Hough voting
Edge Detection
The gradient points in the direction of most rapid change in intensity
The gradient direction is given by:
• how does this relate to the direction of the edge?
2. Aggregation of edgels into extended edges (maybe parametric description)
Edgel detection
• Difference operators
• Parametric-model matchers
Edge is Where Change Occurs
(scaled by 4, offset +128)
filter demo
Gaussian - image filter
Gaussian
delta function
Laplacian of Gaussian
An edge is not a line...
How can we detect lines ?
– given a set of points (x,y), find all (m,b) such that y = mx + b
Finding lines in an image
y b
y0 x0
x
m
image space
Hough space
Connection between image (x,y) and Hough (m,b) spaces
This saves us one operation:
论表决权拘束协议
论表决权拘束协议摘要表决权是股东权利中极为重要的一项权利。
表决权拘束协议，从广义上讲，是股东之间或股东与第三人之间就表决权如何行使而达成的协议，是股东行使表决权的一种方式，是公司法与合同法碰撞的结晶。
在实践中，通过表决权拘束协议获得话语权的现象十分普遍。
因其灵活性和弹性，表决权拘束协议在英美德等国广为应用，并已经形成了比较完善的规范体系，而在我国，《公司法》中对此没有明确规定，在理论界探讨也不是很多，但在实践中已有尝试，如果出现纠纷，则会出现无法可依的情况。
因此，本文旨在对表决权拘束协议在各国的发展的考察和对制度自身的分析，结合我国的公司治理环境和实践的需要，试图为该制度的本土化理论和实践提供一些思路。
全文共分为绪论、正文和结论三部分，其中正文分为五章。
第一章为表决权拘束协议的概述。
各国学者对表决权拘束协议的概念都或多或少有不同的理解，而本文则采较为广义的含义，即表决权拘束协议是就表决权的行使而达成的协议，不仅包括以“表决权拘束协议”为名称的合同，也包括股东协议中的关于表决权行使拘束的条款。
继而本文分析了表决权拘束协议的运行机理，以及表决权拘束协议在公众公司、封闭公司中的静态、动态的功能。
第二章为表决权拘束协议在各国的发展状态的总结，即表决权拘束协议在各国都经历了一个从禁止到认可的过程。
表决权拘束协议之所以在各国都存在争议的核心问题是，表决权能否成为合同的客体？表决权是一种特殊的权利，其不同于传统民法中的人身权或者是物权，却与股东的身份紧密相连。
但表决权的客体化是意思自治的体现，是时代发展的需要，因而笔者赞成表决权可以作为表决权拘束协议的客体。
第三章是表决权拘束协议与相似制度的辨析。
虽然表决权拘束协议与表决权代理、表决权信托以及累积投票制等表决权相关的制度在功能上有很多相似之处，但是在制度构建上各有不同，且应用的时间和范围也不相同，因而它们不能相互替代。
而且法律应当尽可能为当事人提供多种法律选择，而不能因为功能上的相似性而否定其存在的必要。
姿态估计算法汇总基于RGB、RGB-D以及点云数据
姿态估计算法汇总基于RGB、RGB-D以及点云数据作者：Tom Hardy点击上⽅“3D视觉⼯坊”，选择“星标”⼲货第⼀时间送达作者⼁Tom Hardy@知乎编辑⼁3D视觉⼯坊姿态估计算法汇总|基于RGB、RGB-D以及点云数据主要有整体⽅式、霍夫投票⽅式、Keypoint-based⽅式、Dense Correspondence⽅式等。
实现⽅法：传统⽅法、深度学习⽅式。
数据不同：RGB、RGB-D、点云数据等；标注⼯具实现⽅式不同整体⽅式整体⽅法直接估计给定图像中物体的三维位置和⽅向。
经典的基于模板的⽅法构造刚性模板并扫描图像以计算最佳匹配姿态。
这种⼿⼯制作的模板对集群场景不太可靠。
最近，⼈们提出了⼀些基于深度神经⽹络的⽅法来直接回归相机或物体的6D姿态。
然⽽，旋转空间的⾮线性使得数据驱动的DNN难以学习和推⼴。
1.Discriminative mixture-of-templates for viewpoint classification2.Gradient response maps for realtime detection of textureless objects.paring images using the hausdorff distance4.Implicit 3d orientation learning for 6d object detection from rgb images.5.Instance- and Category-level 6D Object Pose Estimation基于模型2.Deep model-based 6d pose refinement in rgbKeypoint-based⽅式⽬前基于关键点的⽅法⾸先检测图像中物体的⼆维关键点，然后利⽤PnP算法估计6D姿态。
1.Surf: Speeded up robust features.2.Object recognition from local scaleinvariant features3.3d object modeling and recognition using local affine-invariant image descriptors and multi-view spatial constraints.5.Stacked hourglass networks for human pose estimation6.Making deep heatmaps robust to partial occlusions for 3d object pose estimation.7.Bb8: A scalable, accurate, robust to partial occlusion method for predicting the 3d poses of challenging objects without using depth8.Real-time seamless single shot 6d object pose prediction.9.Discovery of latent 3d keypoints via end-toend geometric reasoning.10.Pvnet: Pixel-wise voting network for 6dof pose estimation.Dense Correspondence/霍夫投票⽅式1.Independent object class detection using 3d feature maps.2.Depth encoded hough voting for joint object detection and shape recovery.3.aware object detection and pose estimation.4.Learning 6d object pose estimation using 3d object coordinates.5.Global hypothesis generation for 6d object pose estimation.6.Deep learning of local rgb-d patches for 3d object detection and 6d pose estimation.7.Cdpn: Coordinates-based disentangled pose network for real-time rgb-based 6-dof object pose estimation.8.Pix2pose: Pixel-wise coordinate regression of objects for 6d pose estimation.9.Normalized object coordinate space for categorylevel 6d object pose and size estimation.10.Recovering 6d object pose and predicting next-bestview in the crowd.基于分割深度学习相关⽅法1.PoseCNN: A convolutional neural network for 6d object pose estimation in cluttered scenes.2.Render for cnn: Viewpoint estimation in images using cnns trained with rendered 3d model views.6.Robust 6D Object Pose Estimation in Cluttered Scenesusing Semantic Segmentation and Pose Regression Networks - Arul Selvam Periyasamy, Max Schwarz, and Sven Behnke. [[Paper]数据格式不同根据数据格式的不同，⼜可分为基于RGB、RGB-D、点云数据的识别算法。
hough变换的英文文献5
indicate that the proposed method has achieved a much better performance than the previous variations of Hough transform.
Keywords- Hough transform; Line detection; Many-to-one mapping; Local sliding window neighborhood
College of Computer and Information Engineering Beijing Technology and Business University Beijing, China wanyl@
Abstract-The
Hough transform is a popular robust method
The Hough transform [1] is a popular robust statistical algorithm for extracting global features such as straight lines, circles, ellipses, etc., from an image, which is widely used in computer vision and pattern recognition. This algorithm is essentially a voting process where each point belonging to the patterns votes for all the possible patterns passing through that point. These votes are accumulated in an accumulator array called bins, and the pattern receiving the maximum votes is recognized as the desired pattern. Given an NxN binary edge image, straight lines are defined in Equation (1). (1) p x cos e + y sin e
Efficient RANSAC for Point-Cloud Shape Detection
Volume0(1981),Number0pp.1–12Efficient RANSAC for Point-Cloud Shape DetectionRuwen Schnabel Roland Wahl Reinhard Klein†Universität Bonn,Computer Graphics GroupAbstractIn this work we present an automatic algorithm to detect basic shapes in unorganized point clouds.The algorithm decomposes the point cloud into a concise,hybrid structure of inherent shapes and a set of remaining points.Each detected shape serves as a proxy for a set of corresponding points.Our method is based on random sampling and detects planes,spheres,cylinders,cones and tori.For models with surfaces composed of these basic shapes only,e.g.CAD models,we automatically obtain a representation solely consisting of shape proxies.We demonstratethat the algorithm is robust even in the presence of many outliers and a high degree of noise.The proposed method scales well with respect to the size of the input point cloud and the number and size of the shapes within the data.Even point sets with several millions of samples are robustly decomposed within less than a minute.Moreover the algorithm is conceptually simple and easy to implement.Application areas include measurement of physical parameters,scan registration,surface compression,hybrid rendering,shape classification,meshing, simplification,approximation and reverse engineering.Categories and Subject Descriptors(according to ACM CCS):I.4.8[Image Processing and Computer Vision]:Scene AnalysisShape;Surface Fitting;I.3.5[Computer Graphics]:Computational Geometry and Object ModelingCurve, surface,solid,and object representations1.IntroductionDue to the increasing size and complexity of geometric data sets there is an ever-growing demand for concise and mean-ingful abstractions of this data.Especially when dealing with digitized geometry,e.g.acquired with a laser scanner,no handles for modification of the data are available to the user other than the digitized points themselves.However,in or-der to be able to make use of the data effectively,the raw digitized data has to be enriched with abstractions and pos-sibly semantic information,providing the user with higher-level interaction possibilities.Only such handles can pro-vide the interaction required for involved editing processes, such as deleting,moving or resizing certain parts and hence can make the data more readily usable for modeling pur-poses.Of course,traditional reverse engineering approaches can provide some of the abstractions that we seek,but usu-ally reverse engineering focuses onfinding a reconstruction of the underlying geometry and typically involves quite te-dious user interaction.This is not justified in a setting where †e-mail:{schnabel,wahl,rk}@cs.uni-bonn.de a complete and detailed reconstruction is not required at all, or shall take place only after some basic editing operations have been applied to the data.On the other hand,detecting instances of a set of primitive geometric shapes in the point sampled data is a means to quickly derive higher levels of ab-straction.For example in Fig.1patches of primitive shapes provide a coarse approximation of the geometry that could be used to compress the point-cloud very effectively. Another problem arising when dealing with digitized geom-etry is the often huge size of the datasets.Therefore the efficiency of algorithms inferring abstractions of the data is of utmost importance,especially in interactive settings. Thus,in this paper we focus especially onfinding an effi-cient algorithm for point-cloud shape detection,in order to be able to deal even with large point-clouds.Our work is a high performance RANSAC[FB81]algorithm that is capa-ble to extract a variety of different types of primitive shapes, while retaining such favorable properties of the RANSAC paradigm as robustness,generality and simplicity.At the heart of our algorithm are a novel,hierarchically structured sampling strategy for candidate shape generation as well as a novel,lazy cost function evaluation scheme,which signif-c The Eurographics Association and Blackwell Publishing2007.Published by Blackwell Publishing,9600Garsington Road,Oxford OX42DQ,UK and350Main Street,Malden, MA02148,USA.(a)Original(b)ApproximationFigure1:The372detected shapes in the choir screen define a coarse approximation of the surface.icantly reduces overall computational cost.Our method de-tects planes,spheres,cylinders,cones and tori,but additional primitives are possible.The goal of our algorithm is to reli-ably extract these shapes from the data,even under adverse conditions such as heavy noise.As has been indicated above,our method is especially well suited in situations where geometric data is automatically acquired and users refrain from applying surface reconstruc-tion methods,either due to the data’s low quality or due to processing time constraints.Such constraints are typical for areas where high level model interaction is required,as is the case when measuring physical parameters or in interactive, semi-automatic segmentation and postprocessing.Further applications are,for instance,registering many scans of an object,where detecting corresponding primitive shapes in multiple scans can provide good initial matches.High compression rates for point clouds can be achieved if prim-itive shapes are used to represent a large number of points with a small set of parameters.Other areas that can benefit from primitive shape information include hybrid rendering and shape classification.Additionally,a fast shape extraction method as ours can serve as building block in applications such as meshing,simplification,approximation and reverse engineering and bears the potential of significant speed up.2.Previous workThe detection of primitive shapes is a common problem en-countered in many areas of geometry related computer sci-ence.Over the years a vast number of methods have been proposed which cannot all be discussed here in depth.In-stead,here we give a short overview of some of the most important algorithms developed in the differentfields.We treat the previous work on RANSAC algorithms separately in section2.1as it is of special relevance to our work. Vision In computer vision,the two most widely known methodologies for shape extraction are the RANSAC paradigm[FB81]and the Hough transform[Hou62].Both have been proven to successfully detect shapes in2D as well as3D.RANSAC and the Hough transform are reliable even in the presence of a high proportion of outliers,but lack of efficiency or high memory consumption remains their ma-jor drawback[IK88].For both schemes,many acceleration techniques have been proposed,but no one on its own,or combinations thereof,have been shown to be able to provide an algorithm as efficient as ours for the3D primitive shape extraction problem.The Hough transform maps,for a given type of parameter-ized primitive,every point in the data to a manifold in the pa-rameter space.The manifold describes all possible variants of the primitive that contain the original point,i.e.in practice each point casts votes for many cells in a discretized param-eter space.Shapes are extracted by selecting those parame-ter vectors that have received a significant amount of votes. If the parameter space is discretized naively using a simple grid,the memory requirements quickly become prohibitive even for primitives with a moderate number of parameters, such as,for instance,cones.Although several methods have been suggested to alleviate this problem[IK87][XO93]its major application area remains the2D domain where the number of parameters typically is quite small.A notable ex-ception is[VGSR04]where the Hough transform is used to detect planes in3D datasets,as3D planes still have only a small number of parameters.They also propose a two-step procedure for the Hough based detection of cylinders that uses estimated normals in the data points.In the vision community many approaches have been pro-posed for segmentation of range images with primitive shapes.When working on range images these algorithms usually efficiently exploit the implicitly given connectiv-ity information of the image grid in some kind of region growing or region merging step[FEF97][GBS03].This is a fundamental difference to our case,where we are given only an unstructured cloud of points that lacks any explicit connectivity information.In[LGB95]and[LJS97]shapes are found by concurrently growing different seed primi-tives from which a suitable subset is selected according to an MDL criterion(coined the recover-and-select paradigm). [GBS03]detect shapes using a genetic algorithm to optimize a robust MSACfitness function(see also sec.2.1).[MLM01]c The Eurographics Association and Blackwell Publishing2007.introduce involved non-linearfitting functions for primitive shapes that are able to handle geometric degeneracy in the context of recover-and-select segmentation.Another robust method frequently employed in the vision community is the tensor voting framework[MLT00]which has been applied to successfully reconstruct surface geome-try from extremely cluttered scenes.While tensor voting can compete with RANSAC in terms of robustness,it is,how-ever,inherently model-free and therefore cannot be applied to the detection of predefined types of primitive shapes. Reverse engineering In reverse engineering,surface re-covery techniques are usually based on either a separate seg-mentation step or on a variety of region growing algorithms [VMC97][SB95][BGV∗02].Most methods call for some kind of connectivity information and are not well equipped to deal with a large amount of outliers[VMC97].Also these approaches try tofind a shape proxy for every part of the pro-cessed surface with the intent of loading the reconstructed geometry information into a CAD application.[BMV01]de-scribe a system which reconstructs a boundary representa-tion that can be imported into a CAD application from an unorganized point-cloud.However,their method is based on finding a triangulation for the point-set,whereas the method presented in this work is able to operate directly on the input points.This is advantageous as computing a suitable tessela-tion may be extremely costly and becomes very intricate or even ill-defined when there is heavy noise in the data.We do not,however,intend to present a method implementing all stages of a typical reverse engineering process.Graphics In computer graphics,[CSAD04]have recently proposed a general variational framework for approximation of surfaces by planes,which was extended to a set of more elaborate shape proxies by[WK05].Their aim is not only to extract certain shapes in the data,but tofind a globally optimal representation of the object by a given number of primitives.However,these methods require connectivity in-formation and are,due to their exclusive use of least squares fitting,susceptible to errors induced by outliers.Also,the optimization procedure is computationally expensive,which makes the method less suitable for large data sets.The out-put of our algorithm,however,could be used to initialize the set of shape proxies used by these methods,potentially accelerating the convergence of the optimization procedure. While the Hough transform and the RANSAC paradigm have been mainly used in computer vision some applica-tions have also been proposed in the computer graphics com-munity.[DDSD03]employ the Hough transform to identify planes for billboard clouds for triangle data.They propose an extension of the standard Hough transform to include a compactness criterion,but due to the high computational de-mand of the Hough transform,the method exhibits poor run-time performance on large or complex geometry.[WGK05] proposed a RANSAC-based plane detection method for hy-brid rendering of point clouds.To facilitate an efficient plane detection,planes are detected only in the cells of a hier-archical space decomposition and therefore what is essen-tially one plane on the surface is approximated by several planar patches.While this is acceptable for their hybrid ren-dering technique,our methodfinds maximal surface patches in order to yield a more concise representation of the ob-ject.Moreover,higher order primitives are not considered in their approach.