒｒｉｎｇ:ａｓｕｒｖｅｙ[Ｊ].ＩｎｔｅｒｎａｔｉｏｎａｌＪｏｕｒｎａｌｏｆＣｏｍｐｕｔｅｒＶｉ￣ｓｉｏｎꎬ２０２２ꎬ１３０(９):２１０３－２１３０.[７]㊀ＫＵＰＹＮＯꎬＢＵＤＺＡＮＶꎬＭＹＫＨＡＩＬＹＣＨＭꎬｅｔａｌ.Ｄｅｂｌｕｒ￣ＧＡＮ:ｂｌｉｎｄｍｏｔｉｏｎｄｅｂｌｕｒｒｉｎｇｕｓｉｎｇｃｏｎｄｉｔｉｏｎａｌａｄｖｅｒｓａｒｉａｌｎｅｔｗｏｒｋｓ[Ｃ]//２０１８ＩＥＥＥ/ＣＶＦＣｏｎｆｅｒｅｎｃｅｏｎＣｏｍｐｕｔｅｒＶｉｓｉｏｎａｎｄＰａｔｔｅｒｎＲｅｃｏｇｎｉｔｉｏｎ.ＳａｌｔＬａｋｅＣｉｔｙꎬＵＳＡ:ＩＥＥＥꎬ２０１８:８１８３－８１９２.[８]㊀ＫＵＰＹＮＯꎬＭＡＲＴＹＮＩＵＫＴꎬＷＵＪＲꎬｅｔａｌ.ＤｅｂｌｕｒＧＡＮ￣ｖ２:ｄｅｂｌｕｒｒｉｎｇ(ｏｒｄｅｒｓ￣ｏｆ￣ｍａｇｎｉｔｕｄｅ)ｆａｓｔｅｒａｎｄｂｅｔｔｅｒ[Ｃ]//２０１９ＩＥＥＥ/ＣＶＦＩｎｔｅｒｎａｔｉｏｎａｌＣｏｎｆｅｒｅｎｃｅｏｎＣｏｍｐｕｔｅｒＶｉｓｉｏｎ(ＩＣ￣ＣＶ).ＳｅｏｕｌꎬＫｏｒｅａ(Ｓｏｕｔｈ):ＩＥＥＥꎬ２０２０:８８７７－８８８６. [９]㊀ＷＡＮＧＺＤꎬＣＵＮＸＤꎬＢＡＯＪＭꎬｅｔａｌ.Ｕｆｏｒｍｅｒ:ａｇｅｎｅｒａｌＵ￣ｓｈａｐｅｄｔｒａｎｓｆｏｒｍｅｒｆｏｒｉｍａｇｅｒｅｓｔｏｒａｔｉｏｎ[Ｃ]//２０２２ＩＥＥＥ/ＣＶＦＣｏｎｆｅｒｅｎｃｅｏｎＣｏｍｐｕｔｅｒＶｉｓｉｏｎａｎｄＰａｔｔｅｒｎＲｅｃｏｇｎｉｔｉｏｎ(ＣＶＰＲ).ＮｅｗＯｒｌｅａｎｓꎬＵＳＡ:ＩＥＥＥꎬ２０２２:１７６６２－１７６７２. [１０]ＺＡＭＩＲＳＷꎬＡＲＯＲＡＡꎬＫＨＡＮＳꎬｅｔａｌ.Ｒｅｓｔｏｒｍｅｒ:ｅｆｆｉｃｉｅｎｔｔｒａｎｓｆｏｒｍｅｒｆｏｒｈｉｇｈ￣ｒｅｓｏｌｕｔｉｏｎｉｍａｇｅｒｅｓｔｏｒａｔｉｏｎ[Ｃ]//２０２２ＩＥＥＥ/ＣＶＦＣｏｎｆｅｒｅｎｃｅｏｎＣｏｍｐｕｔｅｒＶｉｓｉｏｎａｎｄＰａｔｔｅｒｎＲｅｃ￣ｏｇｎｉｔｉｏｎ(ＣＶＰＲ).ＮｅｗＯｒｌｅａｎｓꎬＵＳＡ:ＩＥＥＥꎬ２０２２:５７１８－５７２９.[１１]ＴＡＯＸꎬＧＡＯＨꎬＳＨＥＮＸꎬｅｔａｌ.Ｓｃａｌｅ￣ｒｅｃｕｒｒｅｎｔｎｅｔｗｏｒｋｆｏｒｄｅｅｐｉｍａｇｅｄｅｂｌｕｒｒｉｎｇ[Ｃ]//２０１８ＩＥＥＥ/ＣＶＦＣｏｎｆｅｒｅｎｃｅｏｎＣｏｍｐｕｔｅｒＶｉｓｉｏｎａｎｄＰａｔｔｅｒｎＲｅｃｏｇｎｉｔｉｏｎ.ＳａｌｔＬａｋｅＣｉｔｙꎬＵＳＡ:ＩＥＥＥꎬ２０１８:８１７４－８１８２.[１２]ＣＨＯＳＪꎬＪＩＳＷꎬＨＯＮＧＪＰꎬｅｔａｌ.Ｒｅｔｈｉｎｋｉｎｇｃｏａｒｓｅ￣ｔｏ￣ｆｉｎｅａｐｐｒｏａｃｈｉｎｓｉｎｇｌｅｉｍａｇｅｄｅｂｌｕｒｒｉｎｇ[Ｃ]//２０２１ＩＥＥＥ/ＣＶＦＩｎｔｅｒｎａｔｉｏｎａｌＣｏｎｆｅｒｅｎｃｅｏｎＣｏｍｐｕｔｅｒＶｉｓｉｏｎ(ＩＣＣＶ).Ｍｏｎｔ￣ｒｅａｌꎬＣａｎａｄａ:ＩＥＥＥꎬ２０２１:４６２１－４６３０.[１３]ＺＡＭＩＲＳＷꎬＡＲＯＲＡＡꎬＫＨＡＮＳꎬｅｔａｌ.Ｌｅａｒｎｉｎｇｅｎｒｉｃｈｅｄｆｅａｔｕｒｅｓｆｏｒｒｅａｌｉｍａｇｅｒｅｓｔｏｒａｔｉｏｎａｎｄｅｎｈａｎｃｅｍｅｎｔ[Ｃ]//ＣｏｍｐｕｔｅｒＶｉｓｉｏｎ:ＥＣＣＶ２０２０Ｗｏｒｋｓｈｏｐｓ.ＧｌａｓｇｏｗꎬＵＫ:Ｓｐｒｉｎｇｅｒꎬ２０２０:４９２－５１１.[１４]王光博ꎬ陈亮.改进关系网络的小样本图像分类方法[Ｊ].沈阳理工大学学报ꎬ２０２３ꎬ４２(１):２８－３４.㊀㊀ＷＡＮＧＧＢꎬＣＨＥＮＬ.Ｓｍａｌｌｓａｍｐｌｅｉｍａｇｅｃｌａｓｓｉｆｉｃａｔｉｏｎｍｅｔｈｏｄｂａｓｅｄｏｎｉｍｐｒｏｖｅｄｒｅｌａｔｉｏｎａｌｎｅｔｗｏｒｋ[Ｊ].ＪｏｕｒｎａｌｏｆＳｈｅｎｙａｎｇＬｉｇｏｎｇＵｎｉｖｅｒｓｉｔｙꎬ２０２３ꎬ４２(１):２８－３４.(ｉｎＣｈｉ￣ｎｅｓｅ)(责任编辑:宋颖韬)(上接第５４页)[１１]胡智翔ꎬ鲍胜利ꎬ徐传淇ꎬ等.基于ＤｅｅｐＬａｂｖ３＋的遥感建筑物语义分割算法[Ｊ].计算机应用ꎬ２０２１ꎬ４１(Ｓ２):７１－７５.㊀㊀ＨＵＺＸꎬＢＡＯＳＬꎬＸＵＣＱꎬｅｔａｌ.Ｓｅｍａｎｔｉｃｓｅｇｍｅｎｔａｔｉｏｎａｌ￣ｇｏｒｉｔｈｍｆｏｒｒｅｍｏｔｅｓｅｎｓｉｎｇｂｕｉｌｄｉｎｇｓｂａｓｅｄｏｎＤｅｅｐＬａｂｖ３＋[Ｊ].ＪｏｕｒｎａｌｏｆＣｏｍｐｕｔｅｒＡｐｐｌｉｃａｔｉｏｎｓꎬ２０２１ꎬ４１(Ｓ２):７１－７５.(ｉｎＣｈｉｎｅｓｅ)[１２]陈兵ꎬ贺晟ꎬ刘坚ꎬ等.基于轻量化ＤｅｅｐＬａｂｖ３＋网络的焊缝结构光图像分割[Ｊ].中国激光ꎬ２０２３ꎬ５０(８):４９－５８.㊀㊀ＣＨＥＮＢꎬＨＥＳꎬＬＩＵＪꎬｅｔａｌ.Ｗｅｌｄｓｔｒｕｃｔｕｒｅｄｌｉｇｈｔｉｍａｇｅｓｅｇ￣ｍｅｎｔａｔｉｏｎｂａｓｅｄｏｎｌｉｇｈｔｗｅｉｇｈｔＤｅｅｐＬａｂｖ３＋ｎｅｔｗｏｒｋ[Ｊ].ＣｈｉｎｅｓｅＪｏｕｒｎａｌｏｆＬａｓｅｒｓꎬ２０２３ꎬ５０(８):４９－５８.(ｉｎＣｈｉ￣ｎｅｓｅ)[１３]徐聪ꎬ王丽.基于改进ＤｅｅｐＬａｂｖ３＋网络的图像语义分割方法[Ｊ].激光与光电子学进展ꎬ２０２１ꎬ５８(１６):１６１０００８.㊀㊀ＸＵＣꎬＷＡＮＧＬ.ＩｍａｇｅｓｅｍａｎｔｉｃｓｅｇｍｅｎｔａｔｉｏｎｍｅｔｈｏｄｂａｓｅｄｏｎｉｍｐｒｏｖｅｄＤｅｅｐＬａｂｖ３＋ｎｅｔｗｏｒｋ[Ｊ].Ｌａｓｅｒ＆Ｏｐｔｏｅｌｅｃ￣ｔｒｏｎｉｃｓＰｒｏｇｒｅｓｓꎬ２０２１ꎬ５８(１６):１６１０００８.(ｉｎＣｈｉｎｅｓｅ)(责任编辑:宋颖韬)。

用python实现鸢尾花数据集的朴素贝叶斯算法

用python实现鸢尾花数据集的朴素贝叶斯算法Python是一种功能强大的编程语言，广泛应用于数据科学和机器学习领域。

朴素贝叶斯算法是一种常见的建模方法，可以用于分类问题。

在这篇文章中，我们将使用Python实现朴素贝叶斯算法来处理鸢尾花数据集。

首先，我们需要导入一些必要的库。

在Python中，有很多强大的数据处理和机器学习库可供选择，例如NumPy、Pandas和Scikit-Learn。

这些库可以帮助我们加载和处理数据，以及构建机器学习模型。

我们首先导入NumPy和Pandas库，用于数据处理和分析。

pythonimport numpy as npimport pandas as pd接下来，我们将使用Pandas库加载鸢尾花数据集。

鸢尾花数据集是一个常用的机器学习数据集，其中包含150个样本，每个样本有四个特征：花萼长度、花萼宽度、花瓣长度和花瓣宽度。

鸢尾花数据集由三个类别组成：Setosa、Versicolor和Virginica。

我们可以使用Pandas库的read_csv函数从CSV文件中加载数据集。

pythondata = pd.read_csv('iris.csv')数据集加载完成后，我们可以使用head()函数查看前几行数据，确保数据正确加载。

pythonprint(data.head())现在，我们已经成功加载了鸢尾花数据集。

接下来，我们将进行数据预处理，以准备数据用于训练朴素贝叶斯分类器。

首先，我们需要将数据集拆分为特征和目标变量。

特征是我们用来预测目标变量的变量，而目标变量是我们希望预测的变量。

pythonX = data.drop('species', axis=1)y = data['species']在实施朴素贝叶斯算法之前，我们需要将特征进行标准化处理。