[GG04]detect so-called slippable shapes which is a superset of the shapes recognized by our method. They use the eigenvalues of a symmetric matrix derived from the points and their normals to determine the slippability of a point-set.Their detection is a bottom-up approach that merges small initial slippable surfaces to obtain a global de-composition of the model.However the computation of the eigenvalues is costly for large models,the method is sen-sitive to noise and it is hard to determine the correct size of the initial surface patches.A related approach is taken by [HOP∗05].They also use the eigenvalues of a matrix derived from line element geometry to classify surfaces.A RANSAC based segmentation algorithm is employed to detect several shapes in a point-cloud.The method is aimed mainly at mod-els containing small numbers of points and shapes as no opti-mizations or extensions to the general RANSAC framework are adopted.2.1.RANSACThe RANSAC paradigm extracts shapes by randomly draw-ing minimal sets from the point data and constructing cor-responding shape primitives.A minimal set is the smallest number of points required to uniquely define a given type of geometric primitive.The resulting candidate shapes are tested against all points in the data to determine how many of the points are well approximated by the primitive(called the score of the shape).After a given number of trials,the shape which approximates the most points is extracted and the algorithm continues on the remaining data.RANSAC ex-hibits the following,desirable properties:•It is conceptually simple,which makes it easily extensible and straightforward to implement•It is very general,allowing its application in a wide range of settings•It can robustly deal with data containing more than50% of outliers[RL93]Its major deficiency is the considerable computational de-mand if no further optimizations are applied.[BF81]apply RANSAC to extract cylinders from range data,[CG01]use RANSAC and the gaussian image tofind cylinders in3D point clouds.Both methods,though,do not consider a larger number of different classes of shape prim-itives.[RL93]describe an algorithm that uses RANSAC to detect a set of different types of simple shapes.However, their method was adjusted to work in the image domain orc The Eurographics Association and Blackwell Publishing2007.on range images,and they did not provide the optimization necessary for processing large unstructured3D data sets.A vast number of extensions to the general RANSAC scheme have been proposed.Among the more recent ad-vances,methods such as MLESAC[TZ00]or MSAC[TZ98] improve the robustness of RANSAC with a modified score function,but do not provide any enhancement in the perfor-mance of the algorithm,which is the main focus of our work. Nonetheless the integration of a MLESAC scoring function is among the directions of our future work.[Nis05]pro-poses an acceleration technique for the case that the num-ber of candidates isfixed in advance.As it is a fundamen-tal property of our setup that an unknown large number of possibly very small shapes has to be detected in huge point-clouds,the amount of necessary candidates cannot,however, be specified in advance.3.OverviewGiven a point-cloud P={p1,...,p N}with associated nor-mals{n1,...,n N}the output of our algorithm is a set of primitive shapesΨ={ψ1,...,ψn}with corresponding dis-joint sets of points Pψ1⊂P,...,Pψn⊂P and a set of re-maining points R=P\{Pψ1,...,Pψn}.Similar to[RL93]and[DDSD03],we frame the shape extraction problem as an optimization problem defined by a score function.The overall structure of our method is outlined in pseudo-code in algorithm1.In each iteration of the algorithm,the prim-itive with maximal score is searched using the RANSAC paradigm.New shape candidates are generated by randomly sampling minimal subsets of P using our novel sampling strategy(see sec.4.3).Candidates of all considered shape types are generated for every minimal set and all candidates are collected in the set C.Thus no special ordering has to be imposed on the detection of different types of shapes.After new candidates have been generated the one with the high-est score m is computed employing the efficient lazy score evaluation scheme presented in sec.4.5.The best candidate is only accepted if,given the size|m|(in number of points) of the candidate and the number of drawn candidates|C|, the probability P(|m|,|C|)that no better candidate was over-looked during sampling is high enough(see sec.4.2.1).We provide an analysis of our sampling strategy to derive a suit-able probability computation.If a candidate is accepted,the corresponding points P m are removed from P and the can-didates C m generated with points in P m are deleted from C. The algorithm terminates as soon as P(τ,|C|)for a user de-fined minimal shape sizeτis large enough.In our implementation we use a standard score function that counts the number of compatible points for a shape candi-date[RL93][GBS03].The function has two free parame-ters:εspecifies the maximum distance of a compatible point whileαrestricts the deviation of a points’normal from that of the shape.We also ensure that only points forming a con-nected component on the surface are considered(see sec.4.4).Algorithm1Extract shapes in the point cloud PΨ←/0{extracted shapes}C←/0{shape candidates}repeatC←C∪newCandidates(){see sec.4.1and4.3}m←bestCandidate(C){see sec.4.4}if P(|m|,|C|)>p t thenP←P\P m{remove points}Ψ←Ψ∪mC←C\C m{remove invalid candidates}end ifuntil P(τ,|C|)>p treturnΨ4.Our method4.1.Shape estimationAs mentioned above,the shapes we consider in this work are planes,spheres,cylinders,cones and tori which have be-tween three and seven parameters.Every3D-point p i sam-plefixes only one parameter of the shape.In order to reduce the number of required points we compute an approximate surface normal n i for each point[HDD∗92],so that the ori-entation gives us two more parameters per sample.That way it is possible to estimate each of the considered basic shapes from only one or two point samples.However,always using one additional sample is advantageous,because the surplus parameters can be used to immediately verify a candidate and thus eliminate the need of evaluating many relatively low scored shapes[MC02].Plane For a plane,{p1,p2,p3}constitutes a minimal set when not taking into account the normals in the points.To confirm the plausibility of the generated plane,the deviation of the plane’s normal from n1,n2,n3is determined and the candidate plane is accepted only if all deviations are less than the predefined angleα.Sphere A sphere is fully defined by two points with corre-sponding normal vectors.We use the midpoint of the short-est line segment between the two lines given by the points p1and p2and their normals n1and n2to define the center of the sphere c.We take r= p1−c + p2−c2as the sphere ra-dius.The sphere is accepted as a shape candidate only if all three points are within a distance ofεof the sphere and their normals do not deviate by more thanαdegrees.Cylinder To generate a cylinder from two points with nor-mals wefirst establish the direction of the axis with a= n1×n2.Then we project the two parametric lines p1+tn1 and p2+tn2along the axis onto the a·x=0plane and take their intersection as the center c.We set the radius to the dis-tance between c and p1in that plane.Again the cylinder isc The Eurographics Association and Blackwell Publishing2007.verified by applying the thresholds εand αto distance and normal deviation of the samples.Cone Although the cone,too,is fully defined by two points with corresponding normals,for simplicity we use all three points and normals in its generation.To derive the po-sition of the apex c ,we intersect the three planes defined by the point and normal pairs.Then the normal of the plane de-fined by the three points {c +p 1−c p 1−c ,...,c +p 3−c p 3−c }givesthe direction of the axis a .Now the opening angle ωis givenas ω=∑i arccos ((p i −c )·a )3.Afterwards,similar to above,the cone is verified before becoming a candidate shape.Torus Just as in the case of the cone we use one more point than theoretically necessary to ease the computations required for estimation,i.e.four point and normal pairs.The rotational axis of the torus is found as one of the up to two lines intersecting the four point-normal lines p i +λn i [MLM01].To choose between the two possible axes,a full torus is estimated for both choices and the one which causes the smaller error in respect to the four points is selected.To find the minor radius,the points are collected in a plane that is rotated around the axis.Then a circle is computed using three points in this plane.The major radius is given as the distance of the circle center to the plexityThe complexity of RANSAC is dominated by two major fac-tors:The number of minimal sets that are drawn and the cost of evaluating the score for every candidate shape.As we de-sire to extract the shape that achieves the highest possible score,the number of candidates that have to be considered is governed by the probability that the best possible shape is indeed detected,i.e.that a minimal set is drawn that defines this shape.4.2.1.ProbabilitiesConsider a point cloud P of size N and a shape ψtherein consisting of n points.Let k denote the size of a minimal set required to define a shape candidate.If we assume that any k points of the shape will lead to an appropriate candidate shape then the probability of detecting ψin a single pass is:P (n )= n k N k ≈ n N k(1)The probability of a successful detection P (n ,s )after s can-didates have been drawn equals the complementary of s con-secutive failures:P (n ,s )=1−(1−P (n ))s(2)Solving for s tells us the number of candidates T required to detect shapes of size n with a probability P (n ,T )≥p t :T ≥ln (1−p t )ln (1−P (n ))(3)Figure 2:A small cylinder that has been detected by ourmethod.The shape consists of 1066points and was detected among 341,587points.That corresponds to a relative size of 1/3000.For small P (n )the logarithm in the denominator can be approximated by its Taylor series ln (1−P (n ))=−P (n )+O (P (n )2)so that:T ≈−ln (1−p t )P (n )(4)Given the cost C of evaluating the cost function,the asymp-totic complexity of the RANSAC approach is O (TC )=O (1P (n )C ).4.3.Sampling strategyAs can be seen from the last formula,the runtime complexity is directly linked to the success rate of finding good sample sets.Therefore we will now discuss in detail how sampling is performed.4.3.1.Localized samplingSince shapes are local phenomena,the a priori probability that two points belong to the same shape is higher the smaller the distance between the points.In our sampling strategy we want to exploit this fact to increase the probability of draw-ing minimal sets that belong to the same shape.[MTN ∗02]have shown that non-uniform sampling based on locality leads to a significantly increased probability of selecting a set of inliers.From a ball of given radius around an ini-tially unrestrainedly drawn sample the remaining samples are picked to obtain a complete minimal set.This requires to fix a radius in advance,which they derive from a known (or assumed)outlier density and distribution.In our setup however,outlier density and distribution vary strongly for different models and even within in a single model,which renders a fixed radius inadequate.Also,in our case,using minimal sets with small diameter introduces unnecessary stability issues in the shape estimation procedure for shapes that could have been estimated from samples spread farther apart.Therefore we propose a novel sampling strategy that is able to adapt the diameter of the minimal sets to both,outlier density and shape size.cThe Eurographics Association and Blackwell Publishing 2007.We use an octree to establish spatial proximity between sam-ples very efficiently.When choosing points for a new candi-date,we draw thefirst sample p1without restrictions among all points.Then a cell C is randomly chosen from any level of the octree such that p1is contained in C.The k−1other samples are then drawn only from within cell C.The effect of this sampling strategy can be expressed in a new probability P local(n)forfinding a shapeψof size n:P local(n)=P(p1∈ψ)P(p2...p k∈ψ|p2...p k∈C)(5) Thefirst factor evaluates to n/N.The second factor obvi-ously depends on the choice of C.C is well chosen if it con-tains mostly points belonging toψ.The existence of such a cell is backed by the observation that for most points on a shape,except on edges and corners,there exists a neighbor-hood such that all of the points therein belong to that shape. Although in general it is not guaranteed that this neighbor-hood is captured in the cells of the octree,in the case of real-life data,shapes have to be sampled with an adequate density for reliable representation and,as a consequence,for all but very few points such a neighborhood will be at least as large as the smallest cells of the octree.For the sake of analysis,we assume that there exists a C for every p i∈ψsuch thatψwill be supported by half of the points in C, which accounts for up to50%local noise and outliers.We conservatively estimate the probability offinding a good C by1d where d is the depth of the octree(in practice a path of cells starting at the highest good cell to a good leaf will be good as well).The conditional probability for p2,p3∈ψinthe case of a good cell is then described by (|C|/2k−1)(|C|k−1)≈(12)k−1.And substituting yields:P local(n)=nNd2k−1(6)As large shapes can be estimated from large cells(and with high probability this will happen),the stability of the shape estimation is not affected by the sampling strategy.The impact of this sampling strategy is best illustrated with an example.The cylinder depicted in Figure2consists of 1066points.At the time that it belongs to one of the largest shapes in the point-cloud,341,547points of the original2 million still remain.Thus,it then comprises only three thou-sandth of the point-cloud.If an ordinary uniform sampling strategy were to be applied,151,522,829candidates would have to be drawn to achieve a detection probability of99%. With our strategy only64,929candidates have to be gen-erated for the same probability.That is an improvement by three orders of magnitude,i.e.in this case that is the differ-ence between hours and seconds.4.3.1.1.Level weighting Choosing C from a proper level is an important aspect of our sampling scheme.Therefore we can further improve the sampling efficiency by choosing C from a level according to a non-uniform distribution that re-flects the likelihood of the respective level to contain a good cell.To this end,the probability P l of choosing C from level l isfirst initialized with1d.Then for every level l,we keep track of the sumσl of the scores achieved by the candidates generated from a cell on level l.After a given number of candidates has been tested,a new distribution for the levels is computed.The new probabilityˆP l of the level l is given asˆPl=xσlwP l+(1−x)1d,(7)where w=∑d i=1σPi.We set x=.9to ensure that at all times at least10%of the samples are spread uniformly over the levels to be able to detect when new levels start to become of greater importance as more and more points are removed from P.4.3.2.Number of candidatesIn section4.2we gave a formula for the number of candi-dates necessary to detect a shape of size n with a given prob-ability.However,in our case,the size n of the largest shape is not known in advance.Moreover,if the largest candidate has been generated early in the process we should be able to de-tect this lucky case and extract the shape well before achiev-ing a precomputed number of candidates while on the other hand we should use additional candidates if it is still unsure that indeed the best candidate has been detected.Therefore, instead offixing the number of candidates,we repeatedly an-alyze small numbers t of additional candidates and consider the best oneψm generated so far each time.As we want to achieve a low probability that a shape is extracted which is not the real maximum,we observe the probability P(|ψm|,s) with which we would have found another shape of the same size asψm.Once this probability is higher than a threshold p t(we use99%)we conclude that there is a low chance that we have overlooked a better candidate and extractψm.The algorithm terminates as soon as P(τ,s)>p t.4.4.ScoreThe score functionσP is responsible for measuring the qual-ity of a given shape candidate.We use the following aspects in our scoring function:•To measure the support of a candidate,we use the number of points that fall within anε-band around the shape.•To ensure that the points inside the band roughly follow the curvature pattern of the given primitive,we only count those points inside the band whose normals do not deviate from the normal of the shape more than a given angleα.•Additionally we incorporate a connectivity measure: Among the points that fulfill the previous two conditions, only those are considered that constitute the largest con-nected component on the shape.c The Eurographics Association and Blackwell Publishing2007.。
Today Remote eVoting
8
Introduction to Remote E-Voting (2/5)
•
secrecy
Postal vote: Privacy of Correspondence Remote E-Voting: PC hacked (trojan horse, key logger)? Security state of the PC: Responsibility of the user? Long term security of cryptographic mechanisms?