标准化可以确保数据具有相似的范围和分布，有助于提高算法的性能。

我们可以使用Scikit-Learn库中的StandardScaler进行标准化处理。

蓝莓拟茎点枝枯病的病原_岳清华

蓝莓拟茎点枝枯病的病原
岳清华
1 2
1
赵洪海 1*
梁晨 1
李晓东 2
青岛农业大学农学与植物保护学院山东省植物病虫害综合防控重点实验室山东青岛 266109 青岛市果茶卉工作站山东青岛 266071
摘
要：近期在山东省蓝莓种植区发现一种枝枯病害。为明确其致病菌，通过单孢分离方法和接种试验获得 3 个菌
YUE Qing-Hua1
1
ZHAO Hong-Hai1*
LIANG Chen1
LI Xiao-Dong2
Key Lab of Integrated Crop Pest Management of Shandong Province, College of Agronomy and Plant Protection, Qingdao Agricultural
*
Corresponding author. E-mail: liangchen@
收稿日期: 2012-10-16, 接受日期: 2013-05-10
960
ISSN1672-6472 CN11-5180/Q Mycosystema
November 15, 2013 Vol.32
株。通过形态学特征观察和 rDNA-ITS 序列分析，确定该病原菌为乌饭树拟茎点霉 Phomopsis vaccinii，即越橘间座壳 Diaporthe vaccinii 的无性阶段。这是乌饭树拟茎点霉所致蓝莓枝枯病在国内的首次发现。关键词：乌饭树拟茎点霉，形态特征，分子鉴定，枝干病害
The pathogen causing Phomopsis twig blight of blueberry
岳清华等 /蓝莓拟茎点枝枯病的病原

基于多层特征嵌入的单目标跟踪算法

基于多层特征嵌入的单目标跟踪算法1. 内容描述基于多层特征嵌入的单目标跟踪算法是一种在计算机视觉领域中广泛应用的跟踪技术。

该算法的核心思想是通过多层特征嵌入来提取目标物体的特征表示，并利用这些特征表示进行目标跟踪。

该算法首先通过预处理步骤对输入图像进行降维和增强，然后将降维后的图像输入到神经网络中，得到不同层次的特征图。

通过对这些特征图进行池化操作，得到一个低维度的特征向量。

将这个特征向量输入到跟踪器中，以实现对目标物体的实时跟踪。

为了提高单目标跟踪算法的性能，本研究提出了一种基于多层特征嵌入的方法。

该方法首先引入了一个自适应的学习率策略，使得神经网络能够根据当前训练状态自动调整学习率。

通过引入注意力机制，使得神经网络能够更加关注重要的特征信息。

为了进一步提高跟踪器的鲁棒性，本研究还采用了一种多目标融合的方法，将多个跟踪器的结果进行加权融合，从而得到更加准确的目标位置估计。

通过实验验证，本研究提出的方法在多种数据集上均取得了显著的性能提升，证明了其在单目标跟踪领域的有效性和可行性。

1.1 研究背景随着计算机视觉和深度学习技术的快速发展，目标跟踪在许多领域(如安防、智能监控、自动驾驶等)中发挥着越来越重要的作用。

单目标跟踪(MOT)算法是一种广泛应用于视频分析领域的技术，它能够实时跟踪视频序列中的单个目标物体，并将其位置信息与相邻帧进行比较，以估计目标的运动轨迹。

传统的单目标跟踪算法在处理复杂场景、遮挡、运动模糊等问题时表现出较差的鲁棒性。

为了解决这些问题，研究者们提出了许多改进的单目标跟踪算法，如基于卡尔曼滤波的目标跟踪、基于扩展卡尔曼滤波的目标跟踪以及基于深度学习的目标跟踪等。

这些方法在一定程度上提高了单目标跟踪的性能，但仍然存在一些局限性，如对多目标跟踪的支持不足、对非平稳运动的适应性差等。

开发一种既能有效跟踪单个目标物体，又能应对多种挑战的单目标跟踪算法具有重要的理论和实际意义。

1.2 研究目的本研究旨在设计一种基于多层特征嵌入的单目标跟踪算法，以提高目标跟踪的准确性和鲁棒性。

基于粒子群优化的深度神经网络分类算法

基于粒子群优化的深度神经网络分类算法董晴;宋威【摘要】针对神经网络分类算法中节点函数不可导,分类精度不够高等问题,提出了一种基于粒子群优化(PSO)算法的深度神经网络分类算法.使用深度学习中的自动编码机,结合PSO算法优化权值,利用自动编码机对输入样本数据进行编解码,为提高网络分类精度,以编码机本身的误差函数和Softmax分类器的代价函数加权求和共同作为PSO算法的评价函数,使编码后的数据更加适应分类器.实验结果证明:与其他传统的神经网络相比,在邮件分类问题上,此分类算法有更高的分类精度.%Aiming at problem that classification precision of neural network algorithm is not very high and node function doesn't have derivate,a new classification algorithm of deep neural network based on particle swarm optimization(PSO) is e autoencoder of deep study,and combined with PSO algorithm to optimize the weight,coder and decoder for input sample data using autoencoder.In order to improve the classification precision of network,take the error function of autoencoder and cost function of softmax classifier weight sum as evaluation function of PSO algorithm in common,making coded data more adapter to the classifier.The experimental results show that compared with other traditional neural network,the classification algorithm has higher classification precision on Email classification.【期刊名称】《传感器与微系统》【年(卷),期】2017(036)009【总页数】5页(P143-146,150)【关键词】深度神经网络;自动编码机;粒子群优化算法;分类【作者】董晴;宋威【作者单位】江南大学物联网工程学院,江苏无锡214122;江南大学物联网工程学院,江苏无锡214122【正文语种】中文【中图分类】TP183近年来，神经网络的研究一直受到学者们的关注，如感知机[1]，反向传播(back propogation,BP)神经网络[2]，径向基函数(radial basis function,RBF)神经网络及其各种改进算法[3～5]等。