cf.: Robert Krimmer, Melanie Volkamer. Wählen auf Distanz: Ein Vergleich zwischen elektronischen und nicht elektronischen Verfahren. In Jahrbuch Effizienz von e-Lösungen in Staat und Gesellschaft - Tagungsband IRIS 2005, Boorberg Verlag, 2005.
11
Introduction to Remote E-Voting (5/5)
•
universality
Postal vote: Ballot paper in time? Letter stashed away? Vote counted? Guarantee? Expatriot? Remote E-Voting: Equipment to vote => Digital Divide? Denial of Service?
cf.: Robert Krimmer, Melanie Volkamer. Wählen auf Distanz: Ein Vergleich zwischen elektronischen und nicht elektronischen Verfahren. In Jahrbuch Effizienz von e-Lösungen in Staat und Gesellschaft - Tagungsband IRIS 2005, Boorberg Verlag, 2005.
KWP2000协议
COMMITTEE DRAFT ISO/CD 14230-4Date Reference numberISO/TC 22/SC 3 NSupersedes documentTHIS DOCUMENT IS STILL UNDER STUDY AND SUBJECT TO CHANGE. ITSHOULD NOT BE USED FOR REFERENCE PURPOSES.Circulated to P- and O-members, and to technical committees and ISO/TC 22/SC 3organizations in liaison for :Title discussion atVéhicules routiers - Equipement électrique etélectronique comments byvoting (P-members only: ballot form attached)for approval for registration as a DIS (see 2.5.6 of part 1 of theISO/IEC Directives), bySecretariatFAKRAP-members of the technical committee or sub-committeeconcerned have an obligation to vote.Road vehicles - Diagnostic systems - Keyword Protocol 2000 - Part 4 : Requirements for emission related systemsPartie 4 :Reference language version : English French RussianIntroductory noteFORM 7 (ISO)F ICHIER :C:\XDOM001\ISO\14230-4E.DOC VERSION DU :07/01/9717:01N OMBRE DE PAGE(S):8 PAGE(S)Page 2ISO/CD 14230-4:1997(E)ContentsForeword (3)0 Introduction (4)1 Scope (4)references (4)2 Normativelayer (5)3 Physical4 Data link layer (5)structure (5)4.1 Message4.2 Timing (5)service (6)4.3 StartCommunication4.4 Stop communication service (6)service (6)4.5 AccessTimingParameterservices (6)5 Diagnostic5.1 Emission related services (6)service (6)5.2 TesterPresent5.3 Other diagnostic services (6)responses (7)5.4 NegativeAnnex A (informative) - Bibliography (8)Page 3ISO/CD 14230-4:1997(E)ForewordISO (the International Organization for Standardization) is a worldwide federation of national standards bodies (ISO member bodies). The work of preparing International Standards is normally carried out through ISO technical committees. Each member body interested in a subject for which a technical committee has been established has the right to be represented on that committee. International organisations, governmental and non-governmental, in liaison with ISO, also take part in the work. ISO collaborates closely with the International Electrotechnical Commission (IEC) on all matters of electrotechnical standardization.Draft International Standards adopted by the technical committees are circulated to the member bodies for voting. Publication as an International Standard requires approval by at least 75 % of the member bodies casting a vote.- Part 1: Physical layer- Part 2: Data link layer- Part 3: Application layer- Part 4: KWP 2000 requirements for emission related systems.Annex A of this International Standard is for information only.Page 4ISO/CD 14230-4:1997(E)0 IntroductionThis International Standard contains references to SAE publications, which are regularly amended/updated without any visible change (neither in the numbering, nor any additive letter, etc.). To ensure precisely to which particular edition this International Standard refers to, Annex A gives the precise dates of the SAE publications.1 ScopeThis International Standard specifies the requirements for the Keyword protocol 2000(KWP 2000) data link and connected vehicle and scan tool when used to comply with On-Board Diagnostic (OBD) requirements for emission related test data.This International Standard only specifies restrictions to Keyword protocol 2000 for OBD purpose. Complete specification can be found in ISO/DIS 14230 Parts 1 to 3 (KWP 2000) and inISO/DIS 14229.Only those sections of ISO/DIS 14230-1, ISO/DIS 14230-2 and ISO/DIS 14230-3 explicitely referenced in this standard are applicable for OBD purposes.references2 NormativeThe following standards contain provisions which, through reference in this text, constitute provisions of this International Standard. At the time of the publication, the editions indicated were valid. All standards are subject to revision, and parties to agreements based on the International Standard are encouraged to investigate the possibility of applying the most recent editions of the standards indicated below. Members of IEC and ISO maintain registers of currently valid International Standards.ISO/DIS 14229: 1996, Road vehicles - Diagnostic systems - Diagnostic services specification.ISO/DIS 14230-1: 1996, Road vehicles - Diagnostic systems - Keyword protocol 2000- Part 1: Physical layer.ISO/DIS 14230-2: 1996, Road vehicles - Diagnostic Systems - Keyword protocol 2000- Part 2: Data link layer.ISO/DIS 14230-3: 1996, Road vehicles - Diagnostic systems - Keyword Protocol 2000- Part 3: Application layer.ISO 9141-2: 1994, Road vehicles - Diagnostic systems - Part 2: CARB-requirements for interchange of digital information.ISO/WD 15031-5: Road vehicles - Emission related diagnostic system - Communication between vehicle and external equipment - Part 5: Emission related diagnostic servicesPage 5ISO/CD 14230-4:1997(E) layer3 PhysicalAll sections of ISO/DIS 14230-1 are applicable for OBD purposes, except for the restrictions defined below.There is no restriction for the physical layer. It should be noticed that ISO/DIS 14230-1 andISO 9141-2 physical layers are fully compatible. The only difference between these two standards is that ISO/DIS 14230-1 additionally supports 24 V systems. Testers meeting this standard are not required to support communications with 24 V systems.The baud rate is specified by the protocol and shall not be determined by measurement.4 Data link layerAll sections of ISO/DIS 14230-2 are applicable for OBD purposes, except for the restrictions defined below.4.1 Message structureThe header shall always consist of three bytes.An optional length byte shall not be used. Data length shall be limited to 7 bytes for compatibility with J 1979 - (and/or ISO/WD 15031-5).Bits A1A0 of format byte = 11, with address information and functional addressing shall be used for request messages.Bits A1A0 of format byte = 10, with address information and physical addressing shall be used for response messages.Functional address 33 H shall be used as target address for messages sent to the vehicle.Target address of the responses shall be the source address included in the request.Source address of the responses shall be the physical address of the ECUs.4.2 TimingOnly the normal timing parameter set with default values shall be used for both the vehicle and the scan tool. Timing exceptions as described in ISO/DIS 14230-2 are not allowed.Page 6ISO/CD 14230-4:1997(E)4.3 StartCommunication serviceECU(s) (OBD related) shall only support one of the two following methods of initialisation :- 5 baud initialisation;- fast initialisation.The scan tool shall support both methods:- 5 baud initialisation;- fast initialisation.Keywords received by the scan tool can be 2025, 2027, 2029 and 2031. In any case, the scan tool and the vehicle shall only use the functionality of keyword 2025 (ie. 3 byte header, no additionnal length byte, normal timing).In case 5 baud initialisation is used, then 5 bauds address shall be 33H and subsequent communication shall take place at 10 400 bauds.4.4 Stop communication serviceThis service may be used by the the scan tool to stop the communication but shall be supported by the vehicule.4.5 AccessTimingParameter serviceThe scan tool shall not support this service.services5 DiagnosticAll sections of ISO/DIS 14229 and ISO/DIS 14230-3 are applicable for OBD purposes, except for the restrictions below.5.1 Emission related servicesSpecification and implementation of emission related diagnostic services are specified inSAE J1979.5.2 TesterPresent serviceThe vehicle shall support the testerPresent service to keep the communication active. This service shall be used within P3 by the scan tool to maintain communication with the vehicle in case no test mode request is needed at this moment. No optional parameter shall be used, meaning that there shall always be a response to a request. The scan tool may support the testerPresent service or use another method to keep communication active.5.3 Other diagnostic servicesSupport of the other diagnostic services is not required by this standard.Page 7ISO/CD 14230-4:1997(E)5.4 Negative responsesA module shall always respond to a request either with positive or negative response when no transmission error has been detected. Format and usage of negative responses are defined inISO 14230-3.In case a negative response is used, the response code included shall be one of the following:10 generalReject11 serviceNotSupported12 subFunctionNotSupported-invalidFormat21 busy-RepeatRequest22 conditionsNotCorrect or requestSequenceError.78 requestCorrectlyReceived-ResponsePending.All negative responses response code 78 included shall be sent by the modules within P2. The modules shall terminate with a positive response or a negative response with a code different from 78.The scan tool shall ignore the content of the negative response messages and shall not perform the error handling actions as specified in ISO/DIS 14230-3.Page 8ISO/CD 14230-4:1997(E)Annex A (informative)BibliographyThe attention of the user is drawn so that the relevant version of the SAE publication is the following:SAE J1979 : June 1994, E/E Diagnostic test modes.This list contains no provision that a more updated version is also valid as a reference to be used in this standard.。
图像处理和计算机视觉--基础,经典以及最近发展
图像处理与计算机视觉基础，经典以及最近发展By xdyang（杨晓冬tc@）一、绪论1.为什么要写这篇文章从2002年到现在，接触图像快十年了。
虽然没有做出什么很出色的工作，不过在这个领域摸爬滚打了十年之后，发现自己对图像处理和计算机视觉的感情越来越深厚。
下班之后看看相关的书籍和文献是一件很惬意的事情。
平常的一大业余爱好就是收集一些相关的文章，尤其是经典的文章，到现在我的电脑里面已经有了几十G的文章。
写这个文档的想法源于我前一段时间整理文献时的一个突发奇想，既然有这个多文献，何不整理出其中的经典，抓住重点来阅读，同时也可以共享给大家。
于是当时即兴写了一个《图像处理与计算机视觉中的经典论文》。
现在来看，那个文档写得很一般，所共享的论文也非常之有限。
就算如此，还是得到了一些网友的夸奖，心里感激不尽。
因此，一直想下定决心把这个工作给完善，力求做到尽量全面。
本文是对现有的图像处理和计算机视觉的经典书籍（后面会有推荐）的一个补充。
一般的图像处理书籍都是介绍性的介绍某个方法，在每个领域内都会引用几十上百篇参考文献。
有时候想深入研究这个领域的时候却发现文献太多，不知如何选择。
但实际上在每个领域都有那么三五篇抑或更多是非读不可的经典文献。
这些文献除了提出了很经典的算法，同时他们的Introduction和Related work 也是对所在的领域很好的总结。
读通了这几篇文献也就等于深入了解了这个领域，比单纯的看书收获要多很多。
写本文的目的就是想把自己所了解到的各个领域的经典文章整理出来,不用迷失在参考文献的汪洋大海里。
2.图像处理和计算机视觉的分类按照当前流行的分类方法，可以分为以下三部分：图像处理：对输入的图像做某种变换，输出仍然是图像，基本不涉及或者很少涉及图像内容的分析。
比较典型的有图像变换，图像增强，图像去噪，图像压缩，图像恢复，二值图像处理等等。
基于阈值的图像分割也属于图像处理的范畴。
一般处理的是单幅图像。
关于选秀节目的英语作文
A talent show is a type of television program where contestants with some form of talentsuch as singing,dancing,acting,or playing a musical instrumentcompete against one another for a top prize,often a monetary reward or a contract with a record label or talent agency.Talent shows have been a staple of television programming for many years, and they continue to be popular with audiences around the world.The Appeal of Talent ShowsTalent shows are appealing for a variety of reasons.Firstly,they provide a platform for individuals who may not have had the opportunity to showcase their skills otherwise. This can be particularly important for those who come from disadvantaged backgrounds or who live in areas where opportunities for artistic expression are limited.Secondly,the competitive nature of these shows can be thrilling to watch.Audiences enjoy the suspense of not knowing who will come out on top,and the judges critiques can be both entertaining and informative.The contestants backstories often add an emotional element to the show,making viewers feel more invested in their success.The Format of Talent ShowsTalent shows typically follow a structured format.Contestants audition in front of a panel of judges,and those who impress the judges are selected to move on to the next round. As the competition progresses,the number of contestants is gradually reduced until only a few remain.These finalists then compete in a grand finale,where the winner is chosen.The judging panel usually consists of industry professionals,such as musicians,actors,or dancers,who provide feedback and score the performances.In some shows,the audience also gets a say in the outcome by voting for their favorite contestants.The Impact of Talent ShowsTalent shows have had a significant impact on the entertainment industry.They have launched the careers of many successful artists,such as Susan Boyle from Britains Got Talent and Kelly Clarkson from American Idol.These shows also encourage a culture of performance and competition,which can be both positive and negative.On the positive side,they inspire people to pursue their passions and can lead to increased interest in the arts.However,they can also create unrealistic expectations about fame and success,leading some contestants to believe that winning a talent show is the only path to a successful career.Criticism of Talent ShowsDespite their popularity,talent shows have faced criticism.Some argue that they exploit contestants for entertainment value,with the focus often being more on drama and spectacle than on the actual talent.Others feel that the shows perpetuate a narrow definition of what constitutes talent,often favoring mainstream,commercial styles over more innovative or unconventional performances.ConclusionIn conclusion,talent shows offer a unique blend of entertainment,competition,and opportunity.While they have their drawbacks,they also provide a valuable platform for aspiring artists and can be a source of inspiration for many.As with any form of media, its important to approach talent shows with a critical eye,appreciating the talent on display while also being aware of the broader implications of their format and impact on society.。
财务管理基础工业出版社13版Chap017_PPT
Poison Pills
• A rights offering made to existing shareholders of a company
– Used to avoid a takeover – Makes hostile takeovers very expensive and unattractive – Allows existing shareholders the right to buy additional shares of the stock at a very low price
• Holders of common stock must be given the first option to buy new shares
– Ensures that management cannot subvert the position of present stockholders
1-17
Advantages of ADRs for the U.S. Investor
• Annual reports and financial statements are presented in English according to GAAP • Dividends are paid in dollars and are more easily collected • Considered to be:
1-3
Preferred Stock
• Plays a secondary role in financing the corporate enterprise
– Represents a hybrid security by combining some of the features of debt and common stock – Stockholders do not have an ownership interest in the firm – Stockholders have a priority of claims to dividends superior to that of common stockholders
An anonymous electronic voting protocol for voting over the internet
An Anonymous Electronic Voting Protocol for Voting Over The Internet Indrajit Ray†Indrakshi Ray†Natarajan Narasimhamurthi‡†Department of Computer and Information Science‡Department of Electrical and Computer EngineeringUniversity of Michigan-Dearborn4901Evergreen Road,Dearborn,MI48128AbstractIn this work we propose a secure electronic voting proto-col that is suitable for large scale voting over the Internet. The protocol allows a voter to cast his or her ballot anony-mously,by exchanging untraceable yet authentic messages. The protocol ensures that(i)only eligible voters are able to cast votes,(ii)a voter is able to cast only one vote,(iii)a voter is able to verify that his or her vote is counted in the final tally,(iv)nobody,other than the voter,is able to link a cast vote with a voter,and(v)if a voter decides not to cast a vote,nobody is able to cast a fraudulent vote in place of the voter.1IntroductionSecure electronic voting requires the exchange of un-traceable yet authentic messages.Broadly two different ap-proaches have been proposed:(i)approaches that require complex encryption schemes[1,6,7,10],and(ii)ap-proaches that require an anonymous channel[2,5,8,9,11, 12,13,14]that is used to cast the ballot as an untraceable message.The protocol we propose does not require any complex cryptographic schemes.Our protocol is similar to the ones in[8,9]but does not need an anonymous channel. V oting is similar to a guest ftp session.The session may,at best,be traced back to an IP address but not to a voter.Researchers have identified a set of requirements for a secure electronic voting protocol[8]:1.Accuracy:(i)A cast vote can not be altered.(ii)Aninvalid vote is not counted.(iii)Each voter has the guarantee that his/her ballot is counted.2.Democracy:(i)Only a eligible voter participate.(ii)Each voter can cast only one vote.This work has been partially supported by the National Science Foun-dation under grant EIA99775483.Privacy:A ballot can not be linked back to the voterwho cast it.4.Verifiability:Each voter can verify that his/her vote iscounted.5.We identify an additional property which we call NoUnauthorized Proxy:If a voter decides not to cast his/her ballot,no party can take advantage of this and cast a forged ballot.The protocol we describe next satisfies all of the above properties.2The ProtocolWe assume that the following are available:1.Hard-to-invert permutations:A permutation of afiniteset of numbers whose inverse is hard to compute.2.(Blind)Signature on messages:A verifyable transfor-mation of a message which can only be generated by the signing ing publicly available informa-tion,any one can verify the signature.In a blind signa-ture,the signing entity signs a message without know-ing its contents[3,4].The message that is submitted for blind signature can be freely published without re-vealing the actual message.3.Secure Transit:An encryption scheme that ensures pri-vacy and integrity of messages in transit.2.1Protocol descriptionThe voting protocol employs three,not necessarily trusted,agents for successful operation:1.BD:A ballot distributor who prepares blank ballotsand distributes one to each voter.2.CA:A certifying authority who verifies eligibility,cer-tifies ballots and ensures that a voter gets only one cer-tified ballot.3.VC:A vote compiler who tallies the votes and an-nounces the results.The agents may collude with each other or with a voter to perpetrate fraud.If a fraud is suspected,then the proto-col ensures that the fraud can be proved.When an agent colludes with a voter,it will only affect that voters ballot. Before the voting process,a voter registers with some voter registration authority.This authority prepares a list of regis-tered voters and issues a certificate for each registered voter that contains the voter’s identity and public key.For this discussion we assume that for any party,X,X e represents the party’s encryption key and X d,the decryption (signing)key.The voting proceeds as follows:1.Blank ballot distribution:BD V:[y,[h(y),BD d],V e],[h(voter certificate),BD d].When a voter elec-tronically authenticates himself,BD provides a signed blank ballot and a signed digest of the voter certificate.The blank ballot is a message of twofields(i)the bal-lot serial numberfield,y and(ii)BD signed digest of the ballot serial number,[h(y),BD d].2.Generating a voter mark:The V oter performs a one-way permutation of the serial number to generate a unique voter mark m.3.Voter certification:(a)V CA:[m[r,CA e],[h(m[r,CA e]),V d],V,CA e],[V,voter-certificate,[h(voter-certificate),BD d],CA e],m is the voter markgenerated by the voter.(b)CA V:[[m[r,CA e],CA d],V e]The voter gets a blind signature on the voter mark.The blinded voter mark has the voter’s signature on it, which authenticates the voter at the certifying author-ity.4.Vote casting:(a)V Public FTP site:[vote,[m,CA d],h(vote,[m,CA d]),VC e](b)Public FTP site VC:[vote,[m,CA d],h(vote,[m,CA d]),VC e](c)VC Public FTP site:[h(vote,[m,CA d]),VC d](d)Public FTP site V:[h(vote,[m,CA d]),VC d]When the voter receives the signed salted voter mark, he removes all identifying marks,creates afilled ballot (vote–a message offixed length and pre-determinedformat)and transmits the CA-signed voter mark and vote to VC using a protocol similar to anonymous ftp from a voting kiosk.The CA-signed voter mark to-gether with the correspondingfilled ballot is hence forth called the marked-ballot.5.Vote counting:Once the voting period is over,VCpublishes in a public place all the cast ballots and an-nounces the results.In addition,CA publishes in a pub-lic place all the salted voter marks that were sent to it for signature,and BD publishes the number of blank ballots distributed and their serial numbers.2.2PropertiesFirst suppose that no fraud has been perpetrated.Since all the ballots are published,and each ballot has the voter mark m,any voter can verify that his ballot has been counted.The voter mark being a one-way permutation of the serial number,it is not possible to trace the voter back via the voter mark.In addition,presence of CA’s signature on the voter mark ensures that each voter gets to cast only one vote.Also,uniqueness of m ensures that a voter can cast only once.Finally,for every ballot that is cast,CA should have a salted voter mark signed by a registered voter.This prevents unauthorized proxies.Based on these properties, it is possible to show that the protocol satisfies the require-ment set forth earlier.Next consider the case when a fraud has been perpe-trated.In the absence of fraud,the following will be true: 1.Every ballot that is published by VC will have CA’ssignature.2.For every marked-ballot that is published by VC,thereshould be a corresponding unsigned salted voter mark that was submitted to CA for signature.Each of those must have the signature of an eligible voter and the signature of BD.Also,voter marks submitted for blind signature are publicly available.Since marked-ballots that VC publishes must contain CA’s signature,CA must be involved in any fraud.CA and VC by colluding can produce a marked-ballot with the neces-sary signatures.However,there will not be a corresponding salted voter mark that has been submitted for CA’s signa-ture.Similarly BD and CA can collude to cast a fraudulent ballot with the necessary signatures.However,in both these two cases,when thefinal ballot is published,the number of ballots that are counted would exceed the number of voter marks that were submitted to CA for signature by valid vot-ers.Thus,in these cases,where only two of the three en-tities collude,fraud is easily detected.This observation is equally valid even if all the three agents collude.Thus,in order to perpetrate fraud,CA must generate spurious signedmarked-ballots,and VC must substitute valid ballots with spurious ones.Detection of such a fraud requires the ac-tive participation of the voter.A voter,who does not see his ballot in thefinal tally will detect the fraud and prove it by submitting,anonymously,a copy of his ballot signed by CA.A corresponding ballot will not be found among the published ballots.3ConclusionIn this work we present a secure electronic voting proto-col that is suitable for large scale voting over the Internet. The protocol satisfies the core properties of secure voting systems–namely accuracy,democracy,privacy and verifi-ability.Further the protocol ensures that if an eligible voter decides not to cast a vote(not the same as a voter choosing to abstain),nobody is able to cast a fraudulent vote in place of the voter–a property we call no unauthorized proxy.We are aware of three shortcomings:Since the voter can iden-tify his ballot,we can not prevent vote buying.Secondly, if several voters,after obtaining CA’s signature,decide not to cast their ballot,then the three agents can cast fraudu-lent ballots.This fraud cannot be detected if the number of fraudulent ballots is less than the number of signed ballots that were not cast.If such a fraud is detected,then prov-ing it will require the cooperation of the voters who did not cast their signed ballots.Finally,we allow a cast ballot to be traced back to an IP address(not to a voter).By using public voting kiosk we can avoid the IP address to be linked to a voter.We are currently looking into methods to address these issues.References[1]J.Benaloh and M.Young.Distributing the Power ofa Government to Enhance the Privacy of V oters.InProc.of the5th ACM Symposium on Principles of Dis-tributed Computing,pages52–62,August1986. [2]C.Boyd.A New Multiple Key Cipher and an Im-proved V oting Scheme.In Advances in Cryptology –EUROCRYPT’89,volume434of Lecture Notes in Computer Science,pages617–625.Springer-Verlag, Berlin,1989.[3]D.Chaum.Blind Signatures for Untraceable Pay-ments.In Advances in Cryptology–CRYPTO’82, pages199–203.Springer-Verlag,Berlin,1983. [4]D.Chaum.Security Without Identification:Transac-tion Systems to Make Big Brother -munications of the ACM,28(10):1030–1044,October 1985.[5]D.Chaum.Elections with Unconditionally Secret Bal-lots and Disruption Equivalent to Breaking RSA.In Advances in Cryptology–EUROCRYPT’88,volume 330of Lecture Notes in Computer Science,pages177–182.Springer-Verlag,Berlin,1988.[6]R.Cramer,M.Franklin, B.Schoenmakers,andM.Young.Multi-Authority Secret-Ballot Elections with Linear Work.In Advances in Cryptology–EU-ROCRYPT’96,volume1070of Lecture Notes in Com-puter Science,pages72–83.Springer-Verlag,Berlin, 1996.[7]R.Cramer,R.Gennaro,and B.Schoenmakers.A Se-cure and Optimally Efficient Multi-Authority Election Scheme.In Advances in Cryptology–EUROCRYPT ’97,volume1233of Lecture Notes in Computer Sci-ence,pages103–118.Springer-Verlag,Berlin,1997.[8]L.F.Cranor and R.K.Cytron.Design and Im-plementation of a Practical Security-Conscious Elec-tronic Polling System.Technical Report WUCS-96-02,Dept.Of Computer Science,Washington Univer-sity,January1996.[9]A.Fujioka,T.Okamoto,and K.Ohta.A PracticalSecret V oting Scheme for Large Scale Elections.In Advances in Cryptology–AUSCRYPT’92,volume 718of Lecture Notes in Computer Science,pages244–251.Springer-Verlag,Berlin,1993.[10]K.R.Iversen.A Cryptographic Scheme for Com-puterized General Elections.In Advances in Cryptol-ogy–CRYPTO’91,volume576of Lecture Notes in Computer Science,pages405–419.Springer-Verlag, Berlin,1992.[11]W.S.Juang and C.L.Lei.A Secure and PracticalElectronic V oting Scheme for Real World Environ-ments.IEICE Transaction on Fundamentals of Elec-tronics,Communications and Computer Science,E80-A(1):64–71,January1997.[12]T.Okamoto.An Electronic V oting Scheme.In Proc.of the IFIP Workshop on Advanced IT Tools,pages 21–30.Chapman&Hall,1996.[13]T.Okamoto.Receipt-Free Electronic V oting Schemesfor Large Scale Elections.In Proc.of Security Proto-col Workshop–Paris,pages25–35,1997.[14]C.Park,K.Itoh,and K.Kurosawa.Efficient Anony-mous Channel and All/Nothing Election Scheme.In Advances in Cryptology–EUROCRYPT’93,volume 765of Lecture Notes in Computer Science,pages248–259.Springer-Verlag,Berlin,1993.。