chap11.主要分子生物信息数据库

Volume 38, Database issue, January 2010 AUTHORGuy R. CochraneEric W. SayersCatherine BrooksbankYukiko YamazakiEli KaminumaRasko LeinonenDennis A. BensonWeizhong LiRaphaël LeplaePryavahiny KichenaradjaMichaël BekaertGiorgio GrilloPora KimJun-ichi TakedaBruno Contreras-Moreira Riu YamashitaElodie Portales-Casamar Pavel S. NovichkovJuan WangJian-Hua YangMartin Mokrejs Panagiotis AlexiouThe UniProt Consortium Yan ZhangJian RenChristian J. A. Sigrist Cathryn M. Gould Annalisa MarsicoJ. MullerGabriel ÖstlundRobert D. Finn Thomas RatteiNeil D. Rawlings Richard J. Roberts Saskia Preissner Andreas Schlicker Paula de MatosYanli WangGuohui Zheng Christian Koetschan Douglas H. Turner Michail Yu. Lobanov Yan Yuan Tseng Jonathan LeesFrançois EhrenmannSven GriepDonald S. Berkholz Patrick MayThe Gene Ontology ConsortiumJohannes GollTanja Davidsen Konstantinos Liolios Minoru KanehisaIkuo UchiyamaLubos KlucarJ. PelletMitsuteru Nakao Victor M. Markowitz Renzo Kottmann Paramvir S. DehalLuke E. UlrichLauren M. Brinkac Cristina Aurrecoechea Martha B. Arnaud Marek S. Skrzypek Stacia R. EngelHee Shin KimUlrike PfreundtMoritz GilsdorfJun DuanMartin AslettTodd W. HarrisVineet K. SharmaRon CaspiAlex FrolkisJunfeng GaoLewis Y. GeerAndreas RueppJudice L. Y. KohMilana Frenkel-Morgenstern Petras J. Kundrotas Benjamin A. ShoemakerB. Aranda, P. Achuthan Arnaud CeolPawel SmialowskiPeter VanheeMichael KuhnPaul FlicekP. J. KerseyPhil WilkinsonHugh MorganCarol J. BultAndrew BlakeKeiko AkagiJeff B. BowesBrooke RheadKate R. Rosenbloom Chisato YamasakiJon W. HussSamuel HiardSimon A. ForbesHong LiLishan WangAdnan S. SyedBrian A. Kennedy Stefan M. Woerner Misha Kapushesky Nicholas Paul GauthierLorna RichardsonJun ZhaoAedín C. Culhane Ramil N. NurtdinovLi JiJuan Antonio Vizcaíno Catherine Y. Cormier Ron MiloLynn M. SchrimlLiwei LiShaini ThomasEmilia LimFeng ZhuAthanasia Spandidos Agatha Schlüter Zhenhai ZhangWalter Sanseverino Paulino Pérez-Rodríguez Pawel DurekMotohiro MiharaDavid GrantHifzur Rahman Ansari Randi VitaJames RobinsonMartin ShumwayDATABASE NAMEThe 2010 Nucleic Acids Research Database Issue and online Database Collection: a community of data resources Database resources of the National Center for Biotechnology InformationThe European Bioinformatics Institute’s data resources NBRP databases: databases of biological resources in Japan DDBJ launches a new archive database with analytical tools for next-generation sequence dataImprovements to services at the European Nucleotide Archive GenBankNon-redundant patent sequence databases with value-added annotations at two levelsACLAME: A CLAssification of Mobile genetic Elements, update 2010ISbrowser: an extension of ISfinder for visualizing insertion sequences in prokaryotic genomesRecode-2: new design, new search tools, and many more genes UTRdb and UTRsite (RELEASE 2010): a collection of sequences and regulatory motifs of the untranslated regions of eukaryotic mRNAsChimerDB 2.0—a knowledgebase for fusion genes updatedH-DBAS: human-transcriptome database for alternative splicing: update 20103D-footprint: a database for the structural analysis of protein–DNA complexesDBTSS provides a tissue specific dynamic view of Transcription Start SitesJASPAR 2010: the greatly expanded open-access database of transcription factor binding profilesRegPrecise: a database of curated genomic inferences of transcriptional regulatory interactions in prokaryotesTransmiR: a transcription factor–microRNA regulation databasedeepBase: a database for deeply annotating and mining deep sequencing dataIRESite—a tool for the examination of viral and cellular internal ribosome entry sitesmiRGen 2.0: a database of microRNA genomic information and regulationThe Universal Protein Resource (UniProt) in 2010HHMD: the human histone modification databaseMiCroKit 3.0: an integrated database of midbody, centrosome and kinetochorePROSITE, a protein domain database for functional characterization and annotationELM: the status of the 2010 eukaryotic linear motif resource MeMotif: a database of linear motifs in a-helical transmembrane proteinseggNOG v2.0: extending the evolutionary genealogy of genes with enhanced non-supervised orthologous groups, species and functional annotationsInParanoid 7: new algorithms and tools for eukaryotic orthology analysisPANTHER version 7: improved phylogenetic trees, orthologs and collaboration with the Gene Ontology ConsortiumThe Pfam protein families databaseSIMAP—a comprehensive database of pre-calculated protein sequence similarities, domains, annotations and clusters MEROPS: the peptidase databaseREBASE—a database for DNA restriction and modification: enzymes, genes and genomesSuperCYP: a comprehensive database on Cytochrome P450 enzymes including a tool for analysis of CYP-drug interactionsFunSimMat update: new features for exploring functional similarityChemical Entities of Biological Interest: an updateAn overview of the PubChem BioAssay resource3DNALandscapes: a database for exploring the conformational features of DNAThe ITS2 Database III—sequences and structures for phylogenyNNDB: the nearest neighbor parameter database for predicting stability of nucleic acid secondary structureComSin: database of protein structures in bound (complex) and unbound (single) states in relation to their intrinsic disorderf POP: footprinting functional pockets of proteins by comparative spatial patternsGene3D: merging structure and function for a Thousand genomesIMGT/3Dstructure-DB and IMGT/DomainGapAlign: a database and a tool for immunoglobulins or antibodies, T cell receptors, MHC, IgSF and MhcSFPDBe: Protein Data Bank in EuropePDBselect 1992–2009 and PDBfilter-selectProtein Geometry Database: a flexible engine to explore backbone conformations and their relationships to covalent geometryPTGL: a database for secondary structure-based protein topologiesThe Gene Ontology in 2010: extensions and refinementsThe Protein Naming Utility: a rules database for protein nomenclatureThe comprehensive microbial resourceThe Genomes On Line Database (GOLD) in 2009: status of genomic and metagenomic projects and their associated metadataKEGG for representation and analysis of molecular networks involving diseases and drugsMBGD update 2010: toward a comprehensive resource for exploring microbial genome diversityphiSITE: database of gene regulation in bacteriophages ViralORFeome: an integrated database to generate a versatile collection of viral ORFsCyanoBase: the cyanobacteria genome database update 2010The integrated microbial genomes system: an expanding comparative analysis resource: integrated database resource for marine ecological genomicsMicrobesOnline: an integrated portal for comparative and functional genomicsThe MiST2 database: a comprehensive genomics resource on microbial signal transductionPathema: a clade-specific bioinformatics resource center for pathogen researchEuPathDB: a portal to eukaryotic pathogen databasesThe Aspergillus Genome Database, a curated comparative genomics resource for gene, protein and sequence information for the Aspergillus research communityNew tools at the Candida Genome Database: biochemical pathways and full-text literature searchSaccharomyces Genome Database provides mutant phenotypedataBeetleBase in 2010: revisions to provide comprehensive genomic information for Tribolium castaneumFlyTF: improved annotation and enhanced functionality of the Drosophila transcription factor databaseGenomeRNAi: a database for cell-based RNAi phenotypes. 2009 updateSilkDB v2.0: a platform for silkworm (Bombyx mori ) genome biologyTriTrypDB: a functional genomic resource for the TrypanosomatidaeWormBase: a comprehensive resource for nematode research MetaBioME: a database to explore commercially useful enzymes in metagenomic datasetsThe MetaCyc database of metabolic pathways and enzymes and the BioCyc collection of pathway/genome databasesSMPDB: The Small Molecule Pathway DatabaseThe University of Minnesota Biocatalysis/Biodegradation Database: improving public accessThe NCBI BioSystems databaseCORUM: the comprehensive resource of mammalian protein complexes—2009DRYGIN: a database of quantitative genetic interaction networks in yeastDynamic Proteomics: a database for dynamics andlocalizations of endogenous fluorescently-tagged proteins in living human cellsGWIDD: Genome-wide protein docking databaseInferred Biomolecular Interaction Server—a web server to analyze and predict protein interacting partners and binding sitesThe IntAct molecular interaction database in 2010MINT, the molecular interaction database: 2009 updateThe Negatome database: a reference set of non-interacting protein pairsPepX: a structural database of non-redundant protein–peptide complexesSTITCH 2: an interaction network database for small molecules and proteinsEnsembl’s 10th yearEnsembl Genomes: Extending Ensembl across the taxonomic spaceEMMA—mouse mutant resources for the internationalscientific communityEuroPhenome: a repository for high-throughput mouse phenotyping dataThe Mouse Genome Database: enhancements and updatesMouseBook: an integrated portal of mouse resources MouseIndelDB: a database integrating genomic indel polymorphisms that distinguish mouse strainsXenbase: gene expression and improved integrationThe UCSC Genome Browser database: update 2010ENCODE whole-genome data in the UCSC Genome BrowserH-InvDB in 2009: extended database and data mining resources for human genes and transcriptsThe Gene Wiki: community intelligence applied to human gene annotationPatrocles: a database of polymorphic miRNA-mediated gene regulation in vertebratesCOSMIC (the Catalogue of Somatic Mutations in Cancer): a resource to investigate acquired mutations in human cancerdbDEPC: a database of Differentially Expressed Proteins in human CancersHLungDB: an integrated database of human lung cancer researchNetwork of Cancer Genes: a web resource to analyze duplicability, orthology and network properties of cancer genesHRTBLDb: an informative data resource for hormone receptors target binding lociSelTar base, a database of human mononucleotide-microsatellite mutations and their potential impact to tumorigenesis and immunologyGene Expression Atlas at the European BioinformaticsInstitute: version 2.0, an updated comprehensive, multi-species repository of cell cycle experiments and derived analysis resultsEMAGE mouse embryo spatial gene expression database: 2010 updateFlyTED: the Drosophila Testis Gene Expression Database GeneSigDB—a curated database of gene expression signaturesPLANdbAffy: probe-level annotation database for Affymetrix expression microarraysNCBI Peptidome: a new repository for mass spectrometry proteomics dataThe Proteomics Identifications database: 2010 updateProtein Structure Initiative Material Repository: an open shared public resource of structural genomics plasmids for the biological communityBioNumbers—the database of key numbers in molecular andcell biologyGeMInA, Genomic Metadata for Infectious Agents, a geospatial surveillance pathogen databaseBioDrugScreen: a computational drug design resource for ranking molecules docked to the human proteomeCAMP: a useful resource for research on antimicrobial peptidesT3DB: a comprehensively annotated database of common toxins and their targetsUpdate of TTD: Therapeutic Target DatabasePrimerBank: a resource of human and mouse PCR primer pairs for gene expression detection and quantification PeroxisomeDB 2.0: an integrative view of the global peroxisomal metabolomePMRD: plant microRNA databasePRGdb: a bioinformatics platform for plant resistance gene analysisPlnTFDB: updated content and new features of the plant transcription factor databasePhosPhAt: the Arabidopsis thaliana phosphorylation site database. An updateSALAD database: a motif-based database of protein annotations for plant comparative genomicsSoyBase, the USDA-ARS soybean genetics and genomics database AntigenDB: an immunoinformatics database of pathogen antigensThe Immune Epitope Database 2.0IPD—the Immuno Polymorphism DatabaseArchiving next generation sequencing dataWEB ADDRESS/nar/database/a/ http://www.nbrp.jphttp://www.ddbj.nig.ac.jp/ena/patentdata/nr/http://aclame.ulb.ac.behttp://www-genome.biotoul.fr/ISbrowser.php http://recode.ucc.ier.it/http://ercsb.ewha.ac.kr/fusiongenehttp://h-invitational.jp/h-dbas/http://floresta.eead.csic.es/3dfootprint http://dbtss.hgc.jp//transmir/http://www.microrna.gr/mirgen//hhmd/prosite//http://projects.biotec.tu-dresden.de/memotif http://eggnog.embl.dehttp://InParanoid.sbc.su.se/http://mips.gsf.de/simap/http://bioinformatics.charite.de/supercyp http://www.funsimmat.de/chebi/http://its2.bioapps.biozentrum.uniwuerzburg.de /NNDBhttp://antares.protres.ru/comsin//fpop///pdbe/http://bioinfo.tg.fh-giessen.de/pdbselect/ /http://ptgl.zib.de/pn-utilityhttp://www.genome.jp/kegg/http://mbgd.genome.ad.jp//http://genome.kazusa.or.jp/cyanobase//http://metasystems.riken.jp/metabiome/ http://www.smpdb.ca//biosystems/http://mips.helmholtzmuenchen.de/genre/proj/co rum/index.htmlbr.utoronto.ca//Structure/ibis/ibi s.html/intacthttp://mint.bio.uniroma2.it/minthttp://mips.helmholtz-muenchen.de/proj/ppi/negatomehttp://stitch.embl.de////http://www.h-invitational.jp//wiki/Portal:Gene_Wiki //cosmic//index//bio/hlunghttp://bio.ifom-ieo-campus.it/ncg/hrtbldb/gxa/emage//genesigdb http://affymetrix2.bioinf.fbb.msu.ru/peptidome /pride /http://www.bicnirrh.res.in/antimicrobial .sg/group/cjttd/TTD.asp /primerbank/ /PMRDhttp://plntfdb.bio.uni-potsdam.de/v3.0/ phosphat.mpimp-golm.mpg.dehttp://salad.dna.affrc.go.jp/salad/http://www.imtech.res.in/raghava/antigendb /ipd//Traces/sra。

复合精细多尺度熵 python实现

复合精细多尺度熵 Python实现一、简介复合精细多尺度熵（CMSE）是一种用于研究时间序列数据复杂性的方法，它结合了多粒度分析和非线性动力学方法。

CMSE能够在不同时间尺度上揭示时间序列数据的非线性结构与复杂性，对于信号处理、生物医学工程和金融等领域具有重要的应用价值。

二、CMSE基本原理1. 复合精细多尺度熵的概念复合精细多尺度熵是通过对时间序列数据进行多尺度分解，然后在不同时间尺度上计算局部熵值，并将这些局部熵值进行复合得出最终的CMSE值。