HOG
Histograms of Oriented Gradients (HOG)理解和源码2010年6月1日丕子发表评论阅读评论2152 V iewsHOG descriptors 是应用在计算机视觉和图像处理领域，用于目标检测的特征描述器。
这项技术是用来计算局部图像梯度的方向信息的统计值。
这种方法跟边缘方向直方图（edge orientation histograms）、尺度不变特征变换（scale-invariant feature transform descriptors）以及形状上下文方法（ shape contexts）有很多相似之处，但与它们的不同点是：HOG描述器是在一个网格密集的大小统一的细胞单元（dense grid of uniformly spaced cells）上计算，而且为了提高性能，还采用了重叠的局部对比度归一化（overlapping local contrast normalization）技术。
这篇文章的作者Navneet Dalal和Bill Triggs是法国国家计算机技术和控制研究所French National Institute for Research in Computer Science and Control (INRIA)的研究员。
他们在这篇文章中首次提出了HOG方法。
这篇文章被发表在2005年的CVPR上。
他们主要是将这种方法应用在静态图像中的行人检测上，但在后来，他们也将其应用在电影和视频中的行人检测，以及静态图像中的车辆和常见动物的检测。
HOG描述器最重要的思想是：在一副图像中，局部目标的表象和形状（appearance and shape）能够被梯度或边缘的方向密度分布很好地描述。
具体的实现方法是：首先将图像分成小的连通区域，我们把它叫细胞单元。
然后采集细胞单元中各像素点的梯度的或边缘的方向直方图。
最后把这些直方图组合起来就可以构成特征描述器。
Hough变换
经典的Hough圆检测
缺点：由于参数的累加器是三维数组，所以上述方法的算法复杂度太高， 资源需求大，处理时间长。在大噪声和具有复杂图像背景的情况下，大 量的无用的点也会参与投票，使算法性能大大降低，甚至影响到检测结 果。
一种快速Hough变换检测圆的方法
原理：
未知圆上的三个点可以确定该圆的方程，即可以得到a,b,r。那么在 图像中的三个点可以确定一组参数，即确定一个圆。本方法应用此原理， 在图像空间中随机取三个点，确定一组参数，然后再取三个点，再确定 一组参数，后者与前者相比较，若相同，则此参数的累加器加一，若不 同则将其作为一个新的参数源，放入参数表中，以此类推，直到有一个 参数组的累加器达到我们设定的阈值，或者达到我们设定的循环次数上 限，检测停止。
• Hough变换算法主要应用于二值图像(即边缘图像)，因此在对灰度图像进行 Hough变换前需要对其进行预处理(包括图像的滤波与边缘检测)。Hough变换 是一种使用表决原理的参数估计技术。其原理是利用图像空间和Hough参数 空间的点－线对偶性，把图像空间中的检测问题转换到参数空间。通过在参 数空间里进行简单的累加统计，然后在Hough参数空间寻找累加器峰值的方 法检测直线。Hough变换的实质是将图像空间内具有一定关系的像元进行聚 类，寻找能把这些像元用某一解析形式联系起来的参数空间累积对应点。
的最大值应为 x2 y2 ，所以 [0, (x2 y2)]
通常 [0, ]（每一度为一个点，分成180段） 2.建立一个累加数组（参数变量为2个，数组为二维数组）
检测步骤
3.对图像空间中的点进行hough变换，即算出该点在参数空间上的对应曲线， 并在相应的累加器加1；（图片大小为614*768）
经典的Hough圆检测
排序投票规则介绍
排序投票规则介绍
排序投票（Ranked Voting）是一种选举过程中的投票规则，它允许选民根据他们的偏好对候选人或选择进行排序。
排序投票允许选民按照优先级对候选人进行评估，而不仅仅是选择一个。
这种方式可以更全面地了解选民的意愿，使选民能够同时表达对多个候选人或选择的支持程度。
排序投票的工作原理通常是选民按照他们最喜欢的候选人给出第一选项，其次是第二选项，依此类推，直到列出他们最不喜欢的选项。
然后，根据选民的排序，计算得出每个候选人或选择的得票数，并最终确定选举结果。
常见的排序投票规则包括但不限于：
1. 第一过半制（Majority Judgment）：选民对每个候选人或选择进行打分，然后根据平均得分确定获胜者。
较高的平均分数代表更受欢迎的候选人或选择。
2. 亲合度投票（Score Voting）：选民为每个候选人或选择赋予一个评分，通常是从0到某个最高分数（如10或100）。
最高总分数的候选人或选择将获胜。
3. 鲍德里治瓦报效算法（Borda Count Method）：选民为每个候选人或选择赋予一个排名，圆排名的候选人或选择获得更高的得分。
然后将所有候选人或选择的得分相加，得分最高的候选人或选择获胜。
排序投票在一些国家和地区的选举中被广泛使用，被认为比传统的“一人一票”方式更能准确反映选民的意愿。
它可以减轻少数候选人分散选民支持的问题，并鼓励选民进行更全面的评估。
然而，排序投票也存在一些争议，例如复杂性和可能出现的不稳定性。
hough变换算法
hough变换算法1、算法思想边缘检测⽐如canny算⼦可以识别出图像的边缘，但是实际中由于噪声和光照不均匀等因素，很多情况下获得的边缘点是不连续的，必须通过边缘连接将他们转换为有意义的边缘。
Hough变化是⼀个重要的检测间断点边界形状的⽅法，它通过将图像坐标空间变化到参数空间来实现直线和曲线的拟合。
霍夫变换于1962年由Paul Hough ⾸次提出，后于1972年由Richard Duda和Peter Hart推⼴使⽤，经典霍夫变换⽤来检测图像中的直线，后来霍夫变换扩展到任意形状物体的识别，多为圆和椭圆。
Hough变换是图像处理中从图像中识别⼏何形状的基本⽅法之⼀。
Hough直线检测的基本原理在于利⽤点与线的对偶性，在我们的直线检测任务中，即图像空间中的直线与参数空间中的点是⼀⼀对应的，参数空间中的直线与图像空间中的点也是⼀⼀对应的。
这意味着我们可以得出两个⾮常有⽤的结论：1）图像空间中的每条直线在参数空间中都对应着单独⼀个点来表⽰；2）图像空间中的直线上任何⼀部分线段在参数空间对应的是同⼀个点。
因此Hough直线检测算法就是把在图像空间中的直线检测问题转换到参数空间中对点的检测问题，通过在参数空间⾥寻找峰值来完成直线检测任务，也即把检测整体特性转化为检测局部特性。
2、算法原理1）图像空间和参数空间霍夫变换的数学理解是“换位思考”，⽐如⼀条直线y=a*x+b有两个参数，在给定坐标系下，这条直线就可以⽤a和b进⾏完整的表述。
如果我们把x和y看作参数，把a和b看作变量的话，那么图像空间下的坐标点（x1，y1）对应着参数空间⾥的⼀条直线q=-x1*k+y1, 图像空间直线上的点（x1，y1）就是参数空间的斜率和截距，其中k，q为参数空间的⾃变量。
2）参数空间转换过程下⾯⽤不同空间下的点和线的变换过程⽰例说明。
⼀条直线可由两个点A=(X1,Y1)和B=(X2,Y2)确定(笛卡尔坐标)。
另⼀⽅⾯，y=kx+q也可以写成关于（k,q）的函数表达式（霍夫空间）：对应的变换可以通过图形直观表⽰：变换后的空间成为霍夫空间。
公司理财精要题库
Chapter 07Equity Markets and Stock ValuationMultiple Choice Questions?1.?What is the name given to the model that computes the present value ofa stock by dividing next year's annual dividend amount by the difference between the discount rate and the rate of change in the annual dividend amount?A.?Stock pricing modelB.?Equity pricing modelC.?Capital gain modelD.?Dividend growth modelE.?Present value model2.?The dividend yield is defined as:?A.?the current annual cash dividend divided by the current market price per share.B.?the current annual cash dividend divided by the current book value per share.C.?next year's expected cash dividend divided by the current market price per share.D.?next year's expected cash dividend divided by the current book value per share.E.?next year's expected cash dividend divided by next year's expected market price per share.3.?The capital gains yield equals which one of the following?A.?Total yieldB.?Current discount rateC.?Market rate of returnD.?Dividend yieldE.?Dividend growth rate4.?Which one of the following types of securities has no priority in a bankruptcy proceeding?A.?Convertible bondB.?Senior debtC.?Common stockD.?Preferred stockE.?Straight bond5.?Mary owns 100 shares of stock. Each share entitles her to one vote per open seat on the board of directors. Assume there are 3 open seats in the current election and Mary casts all 300 of her votes for a single candidate. What is the term used to describe this type of voting?A.?ProxyB.?AggregateC.?CumulativeD.?StraightE.?Condensed6.?There are two open seats on the board of directors. If two separate votes occur to elect the new directors, the firm is using a type of voting that is best described as _____ voting.?A.?simultaneousB.?straightC.?proxyD.?cumulativeE.?sequential7.?Kate could not attend the last shareholders meeting and thus she granted the authority to vote on her behalf to the managers of the firm. Which one of the following terms is used to describe the method by which Kate's shares were voted?A.?StraightB.?CumulativeC.?Consent-formD.?ProxyE.?In absentia8.?Dividends are best defined as:?A.?cash payments to shareholders.B.?cash payments to either bondholders or shareholders.C.?cash or stock payments to shareholders.D.?cash or stock payments to either bondholders or shareholders.E.?distributions of stock to current shareholders.9.?Which one of the following generally pays a fixed dividend, receives first priority in dividend payment, and maintains the right to a dividend payment, even if that payment is deferred?A.?Cumulative commonB.?Noncumulative commonC.?Noncumulative preferredD.?Cumulative preferredE.?Senior common10.?Newly issued securities are sold to investors in which one of the following markets?A.?ProxyB.?Stated valueC.?InsideD.?SecondaryE.?Primary11.?What is the market called that allows shareholders to resell their shares to other investors?A.?PrimaryB.?ProxyC.?SecondaryD.?InsideE.?Initial12.?An agent who buys and sells securities from inventory is called a:?A.?floor trader.B.?dealer.C.?commission broker.D.?broker.E.?floor broker.13.?A broker is an agent who:?A.?trades on the floor of an exchange for himself or herself.B.?buys and sells from inventory.C.?offers new securities for sale to dealers only.D.?who is ready to buy or sell at any time.E.?brings buyers and sellers together.14.?Any person who owns a license to trade on the NYSE is called a:?A.?dealer.B.?floor trader.C.?specialist.D.?member.E.?proxy.15.?A person who executes customer orders to buy and sell securities on the floor of the NYSE is called a:?A.?floor trader.B.?specialist.C.?runner.D.?commission broker.E.?market maker.16.?A specialist is a(n):?A.?employee who executes orders to buy and sell for clients of his or her brokerage firm.B.?individual who trades on the floor of an exchange for his or her personal account.C.?NYSE member who functions as a dealer for a limited number of securities.D.?broker who buys and sells securities from a market maker.E.?trader who only deals with primary offerings.17.?An individual who executes buy and sell orders on the floor of an exchange for a fee is called a:?A.?floor broker.B.?specialist.C.?floor trader.D.?proxy.E.?flow specialist.18.?The electronic system that transmits buy and sell orders directly toa specialist on the floor of the NYSE is called:?A.?NASDAQ.B.?SuperDOT.C.?TICKER.D.?ECN.E.?ORDFLOW.19.?The owner of a trading license who trades on the floor of the NYSE for his or her personal account is called a(n):?A.?specialist.B.?independent broker.C.?floor trader.D.?stand-alone agent.E.?dealer.20.?The stream of customer instructions to buy and sell securities is called the:?A.?order flow.B.?market maker.C.?execution stream.D.?operations flow.E.?buyer's stream.21.?The specific location on the floor of an exchange where a particular security is traded is called a:?A.?box office.B.?figure 6.C.?specialist's post.D.?trading booth.E.?seat.22.?Inside quotes are defined as the:?A.?bid and asked prices presented by NYSE specialists.B.?last bid and asked price offered prior to the market close.C.?lowest asked and highest bid offers.D.?daily opening bid and asked quotes.E.?last traded bid and asked prices.23.?Which one of the following is a web site that enables Lester to sell his shares of ABC stock directly to Marti?A.?SuperDOTB.?POSTC.?ECND.?SEATE.?eNET24.?Which one of the following will increase the current value of a stock?A.?Decrease in the dividend growth rateB.?Increase in the required returnC.?Increase in the market rate of returnD.?Decrease in the expected dividend for next yearE.?Increase in the capital gains yield25.?The price of a stock at year 4 can be expressed as:?A.?D0 / (R + G4).B.?D? (1 + R)5.C.?D1? (1 + R)5.D.?D4/(R-g).E.?D5/(R-g).26.?Delfino's expects to pay an annual dividend of $ per share next year. What is the anticipated dividend for year 5 if the firm increases its dividend by 2 percent annually?A.?$ ? 1B.?$ ? 2C.?$ ? 3D.?$ ? 4E.?$ ? 527.?The required return on a stock is equal to which one of the following if the dividend on the stock decreases by 1 percent per year?A.?(P0/D1)-gB.?(D1/P)/gC.?Dividend yield + capital gains yieldD.?Dividend yield - capital gains yieldE.?Dividend yield ? capital gains yield28.?Donuts Delite just paid an annual dividend of $ a share. The firm expects to increase this dividend by 8 percent per year the following 3 years and then decrease the dividend growth to 2 percent annually thereafter. Which one of the following is the correct computation of the dividend for year 7?A.?($ ? 3) ? 4)B.?($ ? 3) ? 3)C.?($ 34D.?($ 33E.?