CMSE包含了时间序列数据的多尺度信息，能够更全面地反映数据的复杂性。

2. CMSE的计算方法在Python中，可以使用pyentrp库来实现CMSE的计算。

需要对时间序列数据进行多尺度分解，可以使用小波变换或滑动窗口等方法。

在每个时间尺度上计算局部熵值，最后将这些局部熵值进行复合得出CMSE值。

三、CMSE在实际应用中的意义1. 信号处理领域在信号处理中，CMSE可以用于分析复杂信号的动态特性，对于识别非线性系统和复杂动力学系统具有重要作用。

2. 生物医学工程领域在生物医学工程中，CMSE可以用于分析生物医学信号的复杂性，例如心电图信号、脑电图信号等，有助于理解生物系统的动力学特性。

3. 金融领域在金融领域，CMSE可以用于分析金融时间序列数据的复杂性，对于预测金融市场的波动和风险具有重要意义。

四、个人观点和理解复合精细多尺度熵是一种很有潜力的分析方法，它能够在不同时间尺度上反映数据的复杂性和动态特性。

在Python中实现CMSE的计算可以帮助研究者更好地理解时间序列数据的内在结构，对于应用在不同领域具有重要意义。

我个人认为，随着对CMSE方法的深入研究和应用，将会有更多精彩的发现和应用出现。

五、总结通过本文的介绍，我们了解了复合精细多尺度熵的基本原理和在Python中的实现方法，以及在实际应用中的意义和个人观点。

CMSE 作为一种新兴的分析方法，有着广阔的应用前景。

特征更新的动态图卷积表面损伤点云分割方法

第41卷　第4期吉林大学学报(信息科学版)Vol.41　No.42023年7月Journal of Jilin University (Information Science Edition)July 2023文章编号:1671⁃5896(2023)04⁃0621⁃10特征更新的动态图卷积表面损伤点云分割方法收稿日期:2022⁃09⁃21基金项目:国家自然科学基金资助项目(61573185)作者简介:张闻锐(1998 ),男,江苏扬州人,南京航空航天大学硕士研究生,主要从事点云分割研究,(Tel)86⁃188****8397(E⁃mail)839357306@;王从庆(1960 ),男,南京人,南京航空航天大学教授,博士生导师,主要从事模式识别与智能系统研究,(Tel)86⁃130****6390(E⁃mail)cqwang@㊂张闻锐,王从庆(南京航空航天大学自动化学院,南京210016)摘要:针对金属部件表面损伤点云数据对分割网络局部特征分析能力要求高,局部特征分析能力较弱的传统算法对某些数据集无法达到理想的分割效果问题,选择采用相对损伤体积等特征进行损伤分类,将金属表面损伤分为6类,提出一种包含空间尺度区域信息的三维图注意力特征提取方法㊂将得到的空间尺度区域特征用于特征更新网络模块的设计,基于特征更新模块构建出了一种特征更新的动态图卷积网络(Feature Adaptive Shifting⁃Dynamic Graph Convolutional Neural Networks)用于点云语义分割㊂实验结果表明,该方法有助于更有效地进行点云分割,并提取点云局部特征㊂在金属表面损伤分割上,该方法的精度优于PointNet ++㊁DGCNN(Dynamic Graph Convolutional Neural Networks)等方法,提高了分割结果的精度与有效性㊂关键词:点云分割;动态图卷积;特征更新;损伤分类中图分类号:TP391.41文献标志码:A Cloud Segmentation Method of Surface Damage Point Based on Feature Adaptive Shifting⁃DGCNNZHANG Wenrui,WANG Congqing(School of Automation,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)Abstract :The cloud data of metal part surface damage point requires high local feature analysis ability of the segmentation network,and the traditional algorithm with weak local feature analysis ability can not achieve the ideal segmentation effect for the data set.The relative damage volume and other features are selected to classify the metal surface damage,and the damage is divided into six categories.This paper proposes a method to extract the attention feature of 3D map containing spatial scale area information.The obtained spatial scale area feature is used in the design of feature update network module.Based on the feature update module,a feature updated dynamic graph convolution network is constructed for point cloud semantic segmentation.The experimental results show that the proposed method is helpful for more effective point cloud segmentation to extract the local features of point cloud.In metal surface damage segmentation,the accuracy of this method is better than pointnet++,DGCNN(Dynamic Graph Convolutional Neural Networks)and other methods,which improves the accuracy and effectiveness of segmentation results.Key words :point cloud segmentation;dynamic graph convolution;feature adaptive shifting;damage classification 0　引　言基于深度学习的图像分割技术在人脸㊁车牌识别和卫星图像分析领域已经趋近成熟,为获取物体更226吉林大学学报(信息科学版)第41卷完整的三维信息,就需要利用三维点云数据进一步完善语义分割㊂三维点云数据具有稀疏性和无序性,其独特的几何特征分布和三维属性使点云语义分割在许多领域的应用都遇到困难㊂如在机器人与计算机视觉领域使用三维点云进行目标检测与跟踪以及重建;在建筑学上使用点云提取与识别建筑物和土地三维几何信息;在自动驾驶方面提供路面交通对象㊁道路㊁地图的采集㊁检测和分割功能㊂2017年,Lawin等[1]将点云投影到多个视图上分割再返回点云,在原始点云上对投影分割结果进行分析,实现对点云的分割㊂最早的体素深度学习网络产生于2015年,由Maturana等[2]创建的VOXNET (Voxel Partition Network)网络结构,建立在三维点云的体素表示(Volumetric Representation)上,从三维体素形状中学习点的分布㊂结合Le等[3]提出的点云网格化表示,出现了类似PointGrid的新型深度网络,集成了点与网格的混合高效化网络,但体素化的点云面对大量点数的点云文件时表现不佳㊂在不规则的点云向规则的投影和体素等过渡态转换过程中,会出现很多空间信息损失㊂为将点云自身的数据特征发挥完善,直接输入点云的基础网络模型被逐渐提出㊂2017年,Qi等[4]利用点云文件的特性,开发了直接针对原始点云进行特征学习的PointNet网络㊂随后Qi等[5]又提出了PointNet++,针对PointNet在表示点与点直接的关联性上做出改进㊂Hu等[6]提出SENET(Squeeze⁃and⁃Excitation Networks)通过校准通道响应,为三维点云深度学习引入通道注意力网络㊂2018年,Li等[7]提出了PointCNN,设计了一种X⁃Conv模块,在不显著增加参数数量的情况下耦合较远距离信息㊂图卷积网络[8](Graph Convolutional Network)是依靠图之间的节点进行信息传递,获得图之间的信息关联的深度神经网络㊂图可以视为顶点和边的集合,使每个点都成为顶点,消耗的运算量是无法估量的,需要采用K临近点计算方式[9]产生的边缘卷积层(EdgeConv)㊂利用中心点与其邻域点作为边特征,提取边特征㊂图卷积网络作为一种点云深度学习的新框架弥补了Pointnet等网络的部分缺陷[10]㊂针对非规律的表面损伤这种特征缺失类点云分割,人们已经利用各种二维图像采集数据与卷积神经网络对风扇叶片㊁建筑和交通工具等进行损伤检测[11],损伤主要类别是裂痕㊁表面漆脱落等㊂但二维图像分割涉及的损伤种类不够充分,可能受物体表面污染㊁光线等因素影响,将凹陷㊁凸起等损伤忽视,或因光照不均匀判断为脱漆㊂笔者提出一种基于特征更新的动态图卷积网络,主要针对三维点云分割,设计了一种新型的特征更新模块㊂利用三维点云独特的空间结构特征,对传统K邻域内权重相近的邻域点采用空间尺度进行区分,并应用于对金属部件表面损伤分割的有用与无用信息混杂的问题研究㊂对邻域点进行空间尺度划分,将注意力权重分组,组内进行特征更新㊂在有效鉴别外邻域干扰特征造成的误差前提下,增大特征提取面以提高局部区域特征有用性㊂1　深度卷积网络计算方法1.1　包含空间尺度区域信息的三维图注意力特征提取方法由迭代最远点采集算法将整片点云分割为n个点集:{M1,M2,M3, ,M n},每个点集包含k个点:{P1, P2,P3, ,P k},根据点集内的空间尺度关系,将局部区域划分为不同的空间区域㊂在每个区域内,结合局部特征与空间尺度特征,进一步获得更有区分度的特征信息㊂根据注意力机制,为K邻域内的点分配不同的权重信息,特征信息包括空间区域内点的分布和区域特性㊂将这些特征信息加权计算,得到点集的卷积结果㊂使用空间尺度区域信息的三维图注意力特征提取方式,需要设定合适的K邻域参数K和空间划分层数R㊂如果K太小,则会导致弱分割,因不能完全利用局部特征而影响结果准确性;如果K太大,会增加计算时间与数据量㊂图1为缺损损伤在不同参数K下的分割结果图㊂由图1可知,在K=30或50时,分割结果效果较好,K=30时计算量较小㊂笔者选择K=30作为实验参数㊂在分析确定空间划分层数R之前,简要分析空间层数划分所应对的问题㊂三维点云所具有的稀疏性㊁无序性以及损伤点云自身噪声和边角点多的特性,导致了点云处理中可能出现的共同缺点,即将离群值点云选为邻域内采样点㊂由于损伤表面多为一个面,被分割出的损伤点云应在该面上分布,而噪声点则被分布在整个面的两侧,甚至有部分位于损伤内部㊂由于点云噪声这种立体分布的特征,导致了离群值被选入邻域内作为采样点存在㊂根据采用DGCNN(Dynamic Graph Convolutional Neural Networks)分割网络抽样实验结果,位于切面附近以及损伤内部的离群值点对点云分割结果造成的影响最大,被错误分割为特征点的几率最大,在后续预处理过程中需要对这种噪声点进行优先处理㊂图1　缺损损伤在不同参数K 下的分割结果图Fig.1　Segmentation results of defect damage under different parameters K 基于上述实验结果,在参数K =30情况下,选择空间划分层数R ㊂缺损损伤在不同参数R 下的分割结果如图2所示㊂图2b 的结果与测试集标签分割结果更为相似,更能体现损伤的特征,同时屏蔽了大部分噪声㊂因此,选择R =4作为实验参数㊂图2　缺损损伤在不同参数R 下的分割结果图Fig.2　Segmentation results of defect damage under different parameters R 在一个K 邻域内,邻域点与中心点的空间关系和特征差异最能表现邻域点的权重㊂空间特征系数表示邻域点对中心点所在点集的重要性㊂同时,为更好区分图内邻域点的权重,需要将整个邻域细分㊂以空间尺度进行细分是较为合适的分类方式㊂中心点的K 邻域可视为一个局部空间,将其划分为r 个不同的尺度区域㊂再运算空间注意力机制,为这r 个不同区域的权重系数赋值㊂按照空间尺度多层次划分,不仅没有损失核心的邻域点特征,还能有效抑制无意义的㊁有干扰性的特征㊂从而提高了深度学习网络对点云的局部空间特征的学习能力,降低相邻邻域之间的互相影响㊂空间注意力机制如图3所示,计算步骤如下㊂第1步,计算特征系数e mk ㊂该值表示每个中心点m 的第k 个邻域点对其中心点的权重㊂分别用Δp mk 和Δf mk 表示三维空间关系和局部特征差异,M 表示MLP(Multi⁃Layer Perceptrons)操作,C 表示concat 函数,其中Δp mk =p mk -p m ,Δf mk =M (f mk )-M (f m )㊂将两者合并后输入多层感知机进行计算,得到计算特征系数326第4期张闻锐,等:特征更新的动态图卷积表面损伤点云分割方法图3　空间尺度区域信息注意力特征提取方法示意图Fig.3　Schematic diagram of attention feature extraction method for spatial scale regional information e mk =M [C (Δp mk ‖Δf mk )]㊂(1) 第2步,计算图权重系数a mk ㊂该值表示每个中心点m 的第k 个邻域点对其中心点的权重包含比㊂其中k ∈{1,2,3, ,K },K 表示每个邻域所包含点数㊂需要对特征系数e mk 进行归一化,使用归一化指数函数S (Softmax)得到权重多分类的结果,即计算图权重系数a mk =S (e mk )=exp(e mk )/∑K g =1exp(e mg )㊂(2) 第3步,用空间尺度区域特征s mr 表示中心点m 的第r 个空间尺度区域的特征㊂其中k r ∈{1,2,3, ,K r },K r 表示第r 个空间尺度区域所包含的邻域点数,并在其中加入特征偏置项b r ,避免权重化计算的特征在动态图中累计单面误差指向,空间尺度区域特征s mr =∑K r k r =1[a mk r M (f mk r )]+b r ㊂(3) 在r 个空间尺度区域上进行计算,就可得到点m 在整个局部区域的全部空间尺度区域特征s m ={s m 1,s m 2,s m 3, ,s mr },其中r ∈{1,2,3, ,R }㊂1.2　基于特征更新的动态图卷积网络动态图卷积网络是一种能直接处理原始三维点云数据输入的深度学习网络㊂其特点是将PointNet 网络中的复合特征转换模块(Feature Transform),改进为由K 邻近点计算(K ⁃Near Neighbor)和多层感知机构成的边缘卷积层[12]㊂边缘卷积层功能强大,其提取的特征不仅包含全局特征,还拥有由中心点与邻域点的空间位置关系构成的局部特征㊂在动态图卷积网络中,每个邻域都视为一个点集㊂增强对其中心点的特征学习能力,就会增强网络整体的效果[13]㊂对一个邻域点集,对中心点贡献最小的有效局部特征的边缘点,可以视为异常噪声点或低权重点,可能会给整体分割带来边缘溢出㊂点云相比二维图像是一种信息稀疏并且噪声含量更大的载体㊂处理一个局域内的噪声点,将其直接剔除或简单采纳会降低特征提取效果,笔者对其进行低权重划分,并进行区域内特征更新,增强抗噪性能,也避免点云信息丢失㊂在空间尺度区域中,在区域T 内有s 个点x 被归为低权重系数组,该点集的空间信息集为P ∈R N s ×3㊂点集的局部特征集为F ∈R N s ×D f [14],其中D f 表示特征的维度空间,N s 表示s 个域内点的集合㊂设p i 以及f i 为点x i 的空间信息和特征信息㊂在点集内,对点x i 进行小范围内的N 邻域搜索,搜索其邻域点㊂则点x i 的邻域点{x i ,1,x i ,2, ,x i ,N }∈N (x i ),其特征集合为{f i ,1,f i ,2, ,f i ,N }∈F ㊂在利用空间尺度进行区域划分后,对空间尺度区域特征s mt 较低的区域进行区域内特征更新,通过聚合函数对权重最低的邻域点在图中的局部特征进行改写㊂已知中心点m ,点x i 的特征f mx i 和空间尺度区域特征s mt ,目的是求出f ′mx i ,即中心点m 的低权重邻域点x i 在进行邻域特征更新后得到的新特征㊂对区域T 内的点x i ,∀x i ,j ∈H (x i ),x i 与其邻域H 内的邻域点的特征相似性域为R (x i ,x i ,j )=S [C (f i ,j )T C (f i ,j )/D o ],(4)其中C 表示由输入至输出维度的一维卷积,D o 表示输出维度值,T 表示转置㊂从而获得更新后的x i 的426吉林大学学报(信息科学版)第41卷特征㊂对R (x i ,x i ,j )进行聚合,并将特征f mx i 维度变换为输出维度f ′mx i =∑[R (x i ,x i ,j )S (s mt f mx i )]㊂(5) 图4为特征更新网络模块示意图,展示了上述特征更新的计算过程㊂图5为特征更新的动态图卷积网络示意图㊂图4　特征更新网络模块示意图Fig.4　Schematic diagram of feature update network module 图5　特征更新的动态图卷积网络示意图Fig.5　Flow chart of dynamic graph convolution network with feature update 动态图卷积网络(DGCNN)利用自创的边缘卷积层模块,逐层进行边卷积[15]㊂其前一层的输出都会动态地产生新的特征空间和局部区域,新一层从前一层学习特征(见图5)㊂在每层的边卷积模块中,笔者在边卷积和池化后加入了空间尺度区域注意力特征,捕捉特定空间区域T 内的邻域点,用于特征更新㊂特征更新会降低局域异常值点对局部特征的污染㊂网络相比传统图卷积神经网络能获得更多的特征信息,并且在面对拥有较多噪声值的点云数据时,具有更好的抗干扰性[16],在对性质不稳定㊁不平滑并含有需采集分割的突出中心的点云数据时,会有更好的抗干扰效果㊂相比于传统预处理方式,其稳定性更强,不会发生将突出部分误分割或漏分割的现象[17]㊂2　实验结果与分析点云分割的精度评估指标主要由两组数据构成[18],即平均交并比和总体准确率㊂平均交并比U (MIoU:Mean Intersection over Union)代表真实值和预测值合集的交并化率的平均值,其计算式为526第4期张闻锐,等:特征更新的动态图卷积表面损伤点云分割方法U =1T +1∑Ta =0p aa ∑Tb =0p ab +∑T b =0p ba -p aa ,(6)其中T 表示类别,a 表示真实值,b 表示预测值,p ab 表示将a 预测为b ㊂总体准确率A (OA:Overall Accuracy)表示所有正确预测点p c 占点云模型总体数量p all 的比,其计算式为A =P c /P all ,(7)其中U 与A 数值越大,表明点云分割网络越精准,且有U ≤A ㊂2.1　实验准备与数据预处理实验使用Kinect V2,采用Depth Basics⁃WPF 模块拍摄金属部件损伤表面获得深度图,将获得的深度图进行SDK(Software Development Kit)转化,得到pcd 格式的点云数据㊂Kinect V2采集的深度图像分辨率固定为512×424像素,为获得更清晰的数据图像,需尽可能近地采集数据㊂选择0.6~1.2m 作为采集距离范围,从0.6m 开始每次增加0.2m,获得多组采量数据㊂点云中分布着噪声,如果不对点云数据进行过滤会对后续处理产生不利影响㊂根据统计原理对点云中每个点的邻域进行分析,再建立一个特别设立的标准差㊂然后将实际点云的分布与假设的高斯分布进行对比,实际点云中误差超出了标准差的点即被认为是噪声点[19]㊂由于点云数据量庞大,为提高效率,选择采用如下改进方法㊂计算点云中每个点与其首个邻域点的空间距离L 1和与其第k 个邻域点的空间距离L k ㊂比较每个点之间L 1与L k 的差,将其中差值最大的1/K 视为可能噪声点[20]㊂计算可能噪声点到其K 个邻域点的平均值,平均值高出标准差的被视为噪声点,将离群噪声点剔除后完成对点云的滤波㊂2.2　金属表面损伤点云关键信息提取分割方法对点云损伤分割,在制作点云数据训练集时,如果只是单一地将所有损伤进行统一标记,不仅不方便进行结果分析和应用,而且也会降低特征分割的效果㊂为方便分析和控制分割效果,需要使用ArcGIS 将点云模型转化为不规则三角网TIN(Triangulated Irregular Network)㊂为精确地分类损伤,利用图6　不规则三角网模型示意图Fig.6　Schematic diagram of triangulated irregular networkTIN 的表面轮廓性质,获得训练数据损伤点云的损伤内(外)体积,损伤表面轮廓面积等㊂如图6所示㊂选择损伤体积指标分为相对损伤体积V (RDV:Relative Damege Volume)和邻域内相对损伤体积比N (NRDVR:Neighborhood Relative Damege Volume Ratio)㊂计算相对平均深度平面与点云深度网格化平面之间的部分,得出相对损伤体积㊂利用TIN 邻域网格可获取某损伤在邻域内的相对深度占比,有效解决制作测试集时,将因弧度或是形状造成的相对深度判断为损伤的问题㊂两种指标如下:V =∑P d k =1h k /P d -∑P k =1h k /()P S d ,(8)N =P n ∑P d k =1h k S d /P d ∑P n k =1h k S ()n -()1×100%,(9)其中P 表示所有点云数,P d 表示所有被标记为损伤的点云数,P n 表示所有被认定为损伤邻域内的点云数;h k 表示点k 的深度值;S d 表示损伤平面面积,S n 表示损伤邻域平面面积㊂在获取TIN 标准包络网视图后,可以更加清晰地描绘损伤情况,同时有助于量化损伤严重程度㊂笔者将损伤分为6种类型,并利用计算得出的TIN 指标进行损伤分类㊂同时,根据损伤部分体积与非损伤部分体积的关系,制定指标损伤体积(SDV:Standard Damege Volume)区分损伤类别㊂随机抽选5个测试组共50张图作为样本㊂统计非穿透损伤的RDV 绝对值,其中最大的30%标记为凹陷或凸起,其余626吉林大学学报(信息科学版)第41卷标记为表面损伤,并将样本分类的标准分界值设为SDV㊂在设立以上标准后,对凹陷㊁凸起㊁穿孔㊁表面损伤㊁破损和缺损6种金属表面损伤进行分类,金属表面损伤示意图如图7所示㊂首先,根据损伤是否产生洞穿,将损伤分为两大类㊂非贯通伤包括凹陷㊁凸起和表面损伤,贯通伤包括穿孔㊁破损和缺损㊂在非贯通伤中,凹陷和凸起分别采用相反数的SDV 作为标准,在这之间的被分类为表面损伤㊂贯通伤中,以损伤部分平面面积作为参照,较小的分类为穿孔,较大的分类为破损,而在边缘处因腐蚀㊁碰撞等原因缺角㊁内损的分类为缺损㊂分类参照如表1所示㊂图7　金属表面损伤示意图Fig.7　Schematic diagram of metal surface damage表1　损伤类别分类Tab.1　Damage classification 损伤类别凹陷凸起穿孔表面损伤破损缺损是否形成洞穿××√×√√RDV 绝对值是否达到SDV √√\×\\S d 是否达到标准\\×\√\2.3　实验结果分析为验证改进的图卷积深度神经网络在点云语义分割上的有效性,笔者采用TensorFlow 神经网络框架进行模型测试㊂为验证深度网络对损伤分割的识别准确率,采集了带有损伤特征的金属部件损伤表面点云,对点云进行预处理㊂对若干金属部件上的多个样本金属面的点云数据进行筛选,删除损伤占比低于5%或高于60%的数据后,划分并装包制作为点云数据集㊂采用CloudCompare 软件对样本金属上的损伤部分进行分类标记,共分为6种如上所述损伤㊂部件损伤的数据集制作参考点云深度学习领域广泛应用的公开数据集ModelNet40part㊂分割数据集包含了多种类型的金属部件损伤数据,这些损伤数据显示在510张总点云图像数据中㊂点云图像种类丰富,由各种包含损伤的金属表面构成,例如金属门,金属蒙皮,机械构件外表面等㊂用ArcGIS 内相关工具将总图进行随机点拆分,根据数据集ModelNet40part 的规格,每个独立的点云数据组含有1024个点,将所有总图拆分为510×128个单元点云㊂将样本分为400个训练集与110个测试集,采用交叉验证方法以保证测试的充分性[20],对多种方法进行评估测试,实验结果由单元点云按原点位置重新组合而成,并带有拆分后对单元点云进行的分割标记㊂分割结果比较如图8所示㊂726第4期张闻锐,等:特征更新的动态图卷积表面损伤点云分割方法图8　分割结果比较图Fig.8　Comparison of segmentation results在部件损伤分割的实验中,将不同网络与笔者网络(FAS⁃DGCNN:Feature