($ 3229.?Aardvark, Inc. pays a constant annual dividend. At the end of trading on Wednesday, the price of its stock was $28. At the end of trading on the following day, the stock price was $27. As a result of the decline in the stock's price, the dividend yield _____ while the capital gains yield _____.?A.?remained constant; remained constantB.?increased; remained constantC.?increased; increasedD.?decreased; remained constantE.?decreased; decreased30.?Which one of the following must equal zero if a firm pays a constant annual dividend?A.?Dividend yieldB.?Capital gains yieldC.?Total returnD.?Market value per shareE.?Book value per share31.?The dividend growth model can be used to value the stock of firms which pay which type of dividendsI. constant annual dividendII. annual dividend with a constant increasing rate of growthIII. annual dividend with a constant decreasing rate of growthIV. zero dividend?A.?I onlyB.?II onlyC.?II and III onlyD.?I, II, and III onlyE.?I, II, III, and IV32.?Kate owns a stock with a market price of $31 per share. This stock paysa constant annual dividend of $ per share. If the price of the stock suddenly increases to $36 a share, you would expect the:I. dividend yield to increase.II. dividend yield to decrease.III. capital gains yield to increase.IV. capital gains yield to decrease.?A.?I onlyB.?II onlyC.?III onlyD.?I and III onlyE.?II and IV only33.?Computing the present value of a growing perpetuity is most similar to computing the current value of which one of the following?A.?Non-dividend-paying stockB.?Stock with a constant dividendC.?Stock with irregular dividendsD.?Stock with a constant growth dividendE.?Stock with growing dividends for a limited period of time34.?Jensen Shipping has four open seats on its board of directors. How many shares will a shareholder need to control to ensure that his or her candidate is elected to the board given the fact that the firm uses straight voting Assume one share equals one vote.?A.?20 percent of the shares plus one voteB.?25 percent of the shares plus one voteC.?1/3 of the shares plus one voteD.?50 percent of the shares plus one voteE.?51 percent of the shares plus one vote35.?Gleason, Inc. elects its board of directors on a staggered basis using cumulative voting. This implies that:?A.?if there are two open seats, then the candidate with the highest number of votes and the candidate with the lowest number of votes will be selected.B.?the candidates for the open seats are voted for in individual elections.C.?all open positions are filled with one round of voting, assuming there are no tie votes.D.?shareholders can accumulate their votes over multiple years and cast all those votes in one election.E.?the firm's entire board of directors is elected annually in one combined election.36.?Which one of the following statements is correct?A.?From a legal perspective, preferred stock is a form of corporate equity.B.?All classes of stock must have equal voting rights per share.C.?Common shareholders elect the corporate directors while the preferred shareholders vote on mergers and acquisitions.D.?Dividends are tax-free income for individual investors.E.?Shareholders prefer noncumulative dividends over cumulative dividends.37.?Which one of the following statements is correct?A.?Both preferred stock and corporate bonds can be callable.B.?Both preferred stock and corporate bonds have a stated liquidation value of $1,000 each.C.?Interest payments to bondholders as well as dividend payments to preferred shareholders are tax deductible expenses for the issuing firm.D.?Bondholders generally receive a fixed payment while preferred shareholders receive a variable payment.E.?Preferred shareholders receive preferential treatment over bondholders in a liquidation.38.?If shareholders are granted a preemptive right they will:?A.?be given the choice of receiving dividends either in cash or in additional shares of stock.B.?be paid dividends prior to the preferred shareholders during the preemptive period.C.?be entitled to two votes per share of stock.D.?be able to choose the timing and amount of any future dividends.E.?have priority in the purchase of any newly issued shares.39.?On which one of the following dates do dividends become a liability of the issuer for accounting purposes?A.?First day of the fiscal year in which the dividend is expected to be paidB.?Twelve months prior to the expected dividend payment dateC.?On the declaration dateD.?On the date of recordE.?On the date of payment40.?Dividends are which one of the following?A.?Payable at the discretion of a firm's presidentB.?Treated as a tax-deductible expense to the paying firmC.?Paid out of aftertax profitsD.?Paid to holders of record as of the declaration dateE.?Only partially taxable to high-income individual shareholders41.?You have agreed to pay a five percent commission to your best friend if he can locate a buyer for your car. This arrangement is most similar to the compensation arrangement for which one of these individuals who is involved with the stock market?A.?SpecialistB.?Floor traderC.?Market makerD.?Commission brokerE.?Dealer42.?To be a member of the NYSE, you must:?A.?be a primary dealer.B.?buy a seat.C.?own a trading license.D.?be registered as a floor trader.E.?be a specialist.43.?Which one of the following players on the floor of the NYSE is obligated to maintain a fair, orderly market for a limited number of securities?A.?SpecialistB.?Floor traderC.?$2 brokerD.?Commission brokerE.?Floor broker44.?The NYSE:?A.?presently conducts all of its trading through SuperDOT.B.?is a dealer market.C.?is in the business of attracting order flow.D.?is solely a primary market.E.?is based on a multiple market maker system.45.?Which one of the following parties on the NYSE floor post bid and asked prices?A.?Floor tradersB.?SpecialistsC.?Floor brokersD.?Commission brokersE.?Fee brokers46.?Many of the smaller sell orders sent to the floor of the NYSE are:?A.?handled by the floor traders.B.?purchased by the commission brokers.C.?electronically transmitted to the specialists.D.?executed on an ECN.E.?executed in the primary market.47.?If a trade is made "in the crowd", the trade has occurred:?A.?between a broker and a specialist.B.?between two brokers.C.?electronically on NASDAQ.D.?on SuperDOT.E.?on an ECN.48.?The more actively traded large companies that are listed on NASDAQ are traded in which one of the NASDAQ markets?A.?NationalB.?CapitalC.?RegionalD.?Global SelectE.?Global49.?Which one of the following features applies to NASDAQ but not the NYSE?A.?Trading in the crowdB.?Multiple market maker systemC.?SuperDotD.?Broker marketE.?Physical trading floor50.?Companies can list their stock on which one of the following without having to meet listing requirements or filing financial statements with the SEC?A.?NASDAQ Capital MarketB.?Over-the-Counter Bulletin BoardC.?Pink sheetsD.?NASDAQ Global MarketE.?NYSE51.?Keller Metals common stock is selling for $36 a share and has a dividend yield of percent. What is the dividend amount?A.?$B.?$C.?$D.?$E.?$52.?The Glass Ceiling paid an annual dividend of $ per share last year. Management just announced that future dividends will increase by percent annually. What is the amount of the expected dividend in year 5?A.?$B.?$C.?$D.?$E.?$53.?The Pancake House pays a constant annual dividend of $ per share. How much are you willing to pay for one share if you require a 15 percent rate of return?A.?$B.?$C.?$D.?$E.?$54.?Shoreline Foods pays a constant annual dividend of $ a share and currently sells for $ a share. What is the rate of return?A.? percentB.? percentC.? percentD.? percentE.? percent55.?The common stock of Green Garden Flowers is selling for $24 a share. The company pays a constant annual dividend and has a total return of percent. What is the amount of the dividend?A.?$B.?$C.?$D.?$E.?$56.?Healthy Foods just paid its annual dividend of $ a share. The firm recently announced that all future dividends will be increased by percent annually. What is one share of this stock worth to you if you require a14 percent rate of return?A.?$B.?$C.?$D.?$E.?$57.?Plastics, Inc. will pay an annual dividend of $ next year. The company just announced that future dividends will be increasing by percent annually. How much are you willing to pay for one share of this stock if you require a 16 percent return?A.?$B.?$C.?$D.?$E.?$58.?The Printing Company stock is selling for $ a share based on a 14 percent rate of return. What is the amount of the next annual dividend if the dividends are increasing by percent annually?A.?$B.?$C.?$D.?$E.?$59.?The common stock of Mid-Towne Movers is selling for $33 a share and has a 9 percent rate of return. The growth rate of the dividends is 1 percent annually. What is the amount of the next annual dividend?A.?$B.?$C.?$D.?$E.?$60.?Delphin's Marina is expected to pay an annual dividend of $ next year. The stock is selling for $ a share and has a total return of 12 percent. What is the dividend growth rate?A.? percentB.? percentC.? percentD.? percentE.? percent61.?Klaus Toys just paid its annual dividend of $. The required return is 16 percent and the dividend growth rate is 2 percent. What is the expected value of this stock five years from now?A.?$B.?$C.?$D.?$E.?$62.?This morning, you purchased a stock that will pay an annual dividend of $ per share next year. You require a 12 percent rate of return and the annual dividend increases at percent annually. What will your capital gain be on this stock if you sell it three years from now?A.?$B.?$C.?$D.?$E.?$63.?Blackwell Ink is losing significant market share and thus its managers have decided to decrease the firm's annual dividend. The last annual dividend was $ a share but all future dividends will be decreased by 5 percent annually. What is a share of this stock worth today at a required return of 15 percent?A.?$B.?$C.?$D.?$E.?$64.?Lamey Headstones increases its annual dividend by percent annually. The stock sells for $ a share at a required return of 14 percent. What is the amount of the last dividend this company paid?A.?$B.?$C.?$D.?$E.?$65.?The common stock of Tasty Treats is valued at $ a share. The company increases its dividend by 8 percent annually and expects its next dividend to be $ per share. What is the total rate of return on this stock?A.? percentB.? percentC.? percentD.? percentE.? percent66.?River Rock, Inc. just paid an annual dividend of $. The company has increased its dividend by percent a year for the past ten years and expects to continue doing so. What will a share of this stock be worth six years from now if the required return is 16 percent?A.?$B.?$C.?$D.?$E.?$67.?The Cart Wheel plans to pay an annual dividend of $ per share next year, $ per share a year for the following two years, and then cease paying dividends altogether. How much is one share of this stock worth to you today if you require a 17 percent rate of return?A.?$B.?$C.?$D.?$E.?$68.?Atlas Home Supply has paid a constant annual dividend of $ a share for the past 15 years. Yesterday, the firm announced the dividend will increase next year by 10 percent and will stay at the level through year three, after which time the dividends will increase by 2 percent annually. The required return on this stock is 12 percent. What is the current value per share?A.?$B.?$C.?$D.?$E.?$69.?Auto Transmissions is expected to pay annual dividends of $ and $ over the next two years, respectively. After that, the company expects to pay a constant dividend of $ a share. What is the value of this stock at a required return of 15 percent?A.?$B.?$C.?$D.?$E.?$70.?General Importers announced today that its next annual dividend will be $ per share. After that dividend is paid, the company expects to encounter some financial difficulties and is going to suspend dividends for 5 years. Following the suspension period, the company expects to pay a constant annual dividend of $ per share. What is the current value of this stock if the required return is 18 percent?A.?$B.?$C.?$D.?$E.?$71.?Business Services, Inc. is expected to pay its first annual dividend of $ per share three years from now. Starting in year six, the company is expected to start increasing the dividend by 2 percent per year. What is the value of this stock today at a required return of 12 percent?A.?$B.?$C.?$D.?$E.?$72.?New Gadgets is growing at a very fast pace. As a result, the company expects to pay annual dividends of $, , and $ per share over the next three years, respectively. After that, the dividend is projected to increase by 5 percent annually. The last annual dividend the firm paid was $ a share. What is the current value of this stock if the required return is 16 percent?A.?$B.?$C.?$D.?$E.?$73.?The Market Place recently announced that it will pay its first annual dividend two years from today. The first dividend will be $ a share with that amount doubling each year for the following two years. After that, the dividend is expected to increase by 4 percent annually. What is the value of this stock today if the required return is 15 percent?A.?$B.?$C.?$D.?$E.?$74.?A firm expects to increase its annual dividend by 20 percent per year for the next two years and by 15 percent per year for the following two years. After that, the company plans to pay a constant annual dividend of $3 a share. The last dividend paid was $ a share. What is the current value of this stock if the required rate of return is 12 percent?A.?$B.?$C.?$D.?$E.?$75.?The Border Crossing just paid an annual dividend of $ per share and is expected to pay annual dividends of $ and $ per share the next two years, respectively. After that, the firm expects to maintain a constant dividend growth rate of 2 percent per year. What is the value of this stock today if the required return is 14 percent?A.?