Adaptive Shifting⁃Dynamic Graph Convolutional Neural Networks)进行对比㊂除了采用不同的分割网络外,其余实验均采用与改进的图卷积深度神经网络方法相同的实验设置㊂实验结果由单一损伤交并比(IoU:Intersection over Union),平均损伤交并比(MIoU),单一损伤准确率(Accuracy)和总体损伤准确率(OA)进行评价,结果如表2~表4所示㊂将6种不同损伤类别的Accuracy 与IoU 进行对比分析,可得出结论:相比于基准实验网络Pointet++,笔者在OA 和MioU 方面分别在贯通伤和非贯通伤上有10%和20%左右的提升,在整体分割指标上,OA 能达到90.8%㊂对拥有更多点数支撑,含有较多点云特征的非贯通伤,几种点云分割网络整体性能均能达到90%左右的效果㊂而不具有局部特征识别能力的PointNet 在贯通伤上的表现较差,不具备有效的分辨能力,导致分割效果相对于其他损伤较差㊂表2　损伤部件分割准确率性能对比 Tab.2　Performance comparison of segmentation accuracy of damaged parts %实验方法准确率凹陷⁃1凸起⁃2穿孔⁃3表面损伤⁃4破损⁃5缺损⁃6Ponitnet 82.785.073.880.971.670.1Pointnet++88.786.982.783.486.382.9DGCNN 90.488.891.788.788.687.1FAS⁃DGCNN 92.588.892.191.490.188.6826吉林大学学报(信息科学版)第41卷表3　损伤部件分割交并比性能对比 Tab.3　Performance comparison of segmentation intersection ratio of damaged parts %IoU 准确率凹陷⁃1凸起⁃2穿孔⁃3表面损伤⁃4破损⁃5缺损⁃6PonitNet80.582.770.876.667.366.9PointNet++86.384.580.481.184.280.9DGCNN 88.786.589.986.486.284.7FAS⁃DGCNN89.986.590.388.187.385.7表4　损伤分割的整体性能对比分析　　出,动态卷积图特征以及有效的邻域特征更新与多尺度注意力给分割网络带来了更优秀的局部邻域分割能力,更加适应表面损伤分割的任务要求㊂3　结　语笔者利用三维点云独特的空间结构特征,将传统K 邻域内权重相近的邻域点采用空间尺度进行区分,并将空间尺度划分运用于邻域内权重分配上,提出了一种能将邻域内噪声点降权筛除的特征更新模块㊂采用此模块的动态图卷积网络在分割上表现出色㊂利用特征更新的动态图卷积网络(FAS⁃DGCNN)能有效实现金属表面损伤的分割㊂与其他网络相比,笔者方法在点云语义分割方面表现出更高的可靠性,可见在包含空间尺度区域信息的注意力和局域点云特征更新下,笔者提出的基于特征更新的动态图卷积网络能发挥更优秀的作用,而且相比缺乏局部特征提取能力的分割网络,其对于点云稀疏㊁特征不明显的非贯通伤有更优的效果㊂参考文献:[1]LAWIN F J,DANELLJAN M,TOSTEBERG P,et al.Deep Projective 3D Semantic Segmentation [C]∥InternationalConference on Computer Analysis of Images and Patterns.Ystad,Sweden:Springer,2017:95⁃107.[2]MATURANA D,SCHERER S.VoxNet:A 3D Convolutional Neural Network for Real⁃Time Object Recognition [C]∥Proceedings of IEEE /RSJ International Conference on Intelligent Robots and Systems.Hamburg,Germany:IEEE,2015:922⁃928.[3]LE T,DUAN Y.PointGrid:A Deep Network for 3D Shape Understanding [C]∥2018IEEE /CVF Conference on ComputerVision and Pattern Recognition (CVPR).Salt Lake City,USA:IEEE,2018:9204⁃9214.[4]QI C R,SU H,MO K,et al.PointNet:Deep Learning on Point Sets for 3D Classification and Segmentation [C]∥IEEEConference on Computer Vision and Pattern Recognition (CVPR).Hawaii,USA:IEEE,2017:652⁃660.[5]QI C R,SU H,MO K,et al,PointNet ++:Deep Hierarchical Feature Learning on Point Sets in a Metric Space [C]∥Advances in Neural Information Processing Systems.California,USA:SpringerLink,2017:5099⁃5108.[6]HU J,SHEN L,SUN G,Squeeze⁃and⁃Excitation Networks [C ]∥IEEE Conference on Computer Vision and PatternRecognition.Vancouver,Canada:IEEE,2018:7132⁃7141.[7]LI Y,BU R,SUN M,et al.PointCNN:Convolution on X⁃Transformed Points [C]∥Advances in Neural InformationProcessing Systems.Montreal,Canada:NeurIPS,2018:820⁃830.[8]ANH VIET PHAN,MINH LE NGUYEN,YEN LAM HOANG NGUYEN,et al.DGCNN:A Convolutional Neural Networkover Large⁃Scale Labeled Graphs [J].Neural Networks,2018,108(10):533⁃543.[9]任伟建,高梦宇,高铭泽,等.基于混合算法的点云配准方法研究[J].吉林大学学报(信息科学版),2019,37(4):408⁃416.926第4期张闻锐,等:特征更新的动态图卷积表面损伤点云分割方法036吉林大学学报(信息科学版)第41卷REN W J,GAO M Y,GAO M Z,et al.Research on Point Cloud Registration Method Based on Hybrid Algorithm[J]. Journal of Jilin University(Information Science Edition),2019,37(4):408⁃416.[10]ZHANG K,HAO M,WANG J,et al.Linked Dynamic Graph CNN:Learning on Point Cloud via Linking Hierarchical Features[EB/OL].[2022⁃03⁃15].https:∥/stamp/stamp.jsp?tp=&arnumber=9665104. [11]林少丹,冯晨,陈志德,等.一种高效的车体表面损伤检测分割算法[J].数据采集与处理,2021,36(2):260⁃269. LIN S D,FENG C,CHEN Z D,et al.An Efficient Segmentation Algorithm for Vehicle Body Surface Damage Detection[J]. Journal of Data Acquisition and Processing,2021,36(2):260⁃269.[12]ZHANG L P,ZHANG Y,CHEN Z Z,et al.Splitting and Merging Based Multi⁃Model Fitting for Point Cloud Segmentation [J].Journal of Geodesy and Geoinformation Science,2019,2(2):78⁃79.[13]XING Z Z,ZHAO S F,GUO W,et al.Processing Laser Point Cloud in Fully Mechanized Mining Face Based on DGCNN[J]. ISPRS International Journal of Geo⁃Information,2021,10(7):482⁃482.[14]杨军,党吉圣.基于上下文注意力CNN的三维点云语义分割[J].通信学报,2020,41(7):195⁃203. YANG J,DANG J S.Semantic Segmentation of3D Point Cloud Based on Contextual Attention CNN[J].Journal on Communications,2020,41(7):195⁃203.[15]陈玲,王浩云,肖海鸿,等.利用FL⁃DGCNN模型估测绿萝叶片外部表型参数[J].农业工程学报,2021,37(13): 172⁃179.CHEN L,WANG H Y,XIAO H H,et al.Estimation of External Phenotypic Parameters of Bunting Leaves Using FL⁃DGCNN Model[J].Transactions of the Chinese Society of Agricultural Engineering,2021,37(13):172⁃179.[16]柴玉晶,马杰,刘红.用于点云语义分割的深度图注意力卷积网络[J].激光与光电子学进展,2021,58(12):35⁃60. CHAI Y J,MA J,LIU H.Deep Graph Attention Convolution Network for Point Cloud Semantic Segmentation[J].Laser and Optoelectronics Progress,2021,58(12):35⁃60.[17]张学典,方慧.BTDGCNN:面向三维点云拓扑结构的BallTree动态图卷积神经网络[J].小型微型计算机系统,2021, 42(11):32⁃40.ZHANG X D,FANG H.BTDGCNN:BallTree Dynamic Graph Convolution Neural Network for3D Point Cloud Topology[J]. Journal of Chinese Computer Systems,2021,42(11):32⁃40.[18]张佳颖,赵晓丽,陈正.基于深度学习的点云语义分割综述[J].激光与光电子学,2020,57(4):28⁃46. ZHANG J Y,ZHAO X L,CHEN Z.A Survey of Point Cloud Semantic Segmentation Based on Deep Learning[J].Lasers and Photonics,2020,57(4):28⁃46.[19]SUN Y,ZHANG S H,WANG T Q,et al.An Improved Spatial Point Cloud Simplification Algorithm[J].Neural Computing and Applications,2021,34(15):12345⁃12359.[20]高福顺,张鼎林,梁学章.由点云数据生成三角网络曲面的区域增长算法[J].吉林大学学报(理学版),2008,46 (3):413⁃417.GAO F S,ZHANG D L,LIANG X Z.A Region Growing Algorithm for Triangular Network Surface Generation from Point Cloud Data[J].Journal of Jilin University(Science Edition),2008,46(3):413⁃417.(责任编辑:刘俏亮)。

生物科学外文翻译--柑桔属类胡萝卜素生物合成

中文5000汉字,2800单词，15900英文字符出处：Journal of AgricμLtural and Food Chemistry.2007, 55(18)：7405~7417Carotenoid Biosynthetic Pathway in the Citrus Genus: Number of Copies and Phylogenetic Diversity of Seven GeneAnneLaure Fanciullino,†,⊥，Claudie DhuiqueMayer,‡,⊥，François Luro,§,⊥，RaphaëlMorillon,†,⊥ and，Patrick Ollitrault,†The first objective of this paper was to analyze the potential role of allelic variability of carotenoid biosynthetic genes in the interspecifi diversity in carotenoid composition of Citrus juices. The second objective was to determine the number of copies for each of these genes. Seven carotenoid biosynthetic genes were analyzed using restriction fragment length polymorphism (RFLP) and simple sequence repeats (SSR) markers. RFLP analyses were performed with the genomic DNA obtained from 25 Citrus genotypes using several restriction enzymes. cDNA fragments of Psy, Pds, Zds, Lcyb, Lcy-e, Hy-b, and Zep genes labeled with [R-32P]dCTP were used as probes. For SSR analyses, two primer pairs amplifying two SSR sequences identified from expressed sequence tags (ESTs) of Lcy-b and Hy-b genes were designed. The number of copies of the seven genes ranged from one for Lcy-b to three for Zds. The genetic diversity revealed by RFLP and SSR profiles was in agreement with the genetic diversity obtained from neutral molecμLar markers. Genetic interpretation of RFLP and SSR profiles of four genes (Psy1, Pds1, Lcy-b, and Lcy-e1) enabled us to make inferences on the phylogenetic origin of alleles for the major commercial citrus species. Moreover, the resμLts of our analyses suggest that the allelic diversity observed at the locus of both of lycopene cyclase genes, Lcy-b and Lcy-e1, is associated with interspecific diversity in carotenoid accumμLation in Citrus. The interspecific differences in carotenoid contents previously reported to be associated with other key steps catalyzed by PSY, HY-b, and ZEP were not linked to specific alleles at the corresponding loci.KEYWORDS: Citrus; carotenoids; biosynthetic genes; allelic variability; phylogeny INTRODUCTIONCarotenoids are pigments common to all photosynthetic organisms. In pigment-protein complexes, they act as light sensors for photosynthesis but also prevent photo-oxidation induced by too strong light intensities. I n horticμLtural crops, they play a major role in fruit, root, or tuber coloration and in nutritional quality. Indeed some of these micronutrients are precursors of vitamin A, an essential component of human and animal diets. Carotenoids may also play a role in chronic disease prevention (such as certain cancers), probably due to their antioxidant properties. The carotenoid biosynthetic pathway is now well established. Carotenoids are synthesized in plastids by nuclear-encoded enzymes. The immediate precursor of carotenoids (and also of gibberellins, plastoquinone, chlorophylls,phylloquinones, and tocopherols) isgeranylgeranyl diphosphate (GGPP). In light-grown plants, GGPP is mainly derived from the methylerythritol phosphate (MEP) pathway). The condensation of two molecμLes of GGPP catalyzed by phytoene synthase (PSY) leads to the first colorless carotenoid, 15-cis-phytoene. Phytoene undergoes four desaturation reactions catalyzed by two enzymes, phytoene desaturase (PDS) and β-carotene desaturase (ZDS), which convert phytoene into the red-colored poly-cis-lycopene. Recently, Isaacson et al. and Park et al. isolated from tomato and Arabidopsis thaliana, respectively, the genes that encode the carotenoid isomerase (CRTISO) which, in turn, catalyzes the isomerization of poly-cis-carotenoids into all-trans-carotenoids. CRTISO acts on prolycopene to form all-trans lycopene, which undergoes cyclization reactions. Cyclization of lycopene is a branching point: one branch leads to β-carotene (β, β-carotene) and the other toα-carotene (β, ε-carotene). Lycopene β-cyclase (LCY-b) then converts lycopene intoβ-carotene in two steps, whereas the formation of α-carotene requires the action of twoenzymes, lycopene ε- cyclase (LCY-e) and lycopene β-cyclase (LCY-b). α- carotene is converted into lutein by hydroxylations catalyzed by ε-carotene hydroxylase (HY-e) andβ-carotene hydroxylase (HY-b). Other xanthophylls are produced fromβ-carotene with hydroxylation reactions catalyzed by HY-b and epoxydation catalyzed by zeaxanthin epoxidase (ZEP). Most of the carotenoid biosynthetic genes have been cloned and sequenced in Citrus varieties . However, our knowledge of the complex regμLation of carotenoid biosynthesis in Citrus fruit is still limited. We need further information on the number of copies of these genes and on their allelic diversity in Citrus because these can influence carotenoid composition within the Citrus genus.Citrus fruit are among the richest sources of carotenoids. The fruit generally display a complex carotenoid structure, and 115 different carotenoids have been identified in Citrus fruit. The carotenoid richness of Citrus flesh depends on environmental conditions, particμLarly on growing conditions and on geographical or igin . However the main factor influencing variability of caro tenoid quality in juice has been shown to be genetic diversity. Kato et al. showed that mandarin and orange juices accumμLated high levels of β-cryptoxanthin and violaxanthin, respectively, whereas mature lemon accumμLated extremely low levels of carotenoids. Goodner et al. demonstrated that mandarins, oranges, and their hybrids coμLd be clearly distinguished by their β-cryptoxanthin contents. Juices of red grapefruit contained two major carotenoids: lycopene and β-carotene. More recently, we conducted a broad study on the organization of the variability of carotenoid contents in different cμLtivated Citrus species in relation with the biosynthetic pathway . Qualitative analysis of presence or absence of the different compounds revealed three main clusters: (1) mandarins, sweet oranges, and sour oranges; (2) citrons, lemons, and limes; (3) pummelos and grapefruit. Our study also enabled identification of key steps in the diversification of the carotenoid profile. Synthesis of phytoene appeared as a limiting step for acid Citrus, while formation of β-carotene and R-carotene from lycopene were dramatically limited in cluster 3(pummelos and grapefruit). Only varieties in cluster 1 were able to produce violaxanthin. In the same study , we concluded that there was a very strong correlation between the classification of Citrus species based on the presence or absence of carotenoids (below, this classification is also referred to as the organization of carotenoid diversity) and genetic diversity evaluated with biochemical or molecμLar markers such as isozy mes or randomLy amplified polymorphic DNA (RAPD). We also concluded that, at the interspecific level, the organization of the diversity of carotenoid composition was linked to the global evolution process of cμLtivated Citrus rather than to more recent mut ation events or human selection processes. Indeed, at interspecific level, a correlation between phenotypic variability and genetic diversity is common and is generally associated with generalized gametic is common and is generally associated with generalized gametic disequilibrium resμLting from the history of cμLtivated Citrus. Thus from numerical taxonomy based on morphological traits or from analysis of molecμLar markers , all authors agreed on the existence of three basic taxa (C. reticμLata, mandarins; C. medica, citrons; and C. maxima, pummelos) whose differentiation was the resμLt of allopatric evolution. All other cμLtivated Citrus species (C. sinensis, sweet oranges; C. aurantium, sour oranges; C. paradisi, grapefruit; and C. limon, lemons) resμLte d from hybridization events within this basic pool except for C. aurantifolia, which may be a hybrid between C. medica and C. micrantha .Our previous resμLts and data on Citrus evolution lead us to propose the hypothesis that the allelic variability supporting the organization of carotenoid diversity at interspecific level preceded events that resμLted in the creation of secondary species. Such molecμLar variability may have two different effects: on the one hand, non-silent substitutions in coding region affect the specific activity of corresponding enzymes of the biosynthetic pathway, and on the other hand, variations in untranslated regions affect transcriptional or post-transcriptional mechanisms.There is no available data on the allelic diversity of Citrus genes of the carotenoid biosynthetic pathway. The objective of this paper was to test the hypothesis that allelic variability of these genes partially determines phenotypic variability at the interspecific level. For this purpose, we analyzed the RFLPs around seven genes of the biosynthetic pathway of carotenoids (Psy, Pds, Zds, Lcy-b, Lcy-e, Hy-b, Zep) and the polymorphism of two SSR sequences found in Lcy-b and Hy-b genes in a representative set of varieties of the Citrus genus already analyzed for carotenoid constitution. Our study aimed to answer the following questions: (a) are those genes mono- or mμLtilocus, (b) is the polymorphism revealed by RFLP and SSR markers in agreement with the general history of cμLtivated Citrus thus permitting inferen ces about the phylogenetic origin of genes of the secondary species, and (c) is this polymorphism associated with phenotypic (carotenoid compound) variations.RESΜLTS AND DISCUSSIONGlobal Diversity of the Genotype Sample Observed by RFLP Analysis. RFLP analyses were performed using probes defined from expressed sequences of seven majorgenes of the carotenoid biosynthetic pathway . One or two restriction enzymes were used for each gene. None of these enzymes cut the cDNA probe sequence except HindIII for the Lcy-e gene. Intronic sequences and restriction sites on genomic sequences were screened with PCR amplification using genomic DNA as template and with digestion of PCR products. The resμLts indicated the absence of an intronic sequence for Psy and Lcy-b fragments. The absence of intron in these two fragments was checked by cloning and sequencing corresponding genomic sequences (data not shown). Conversely, we found introns in Pds, Zds, Hy-b, Zep, and Lcy-e genomic sequences corresponding to RFLP probes. EcoRV did not cut the genomic sequences of Pds, Zds, Hy-b, Zep, and Lcy-e. In the same way, no BamHI restriction site was found in the genomic sequences of Pds, Zds, and Hy-b. Data relative to the diversity observed for the different genes are presented in Table 4. A total of 58 fragments were identified, six of them being monomorphic (present in all individuals). In the limited sample of the three basic taxa, only eight bands out of 58 coμLd not be observed. In the basic taxa, the mean number of bands per genotype observed was 24.7, 24.7, and 17 for C. reticμLata, C. maxima, and C. medica, respectively. It varies from 28 (C. limettioides) to 36 (C. aurantium) for the secondary species. The mean number of RFLP bands per individual was lower for basic taxa than for the group of secondary species. This resμLt indicates that secondary species are much more heterozygous than the basic ones for these genes, which is logical if we assume that the secondary species arise from hybridizations between the three basic taxa. Moreover C. medica appears to be the least heterozygous taxon for RFLP around the genes of the carotenoid biosynthetic pathway, as already shown with isozymes, RAPD, and SSR markers.The two lemons were close to the acid Citrus cluster and the three sour oranges close to the mandarins/sweet oranges cluster. This organization of genetic diversity based on the RFLP profiles obtained with seven genes of the carotenoid pathway is very similar to that previously obtained with neutral molecμLar markers such as genomic SSR as well as the organization obtained with qualitative carotenoi d compositions. All these resμLts suggest that the observed RFLP and SSR fragments are good phylogenetic markers. It seems consistent with our basic hypothesis that major differentiation in the genes involved in the carotenoid biosynthetic pathway preceded the creation of the secondary hybrid species and thus that the allelic structure of these hybrid species can be reconstructed from alleles observed in the three basic taxa.Gene by Gene Analysis: The Psy Gene. For the Psy probe combined with EcoRV or BamHI restriction enzymes, five bands were identified for the two enzymes, and two to three bands were observed for each genotype. One of these bands was present in all individuals. There was no restriction site in the probe sequence. These resμLts lead us to believe that Psy is present at two loci, one where no polymorphism was found with the restriction enzymes used, and one that displayed polymorphism. The number of different profiles observed was six and four with EcoRV and BamHI, respectively, for a total of 10 different profiles among the 25 individuals .Two Psy genes have also been found in tomato, tobacco, maize, and rice . Conversely, only one Psy gene has been found in Arabidopsis thaliana and in pepper (Capsicum annuum), which also accumμLatescarotenoids in fruit. According to Bartley and Scolnik, Psy1 was expressed in tomato fruit chromoplasts, while Psy2 was specific to leaf tissue. In the same way, in Poaceae (maize, rice), Gallagher et al. found that Psy gene was duplicated and that Psy1 and not Psy2 transcripts in endosperm correlated with endosperm carotenoid accumμLation. These resμLts underline the role of gene duplication and the importance of tissue-specific phytoene synthase in the regμLation of carotenoid accumμLation.All the polymorphic bands were present in the sample of the basic taxon genomes. Assuming the hypothesis that all these bands describe the polymorphism at the same locus for the Psy gene, we can conclude that we found allelic differentiation between the three basic taxa with three alleles for C. reticμLata, four for C. maxima, and one for C. medica.The alleles observed for the basic taxa then enabled us to determine the genotypes of all the other species. The presumed genotypes for the Psy polymorphic locus are given in Table 7. Sweet oranges and grapefruit were heterozygous with one mandarin and one pummelo allele. Sour oranges were heterozygous; they shared the same mandarin allele with sweet oranges but had a different pummelo allele. Clementine was heterozygous with two mandarin alleles; one shared with sweet oranges and one with “Willow leaf”mandarin. “Meyer” lemon was heterozygous, with the mandarin allele also found in sweet oranges, and the citron allele. “Eureka”lemon was also heterozygous with the same pummelo allele as sour oranges and the citron allele. The other acid Citrus were homozygous for the citron allele.The Pds Gen. For the Pds probe combined with EcoRV, six different fragments were observed. One was common to all individuals. The number of fragments per individual was two or three. ResμLts for Pds led us to believe that this gene is present at two loci, one where no polymorphism was found with EcoRV restriction, and one displaying polymorphism. Conversely, studies on Arabidopsis, tomato, maize, and rice showed that Pds was a single copy gene. However, a previous study on Citrus suggests that Pds is present as a low-copy gene family in the Citrus genome, which is in agreement with our findings.The Zds Gene. The Zds profiles were complex. Nine and five fragments were observed with EcoRV and BamHI restriction, respectively. For both enzymes, one fragment was common to all individuals. The number of fragments per individual ranged from two to six for EcoRV and three to five for BamHI. There was no restriction site in the probe sequence. It can be assumed that several copies (at least three) of the Zds gene are present in the Citrus genome with polymorphism for at least two of them. In Arabidopsis, maize, and rice, like Pds, Zds was a single-copy gene .In these conditions and in the absence of analysis of controlled progenies, we are unable to conduct genetic analysis of profiles. However it appears that some bands differentiated the basic taxa: one for mandarins, one for pummelos, and one for citrons with EcoRV restriction and one for pummelos and one for citrons with BamHI restriction. Two bands out of the nine obtained with EcoRV were not observed in the samples of basic taxa. One was rare and only observed in “Rangpur” lime. The other was found in sour oranges, “V olkamer” lemon,and “Palestine sweet” lime sug gesting acommon ancestor for these three genotypes.This is in agreement with the assumption of Nicolosi et al. that “V olkamer” lemon resμLts from a complex hybrid combination with C. aurantium as one parent. It will be necessary to extend the analysis of the basic taxa to conclude whether these specific bands are present in the diversity of these taxa or resμLt from mutations after the formation of the secondary species.The Lcy-b Gene with RFLP Analysis.After restriction with EcoRV and hybridization with the Lcy-b probe, we obtained simple profiles with a total of four fragments. One to two fragments were observed for each individual, and seven profiles were differentiated among the 25 genotypes. These resμLts provide evidence that Lcy-b is present at a single locus in the haploid Citrus genome. Two lycopene β-cyclases encoded by two genes have been identified in tomato. The B gene encoded a novel type of lycopene β-cyclase whose sequence was similar to capsanthin-capsorubin synthase. The B gene expressed at a high level in βmutants was responsible for strong accumμLation ofβ-carotene in fruit, while in wild-type tomatoes, B was expressed at a low level.The Lcy-b Gene with SSR Analysis. Four bands were detected at locus 1210 (Lcy-b gene). One or two bands were detected per variety confirming that this gene is mono locus. Six different profiles were observed among the 25 genotypes. As with RFLP analysis, no intrataxon molecμLar polymorphism was found within C. Paradisi, C. Sinensis, and C. Aurantium.Taken together, the information obtained from RFLP and SSR analyses enabled us to identify a complete differentiation among the three basic taxon samples. Each of these taxons displayed two alleles for the analyzed sample. An additional allele was identified for “Mexican” lime. The profiles for all secondary species can be reconstructed from these alleles. Deduced genetic structure is given in. Sweet oranges and clementine were heterozygous with one mandarin and one pummelo allele. Sour oranges were also heterozygous sharing the same mandarin allele as sweet oranges but with another pummelo allele. Grapefruit were heterozygous with two pummelo alleles. All the acid secondary species were heterozygous, having one allele from citrons and the other one from manda rins except for “Mexican” lime, which had a specific allele.柑桔属类胡萝卜素生物合成途径中七个基因拷贝数目及遗传多样性的分析Journal of AgricμLtural and Food Chemistry. 2007, 55(18): 7405~7417.摘要：本文的首要目标是分析类胡萝卜素生物合成相关等位基因在发生变异柑橘属类胡萝卜素组分种间差异的潜在作用；第二个目标是确定这些基因的拷贝数。

ros pcl的滤波算法 -回复

ros pcl的滤波算法-回复ROS（Robotic Operating System）是一个用于机器人开发的开源框架，提供了一系列丰富的软件库和工具，用于实现机器人的感知、控制、仿真和通信等功能。