$B.?$C.?$D.?$E.?$76.?A stock has a market price of $ and pays a $ annual dividend. What is the dividend yield?A.? percentB.? percentC.? percentD.? percentE.? percent77.?The required return on Mountain Meadow stock is 14 percent and the dividend growth rate is percent. The stock is currently selling for $ a share. What is the dividend yield?A.? percentB.? percentC.? percentD.? percentE.? percent78.?For the past six years, the price of Slate Rock stock has been increasing at a rate of percent a year. Currently, the stock is priced at $47 a share and has a required return of 14 percent. What is the dividend yield?A.? percentB.? percentC.? percentD.? percentE.? percent79.?A stock has paid dividends of $, $, $, $, and $ over the past five years, respectively. What is the average capital gains yield?A.? percentB.? percentC.? percentD.? percentE.? percent80.?The Toy Box pays an annual dividend of $ per share and sells for $ a share based on a market rate of return of 15 percent. What is the capital gains yield?A.? percentB.? percentC.? percentD.? percentE.? percent81.?Investors receive a total return of percent on the common stock of Dexter International. The stock is selling for $ a share. What is the dividend growth rate if the company plans to pay an annual dividend of $ a share next year?A.? percentB.? percentC.? percentD.? percentE.? percent82.?Western Beef stock is valued at $ a share. The company pays a constant annual dividend of $ per share. What is the total return on this stock?A.? percentB.? percentC.? percentD.? percentE.? percent83.?Last year, when the stock of Alpha Minerals was selling for $55 a share the dividend yield was percent. Today, the stock is selling for $41 a share. What is the total return on this stock if the company maintains a constant dividend growth rate of percent?A.? percentB.? percentC.? percentD.? percentE.? percent84.?There are four open positions on the board of directors of Double Tree Restaurants. The company has 180,000 shares of stock outstanding. Each share is entitled to one vote. How many shares of stock must you own to guarantee your personal election to the board of directors if the firm uses cumulative voting?A.?36,001 sharesB.?37,501 sharesC.?38,501 sharesD.?40,001 sharesE.?42,001 shares85.?A firm has two open positions on its board of directors. How many shares do you need to own to guarantee your own election to the board if the firm has 12,500 shares of stock outstanding and uses cumulative voting Each share is granted one vote.?A.?3,334 sharesB.?4,168 sharesC.?5,251 sharesD.?5,501 sharesE.?6,251 shares86.?Miller's Hardware has 185,000 shares of stock outstanding with a current market value of $27 a share. You own 38,000 of those shares. Next month, the election will be held to select four new members to the board of directors. The firm uses a cumulative voting system. How much additional money do you need to spend to guarantee that you will be elected to the board assuming that everyone else votes for one of the other candidates?A.?$0B.?$28,512C.?$34,047D.?$222,777E.?$311,02787.?The Chip Dip Co. has 15,500 shares of stock outstanding, grants one vote per share, and uses straight voting. How many shares must you control to guarantee that you will be elected to the firm's board of directors if there are three open seats?A.?5,167 sharesB.?5,134 sharesC.?3,876 sharesD.?7,751 sharesE.?7,134 shares88.?Kathryn owns 18,700 shares of Global Importers. Her shares have a total market value of $787,270. In total, the firm has 65,000 shares outstanding. Each share is entitled to one vote under the straight voting policy of the firm. The next election is in four months at which time two directors are up for election. How much more must Kathryn invest in this firm to guarantee that she is elected to the board?A.?$0B.?$513,361C.?$581,022D.?$647,280E.?$711,01089.?A preferred stock sells for $ a share and has a market return of percent. What is the dividend amount?A.?$B.?$C.?$D.?$E.?$90.?Central Staircase is offering preferred stock which is commonly referred to as 10-10 stock. This stock will pay an annual dividend of $10 a share starting 10 years from now. What is this stock worth to you today if you desire a 16 percent rate of return?A.?$B.?$C.?$D.?$E.?$91.?Graphic Designs has 120,000 shares of cumulative preferred stock outstanding. Preferred shareholders are supposed to be paid $ per quarter per share in dividends. However, the firm has encountered financial problems and has not paid any dividends for the past three quarters. How much will the firm have to pay per share of preferred next quarter if the firm also wishes to pay a common stock dividend?A.?$B.?$C.?$D.?$E.?$92.?Webster Industrial Products has both common and noncumulative preferred stock outstanding. The dividends on these stocks are $ per quarter per share of common and $ per quarter per share of preferred. The company has not paid any dividends for the past two quarters but is expected to pay dividends on both the common and the preferred stock next quarter. What is the minimum amount the firm must pay per share to its preferred stockholders next quarter if it plans to pay a common dividend?A.?$0B.?$C.?$D.?$E.?$93.?Given the following partial stock quote, what was the closing price on the previous trading day if the firm's earnings per share are $？???A.?$B.?$C.?$D.?$E.?$。
数学建模 建模答案.docx
programi :(1) function [accum, varargout] = CircularHough_Grd(img, radrange, varargin) %Detect circular shapes in a grayscale image. Resolve their center %positions and radii.%% [accum, circen, cirrad, dbg_LMmask] = CircularHough_Grd(% img, radrange, grdthres, fltr4LM_R, multirad, fltr4accum)% Circular Hough transform based on the gradient field of an image.% NOTE: Operates on grayscale images, NOT B/W bitmaps.% NO loops in the implementation of Circular Hough transform,% which means faster operation but at the same time larger% memory consumption.%%%%%%%%% INPUT: (img, radrange, grdthres, fltr4LM_R, multirad, fltr4accum) % % img: A 2-D grayscale image (NO B/W bitmap)%% radrange: The possible minimum and maximum radii of the circles% to be searched, in the format of% [minimum radius , maximum_radius] (unit: pixels)% **NOTE**: A smaller range saves computational time and% memory.%% grdthres: (Optional, default is 10, must be non-negative)% The algorithm is based on the gradient field of the% input image. A thresholding on the gradient magnitude% is performed before the voting process of the Circular% Hough transform to remove the Uniform intensity'% (sort-of) image background from the voting process.% In other words, pixels with gradient magnitudes smaller% than 'grdthres' are NOT considered in the computation.% **NOTE**: The default parameter value is chosen for% images with a maximum intensity close to 255. For cases% with dramatically different maximum intensities, e.g.% 10-bit bitmaps in stead of the assumed 8-bit, the default% value can NOT be used. A value of 4% to 10% of the maximum% intensity may work for general cases.%% fltr4LM_R: (Optional, default is 8, minimum is 3)% The radius of the filter used in the search of local% maxima in the accumulation array. To detect circles whose% shapes are less perfect, the radius of the filter needs% to be set larger.%% multirad: (Optional, default is 0.5)% In case of concentric circles, multiple radii may be% detected corresponding to a single center position. This% argument sets the tolerance of picking up the likely% radii values. It ranges from 0.1 to 1, where 0.1% corresponds to the largest tolerance, meaning more radii % values will be detected, and 1 corresponds to the smallest % tolerance, in which case only the "principal" radius will% be picked up.%% fltr4accum: (Optional. A default filter will be used if not given)% Filter used to smooth the accumulation array. Depending % on the image and the parameter settings, the accumulation % array built has different noise level and noise pattern% (e.g. noise frequencies). The filter should be set to an% appropriately size such that ifs able to suppress the% dominant noise frequency.%%%%%%%%% OUTPUT: [accum, circen, cirrad, dbg_LMmask]%% accum: The result accumulation array from the Circular Hough% transform. The accumulation array has the same dimension % as the input image.%% circen: (Optional)% Center positions of the circles detected. Is a N-by-2% matrix with each row contains the (x, y) positions% of a circle. For concentric circles (with the same center% position), say k of them, the same center position will% appear k times in the matrix.%% cirrad: (Optional)% Estimated radii of the circles detected. Is a N-by-1% column vector with a one-to-one correspondance to the% output tircen*. A value 0 for the radius indicates a% failed detection of the circle's radius.%% dbg_LMmask: (Optional, for debugging purpose)% Mask from the search of local maxima in the accumulation % array.%%%%%%%%%% EXAMPLE #0:% rawimg = imread('TestImg_CHT_a2.bmp');% tic;% [accum, circen, cirrad] = CircularHough_Grd(rawimg, [15 60]);% toe;% figure(l); imagesc(accum); axis image;% title(,Accumulation Array from Circular Hough Transfbrm,);% figure(2); imagesc(rawimg); colormap(,gray,); axis image;% hold on;% plot(circen(:,l), circen(:,2), *r+');% for k = 1 : size(circen, 1),% DrawCircle(circen(k, 1), circen(k,2), cirrad(k), 32,，b」)；% end% hold off;% title([*Raw Image with Circles Detected% '(center positions and radii marked)*]);% figure(3); surf(accum, 'EdgeColoF, hone'); axis ij;% title('3-D View of the Accumulation Array*);%% COMMENTS ON EXAMPLE #0:% Kind of an easy case to handle. To detect circles in the image whose% radii range from 15 to 60. Default values for arguments 'grdthres',% 'fltr4LM_R', 'multirad* and ,fltr4accum, are used.%%%%%%%%%% EXAMPLE #1:% rawimg = imread('TestImg_CHT_a3.bmp');% tic;% [accum, circen, cirrad] = CircularHough_Grd(rawimg, [15 60], 10, 20);% toe;% figure(l); imagesc(accum); axis image;% title(,Accumulation Array from Circular Hough Transfbrm,);% figure(2); imagesc(rawimg); colormap('gray'); axis image;% hold on;% plot(circen(:,l), circen(:,2), T+');% for k = 1 : size(circen, 1),% DrawCircle(circen(k, 1), circen(k,2), cirrad(k), 32, 'b-');% end% hold off;% title([*Raw Image with Circles Detected% '(center positions and radii marked)*]);% figure(3); surf(accum, 'EdgeColoF, hone'); axis ij;% title(*3-D View of the Accumulation Array*);%% COMMENTS ON EXAMPLE #1:% The shapes in the raw image are not very good circles. As a result,% the profile of the peaks in the accumulation array are kind of% 'stumpy', which can be seen clearly from the 3-D view of the% accumulation array, (As a comparison, please see the sharp peaks in % the accumulation array in example #0) To extract the peak positions % nicely, a value of 20 (default is 8) is used for argument 'fltr4LM_R', % which is the radius of the filter used in the search of peaks.%%%%%%%%%% EXAMPLE #2:% rawimg = imread(,TestImg_CHT_b3 .bmp1);% fltr4img = [1 1 1 1 1; 1 2 2 2 1; 1 2 4 2 1; 1 2 2 2 1; 1 1 1 1 1];% fltr4img = fltr4img / sum(fltr4img(:));% imgfltrd = filter2( fltr4img , rawimg );% tic;% [accum, circen, cirrad] = CircularHough_Grd(imgfltrd, [15 80], 8, 10); % toe;% figure(l); imagesc(accum); axis image;% title(,Accumulation Array from Circular Hough Transfbrm,);% figure(2); imagesc(rawimg); colormap('gray'); axis image;% hold on;% plot(circen(:,l), circen(:,2), T+');% for k = 1 : size(circen, 1),% DrawCircle(circen(k, 1), circen(k,2), cirrad(k), 32, 'b-');% end% hold off;% title([*Raw Image with Circles Detected% '(center positions and radii marked)*]);%% COMMENTS ON EXAMPLE #2:% The circles in the raw image have small scale irregularities along % the edges, which could lead to an accumulation array that is bad for % local maxima detection. A 5-by-5 filter is used to smooth out the % small scale irregularities. A blurred image is actually good for the % algorithm implemented here which is based on the image's gradient % field.%%%%%%%%%% EXAMPLE #3:% rawimg = imread('TestImg_CHT_c3.bmp');% fltr4img = [1 1 1 1 1; 1 2 2 2 1; 1 2 4 2 1; 1 2 2 2 1; 1 1 1 1 1];% fltr4img = fltr4img / sum(fltr4img(:));% imgfltrd = filter2( fltr4img , rawimg );% tic;% [accum, circen, cirrad]=...% CircularHough_Grd(imgfltrd, [15 105], 8, 10, 0.7);% toe;% figure(l); imagesc(accum); axis image;% figure(2); imagesc(rawimg); colormap(,gray,); axis image;% hold on;% plot(circen(:,l), circen(:,2), *r+');% for k = 1 : size(circen, 1),% DrawCircle(circen(k, 1), circen(k,2), cirrad(k), 32,，b」)；% end% hold off;% title([*Raw Image with Circles Detected% '(center positions and radii marked)*]);%% COMMENTS ON EXAMPLE #3:% Similar to example #2, a filtering before circle detection works for% noisy image too. 