而PCL（Point Cloud Library）是ROS中用于处理点云数据的强大且广泛使用的库。

PCL中包含了许多滤波算法，用于对点云数据进行降噪、平滑和下采样等处理。

本文将详细介绍PCL中的一些常用滤波算法。

1. 点云滤波背景介绍点云数据是三维空间中一系列离散的点的集合，这些点通常用于表示物体的形状和表面。

在进行机器人感知或三维重构时，点云数据往往包含大量的噪声和冗余信息，因此需要对其进行滤波处理。

滤波算法的目标是在保留重要信息的同时，去除噪声和冗余点，从而提高点云数据的质量和准确性。

2. PCL中的滤波算法PCL中提供了多种滤波算法，具体包括：直通滤波、离群点移除、统计滤波、高斯滤波、平滑滤波、体素网格滤波等。

下面将逐一介绍这些算法的原理和使用方法。

2.1 直通滤波（PassThrough Filter）直通滤波是一种常用的基础滤波算法，它通过设置截断范围（即过滤阈值）来剔除位于指定范围之外的点。

直通滤波器首先获取点云数据中某个轴的最小和最大值，然后将处于指定范围之外的点去除。

这一算法常用于移除掉落在机器人传感器盲区之外的点，或者是移除点云数据中的地面或天空等不感兴趣的区域。

使用StraightThrough filter的示例代码如下：pcl::PassThrough<pcl::PointXYZ> pass;pass.setInputCloud(cloud);pass.setFilterFieldName("z");pass.setFilterLimits(0.0, 1.0);pass.filter(*filtered_cloud);以上代码将输入点云数据设置为"cloud"，并使用“z”轴作为过滤字段。

融合改进UNet和迁移学习的棉花根系图像分割方法

2023 年 9 月第 5 卷第 3 期Sept.2023 Vol.5, No.3智慧农业（中英文） Smart Agriculture融合改进UNet和迁移学习的棉花根系图像分割方法唐辉1，王铭2，于秋实1，张佳茜1，刘连涛3，王楠1*（1.河北农业大学机电工程学院，河北保定071001，中国； 2.河北省教育考试院，河北石家庄050091，中国；3.河北农业大学农学院，河北保定071001，中国）摘要：［目的/意义］根系是植物组成的重要部分，其生长发育至关重要。

根系图像分割是根系表型分析的重要方法，受限于图像质量、复杂土壤环境、低效传统方法，根系图像分割存在一定挑战。

［方法］为提高根系图像分割的准确性和鲁棒性，本研究以UNet模型为基础，提出了一种多尺度特征提取根系分割算法，并结合数据增强和迁移学习进一步提高改进UNet模型的泛化性和通用性。

首先，获取棉花根系单一数据集和开源多作物混合数据集，基于单一数据集的消融试验测试多尺度特征提取模块（Conv_2+Add）的有效性，与UNet、PSPNet、SegNet、DeeplabV3Plus算法对比验证其优势。

基于混合数据集验证改进算法（UNet+Conv_2+Add）在迁移学习的优势。

［结果和讨论结果和讨论］］UNet+Conv_2+Add相比其他算法（UNet、PSPNet、SegNet、DeeplabV3Plus），mIoU、mRecall和根系F1调和平均值分别为81.62%、86.90%和78.39%。

UNet+Conv_2+Add算法的迁移学习相比于普通训练在根系的交并比（Intersection over Union，IoU）值提升1.25%，根系的Recall值提升1.79%，F1调和平均值提升0.92%，且模型的整体收敛速度快。

［结论］本研究采用的多尺度特征提取策略能准确、高效地分割根系，为作物根系表型研究提供重要的研究基础。

关键词：深度学习；根系图像分割；UNet；多尺度特征；迁移学习中图分类号：S24；TP181 文献标志码：A 文章编号：SA202308003引用格式：唐辉, 王铭, 于秋实, 张佳茜, 刘连涛, 王楠. 融合改进UNet和迁移学习的棉花根系图像分割方法[J]. 智慧农业(中英文), 2023, 5(3): 96-109. DOI：10.12133/j.smartag.SA202308003TANG Hui, WANG Ming, YU Qiushi, ZHANG Jiaxi, LIU Liantao, WANG Nan. Root image segmentation method based on improved UNet and transfer learning[J]. Smart Agriculture, 2023, 5(3): 96-109. DOI：10.12133/j.smartag.SA202308003 (in Chinese with English abstract)1 引言根系作为植物和外界环境交换的器官，包括代谢、吸收、矿物和有机物交换等，地上部植株的生长也受到地下根系影响［1, 2］。

光线跟踪的自适应多点迭代体素遍历算法

光线跟踪的自适应多点迭代体素遍历算法
冯海文;牛连强;刘晓明;付博文
【期刊名称】《沈阳工业大学学报》
【年(卷),期】2013(035)001
【摘要】光线跟踪算法是提高图形真实感的一种主要技术,为了提高光线跟踪算法的速度,提出了一种快速的三维直线均匀体素遍历算法.该算法借鉴光栅直线行程扫描转换的思想,依据直线斜率定义决策参数,利用迭代计算决策参数控制每一步的光线走向.与单点迭代算法不同的是,一条直线依据斜率被自适应地拆分成由多点组成的m-遍历,而决策参数仅需要针对m-遍历甚至由多个m-遍历组成的周期进行计算,从而有效地减少了运算量.理论分析和实验表明,该算法的运行速度比现存的最快单步算法提高约56％,大幅度提高了光线跟踪的效率,且仅使用简单的整数运算即可实现.
【总页数】8页(P85-92)
【作者】冯海文;牛连强;刘晓明;付博文
【作者单位】沈阳工业大学软件学院,沈阳110870
【正文语种】中文
【中图分类】TP301.6
【相关文献】
1.基于空盒自适应生成的动态场景光线跟踪计算 [J], 李静;王文成;吴恩华
2.自适应统计迭代重建与基于模型迭代重建算法在超低剂量儿童胸部CT中的比较
[J], 赵凯宇;宣伟玲;陆洪江
3.适用于迭代型去模糊算法的自适应迭代终止条件 [J], 江顺亮;姜尹楠;曾小霞;唐祎玲;徐少平
4.一种基于体素的光线跟踪算法的加速方法 [J], 孙朝辉;孙贞寿
5.基于模型迭代重建与自适应统计迭代重建的上腹部低剂量CT扫描图像质量和肝小病灶检出率的比较 [J], 马光明;张喜荣;钟晖;于勇;于楠;窦越群
因版权原因，仅展示原文概要，查看原文内容请购买。

北邮信通院数据结构实验报告三哈夫曼编码器之欧阳化创编

数据结构实验报告实验名称：实验三树——哈夫曼编/解码器学生姓名：班级：班内序号：学号：日期： 2014年12月11日1．实验要求利用二叉树结构实现赫夫曼编/解码器。

基本要求：1、初始化(Init)：能够对输入的任意长度的字符串s进行统计，统计每个字符的频度，并建立赫夫曼树2、建立编码表(CreateTable)：利用已经建好的赫夫曼树进行编码，并将每个字符的编码输出。

3、编码(Encoding)：根据编码表对输入的字符串进行编码，并将编码后的字符串输出。

4、译码(Decoding)：利用已经建好的赫夫曼树对编码后的字符串进行译码，并输出译码结果。

5、打印(Print)：以直观的方式打印赫夫曼树（选作）6、计算输入的字符串编码前和编码后的长度，并进行分析，讨论赫夫曼编码的压缩效果。

测试数据：I love data Structure, I love Computer。

I will try my best to study data Structure.提示：1、用户界面可以设计为“菜单”方式：能够进行交互。

2、根据输入的字符串中每个字符出现的次数统计频度，对没有出现的字符一律不用编码。

2. 程序分析2.1 存储结构Huffman树给定一组具有确定权值的叶子结点，可以构造出不同的二叉树，其中带权路径长度最小的二叉树称为Huffman树，也叫做最优二叉树。

weight lchild rchildparent 2-1-1-15-1-1-16-1-1-17-1-1-19-1-1-1weight lchild rchild parent2-1-155-1-156-1-167-1-169-1-17701713238165482967-12.2关键算法分析（1）计算出现字符的权值利用ASCII码统计出现字符的次数，再将未出现的字符进行筛选，将出现的字符及頻数存储在数组a[]中。

void Huffman::Init(){int nNum[256]= {0}; //记录每一个字符出现的次数int ch = cin.get();int i=0;while((ch!='\r') && (ch!='\n')){nNum[ch]++; //统计字符出现的次数str[i++] = ch; //记录原始字符串ch = cin.get(); //读取下一个字符}str[i]='\0';n = 0;for ( i=0;i<256;i++){if (nNum[i]>0) //若nNum[i]==0，字符未出现{l[n] = (char)i;a[n] = nNum[i];n++;}}}时间复杂度为O（1）；（2）创建哈夫曼树：算法过程：Huffman树采用顺序存储---数组；数组的前n个结点存储叶子结点，然后是分支结点，最后是根结点；首先初始化叶子结点元素—循环实现；以循环结构，实现分支结点的合成，合成规则按照huffman树构成规则进行。

MEGA_3.1分子进化遗传分析软件

分子进化分析软件MEGA 3.1的使用操作作者：Dxyer- lzf战友（山东大学生命科学学院微生物国家重点实验室）受益的话，请到下列地址给我投一票，谢谢☺/bbs/post/view?bid=73&id=5059255&sty=1&tpg=1&ppg=1&age=0#5059255进化树也称种系树，英文名叫“Phyligenetic tree”，主要通过蛋白质序列/核酸序列的同源性比较进而了解基因或物种进化发生的内在关系和规律。

完整的进化分析需要以下几个步骤[1]：A.了解进化树的一些基本的概念和定义；B.集合一个待分析的数据组；C.对数据组进行多重序列比对；D.比对结果的校正；E.进化树的构建:为了说明进化分析结果的真实性，往往需要对校正后的比对结果选择不同的算法、模型和程序进行进化树的构建（必要的话，可利用Gblocks：http://molevol.ibmb.csic.es/Gblocks/Gblocks.html等软件提取数据组中的保守区后再进行进化树构建）；F.树枝可信度评估：可利用bootstrap和interior-branch等测试方法；G.发表数据的优化及展示：可利用MEGA3.1软件打开多种软件所做的进化树并对其进行分枝合并、排序等优化，最后利用Photoshop、Macromedia Fireworks MX 等作图软件对图的清晰度进行优化以获得高质量的发表展示图.目前树状进化树的分析软件很多，其中MEGA 3.1（/）[2]最大的优点在于易于上手操作并可对图形进行优化展示，是分子进化分析最理想的软件之一。

但MEGA软件包中没有目前蛋白序列进化分析中推崇的“最大可能性（maximum likelyhood，ML）”和“贝叶斯推论（bayesian inferences）”这两种算法[3]。

ML算法可使用软件phyml（http://atgc.lirmm.fr/phyml/），贝叶斯分析使用软件MrBayes（/）。

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)。