'multirad* is set to 0.7 to eliminate the false% detections of the circles* radii.%%%%%%%%%% BUG REPORT:% This is a beta version. Please send your bug reports, comments and% suggestions to pengtao@ . Thanks.%%%%%%%%%%% INTERNAL PARAMETERS:% The INPUT arguments are just part of the parameters that are used by% the circle detection algorithm implemented here. Variables in the code% with a prefix ,prm_, in the name are the parameters that control the% judging criteria and the behavior of the algorithm. Default values for% these parameters can hardly work for all circumstances. Therefore, at% occasions, the values of these INTERNAL PARAMETERS (parameters that% are NOT exposed as input arguments) need to be fine-tuned to make% the circle detection work as expected.% The following example shows how changing an internal parameter could% influence the detection result.% 1. Change the value of the internal parameter 'prm LM LoBndRa* to 0.4% (default is 0.2)% 2. Run the following matlab code:% fltr4accum = [1 2 1; 2 6 2; 1 2 1];% fltr4accum = fltr4accum / sum(fltr4accum(:));% rawimg = imread(,Frame_0_0022jportion.jpg,);% tic;% [accum, circen] = CircularHough_Grd(rawimg,...% [4 14], 10, 4, 0.5, fltr4accum);% toe;% figure(l); imagesc(accum); axis image;% title(*Accumulation Array from Circular Hough Transform*);% figure(2); imagesc(rawimg); colormap(,gray,); axis image;% hold on; plot(circen(:,l), circen(:,2), "); hold off;% title('Raw Image with Circles Detected (center positions marked)*);% 3. See how different values of the parameter 'prm LM LoBndRa* could % influence the result.% Author: Tao Peng% Department of Mechanical Engineering% University of Maryland, College Park, Maryland 20742, USA% pengtao@% Version: Beta Revision: Mar. 07, 2007%%%%%%%% Arguments and parameters %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Validation of argumentsif ndims(img)〜=2 || 〜isnumeric(img),error(*CircularHough_Grd: "img" has to be 2 dimensionaf);endif 〜all(size(img) >= 32),erro^'CircularHough Grd: "img" has to be larger than 32-by-32');endif numel(radrange)〜=2 || -isnumeric(radrange),error([*CircularHough_Grd: "radrange" has to be \ ...'a two-element vector1]);endprm_r_range = sort(max( [0,0;radrange( 1 ),radrange(2)]));% Parameters (default values)prmgrdthres = 10;prmfltrLMR = 8;prmmultirad = 0.5;funccompucen = true;funccompuradii = true;% Validation of argumentsvapgrdthres = 1;if nargin > (1 + vap_grdthres),if isnumeric(varargin{vap grdthres}) && ...varargin(vap grdthres} (1) >= 0,prm_grdthres = varargin {vapgrdthres} (1);elseerror(['CircularHough_Grd: "grdthres" has to be'a non-negative number1]);endendvap_fltr4LM = 2; % filter for the search of local maximaif nargin > (1 + vap_fltr4LM),if isnumeric(varargin{vap_fltr4LM}) && varargin{vap_fltr4LM}(1) >= 3, prmfltrLMR = varargin{vap_fltr4LM} (1);elseerror([,CircularHough_Grd: n fltr4LM_R n has to belarger than or equal to 3']);endendvap_multirad = 3;if nargin > (1 + vap multirad),if isnumeric(varargin{vap_multirad}) && ...varargin{vap multirad}(1) >= 0.1 && ...varargin {vap multirad} (1) <= 1,prmmultirad = varargin {vap_mul tirad} (1);elseerror(['CircularHough_Grd: "multirad" has to be'within the range [0.1, 1]*]);endendvap_fltr4accum = 4; % filter for smoothing the accumulation arrayif nargin > (1 + vap_fltr4accum),if isnumeric(varargin{vap_fltr4accum}) && ...ndims(varargin{vap_fltr4accum}) == 2 && ...all(size(varargin {vap_fltr4accum}) >= 3),fltr4accum = varargin {vap_fltr4accum};elseerror(['CircularHough_Grd: n fltr4accum n has to be \ ...*a 2-D matrix with a minimum size of 3-by-3']);endelse% Default filter (5-by-5)fltr4accum = ones(5,5);fltr4accum(2:4,2:4) = 2;fltr4accum(3,3) = 6;end func_compu_cen = (nargout > 1 );func_compu_radii = (nargout > 2 );% Reserved parametersdbg on = false; % debug information dbgbfigno = 4;if nargout > 3, dbg on = true; end%%%%%%%% Buildingaccumulation array %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Convert the image to single if it is not of% class float (single or double) img_is_double = isa(img, double');if ~(img_is_double || isa(img, 'single')),imgf = single(img);end% Compute the gradient and the magnitude of gradientif img_is_double,[grdx, grdy] = gradient(img);else[grdx, grdy] = gradient(imgf);endgrdmag = sqrt(grdx.A2 + grdy.A2);% Get the linear indices, as well as the subscripts, of the pixels% whose gradient magnitudes are larger than the given threshold grdmasklin = find(grdmag > prm_grdthres);[grdmask_ldxl, grdmask_IdxJ] = ind2sub(size(grdmag), grdmasklin);% Compute the linear indices (as well as the subscripts) of% all the votings to the accumulation array.% The Matlab function 'accumarray* accepts only double variable, % so all indices are forced into double at this point.% A row in matrix ,lin2accum_aJ, contains the J indices (into the % accumulation array) of all the votings that are introduced by a % same pixel in the image. Similarly with matrix linZaccum aP. rr_41inaccum = double( prm_r_range );linaccum_dr = [ (-rr_41inaccum(2) + 0.5) : -rr_41inaccum(l),... (rr_41inaccum(l) + 0.5) : rr_41inaccum(2)];lin2accum_aJ = floor(...double(grdx(grdmasklin)./grdmag(grdmasklin)) * linaccum_dr + ...repmat( double(grdmask_IdxJ)+0.5 , [ 1 ,length(linaccum_dr)])...);lin2accum_al = floor(...double(grdy(grdmasklin)./grdmag(grdmasklin)) * linaccum dr + ...repmat( double(grdmask_IdxI)+0.5 , [1 ,length(linaccum_dr)])...)；% Clip the votings that are out of the accumulation arraymask_valid_a J al =...lin2accum_aJ > 0 & lin2accum_aJ < (size(grdmag,2) + 1) & ...Iin2accum_al > 0 & lin2accum_al < (size(grdmag,l) + 1);mask_valid_aJaI_reverse =〜mask_valid_aJaI;lin2accum_aJ = lin2accum_aJ .* maskvalida J al + maskvalidaJ alreverse;lin2accum_al = lin2accum_al .* mask_valid_aJaI + mask_valid_aJaI_reverse;clear mask_valid_aJ alre verse;% Linear indices (of the votings) into the accumulation arraylin2accum = sub2ind( size(grdmag), lin2accum_al, lin2accum_aJ );lin2accum_size = size( lin2accum );lin2accum = reshape( lin2accum, [numel(lin2accum),l]);clear lin2accum_al lin2accum_aJ;% Weights of the votings, currently using the gradient maginitudes% but in fact any scheme can be used (application dependent)weight4accum =...repmat( double(grdmag(grdmasklin)) , [lin2accum_size(2), 1 ]) .* ...mask_valid_aJ al(:);clear mask_valid_aJaI;% Build the accumulation array using Matlab function 'accumarray'accum = accumarray( lin2accum , weight4accum );accum = [ accum ; zeros( numel(grdmag) - numel(accum), 1 )];accum = reshape( accum, size(grdmag));%%%%%%%% Locating local maxima in the accumulation array %%%%%%%%%%%%% Stop if no need to locate the center positions of circlesif ~func_compu_cen,return;endclear lin2accum weight4accum;% Parameters to locate the local maxima in the accumulation array% — Segmentation of 'accum' before locating LM prmuseaoi = true;prm_aoithres_s = 2;prm aoiminsize = floor(min([ min(size(accum)) * 0.25,... prm_r_range(2) * 1.5 ]));% — Filter for searching for local maxima prmfltrLMs = 1.35;prm fltrLM r = ceil( prm fltrLM R * 0.6 );prm fltrLM npix = max([ 6, ceil((prm_fltrLM_R/2)A 1.8)]);% — Lower bound of the intensity of local maximaprm LM LoBndRa = 0.2; % minimum ratio of LM to the max of'accum'% Smooth the accumulation arrayfltr4accum = fltr4accum / sum(fltr4accum(:));accum = filter2( fltr4accum, accum );% Select a number of Areas-Of^Interest from the accumulation array if prmuseaoi, % Threshold value for 'accum1prm_llm_thresl = prm_grdthres * prm_aoithres_s;% Thresholding over the accumulation array accummask = ( accum > prm llm thres 1 );% Segmentation over the mask[accumlabel, accum nRgn] = bwlabel( accummask, 8 );% Select AOIs from segmented regionsaccumAOI = ones(0,4);for k = 1 : accum nRgn,accumrgn lin = find( accumlabel = k);[accumrgn_ldxl, accumrgn_IdxJ]=...ind2sub( size(accumlabel), accumrgn lin);rgn top = min( accumrgn ldxl);rgn bottom = max( accumrgn_ldxl);rgn left = min( accumrgn ldxJ );rgn_right = max( accumrgn ldxJ );% The AOIs selected must satisfy a minimum sizeif ((rgn_right - rgn_left + 1) >= prm_aoiminsize && ...(rgn_bottom - rgn top + 1) >= prm aoiminsize ),accumAOI = [ accumAOI;...rgn top, rgn bottom, rgn left, rgn right ];endendelse% Whole accumulation array as the one AOIaccumAOI = [1, size(accum,l), 1, size(accum,2)];end% Thresholding of 'accum' by a lower boundprm LM LoBnd = max(accum(:)) * prm LM LoBndRa;% Build the filter for searching for local maxima fltr4LM = zeros(2 * prm_fltrLM_R + 1);[mesh4fLM_x, mesh4fLM_y] = meshgrid(-prm_fltrLM_R : prm fltrLM R);mesh4fLM_r = sqrt( mesh4fLM_x.A2 + mesh4fLM_y.A2 );fltr4LM_mask =...(mesh4fLM_r > prm_fltrLM_r & mesh4fLM_r <= prm fltrLM R );fltr4LM = fltr4LMfltr4LM_mask * (prm fltrLM s / sum(fltr4LM_mask(:)));if prm_fltrLM_R >= 4,fltr4LM_mask = ( mesh4fLM_r < (prm_fltrLM_r - 1))；elsefltr4LM_mask = ( mesh4fLM_r < prm fltrLM r );endfltr4LM = fltr4LM + fltr4LM mask / sum(fltr4LM_mask(:));% **** Debug code (begin)if dbg_on,dbg_LMmask = zeros(size(accum));end% **** Debug code (end)% For each of the AOIs selected, locate the local maximacircen = zeros(0,2);fbrk = 1 : size(accumAOI, 1),aoi = accumAOI(k,:); % just for referencing convenience% Thresholding of 'accum* by a lower boundaccumaoi_LBMask =...(accum(aoi(l):aoi(2), aoi(3):aoi(4)) > prm LM LoBnd );% Apply the local maxima filtercandLM = conv2( accum(aoi( 1):aoi(2), aoi(3):aoi(4)),...fltr4LM, 'same*);candLM mask = ( candLM > 0 );% Clear the margins of 'candLM mask*candLM_mask([l :prm_fltrLM_R, (end-prm_fltrLM_R+l):end], :) = 0;candLM mask(:, [l:prm_fltrLM_R, (end-prm_fltrLM_R+l):end]) = 0;% **** Debug code (begin)if dbg_on,dbg_LMmask(aoi( 1 ):aoi(2), aoi(3):aoi(4))=...dbg_LMmask(aoi( 1 ):aoi(2), aoi(3):aoi(4)) + ...accumaoi LBMask + 2 * candLM mask;end% **** Debug code (end)% Group the local maxima candidates by adjacency, compute the% centroid position for each group and take that as the center% of one circle detected[candLM label, candLM nRgn] = bwlabel( candLM_mask, 8 );fbr ilabel = 1 : candLM nRgn,% Indices (to current AOI) of the pixels in the groupcandgrp masklin = find( candLM label == ilabel);[candgrp_ldxl, candgrp_IdxJ]=...ind2sub( size(candLM label), candgrp masklin );% Indices (to 'accum') of the pixels in the groupcandgrp_ldxl = candgrp_ldxl + ( aoi(l) - 1 );candgrp IdxJ = candgrp IdxJ + ( aoi(3) - 1 );candgrp_idx2acm =...sub2ind( size(accum) , candgrp ldxl, candgrp IdxJ );% Minimum number of qulified pixels in the groupif sum(accumaoi_LBMask(candgrp_masklin)) < prm_fltrLM_npix, continue;end% Compute the centroid positioncandgrp_acmsum = sum( accum(candgrp_idx2acm));cc_x = sum( candgrp IdxJ .* accum(candgrp_idx2acm) ) / ...candgrpacmsum;cc_y = sum( candgrp_ldxl .* accum(candgrp_idx2acm) ) / ...candgrpacmsum;circen = [circen; cc_x, cc_y];endend% **** Debug code (begin)if dbg_on,figure(dbg bfigno); imagesc(dbg LMmask); axis image;title(*Generated map of local maxima1);if size(accumAOI, 1) == 1,figure(dbg_bfigno+1);surf(candLM, 'EdgeColor1, hone'); axis ij;title(,Accumulation array after local maximum filtering*);endend% **** Debug code (end)%%%%%%%% Estimation of the Radii of Circles %%%%%%%%%%%%% Stop if no need to estimate the radii of circlesif ~func_compu_radii,varargout{l} = circen;return;end% Parameters for the estimation of the radii of circlesfltr4SgnCv=[2 1 1];fltr4SgnCv = fltr4SgnCv / sum(fltr4SgnCv);% Find circle's radius using its signature curve cirrad = zeros( size(circen,l), 1 );for k = 1 : size(circen,l),% Neighborhood region of the circle for building the sgn. curve circen_round = round( circen(k,:));SCvR IO = circen_round(2) - prm_r_range(2) - 1;ifSCvR_IO<l,SCvR_I0= 1;endSCvRIl = circen_round(2) + prm_r_range(2) + 1;if SCvR Il > size(grdx,l),SCvRIl = size(grdx,l);endSCvR JO = circen round(l) - prm_r_range(2) - 1;ifSCvR_JO<l,SCvRJO = 1;endSCvRJ 1 = circenround(l) + prm_r_range(2) + 1;if SCvR Jl > size(grdx,2),SCvRJl = size(grdx,2);end% Build the sgn. curveSgnCvMat_dx = repmat( (SCvR J0:SCvR J 1) - circen(k,l),...[SCvRJl - SCvRJO +1,1]);SgnCvMat_dy = repmat( (SCvR_IO:SCvR_Il)' - circen(k,2),...[1 , SCvRJl - SCvRJO + 1]);SgnCvMat_r = sqrt( SgnCvMat dx .A2 + SgnCvMat_dy .A2 );SgnCvMatrpl = round(SgnCvMatr) + 1;f4SgnCv = abs(...double(grdx(SCvR_IO:SCvR_Il, SCvRJO:SCvRJ 1)) .* SgnCvMat_dx + ...double(grdy(SCvR_IO:SCvR Il, SCvR JO:SCvR J 1)) .* SgnCvMat dy...)./ SgnCvMat r;SgnCv = accumarray( SgnCvMat rp 1(:) , f4SgnCv(:));SgnCv_Cnt = accumarray( SgnCvMat rp 1 (:) , ones(numel(f4SgnCv), 1));SgnCv_Cnt = SgnCv_Cnt + (SgnCv_Cnt == 0);SgnCv = SgnCv ./ SgnCv_Cnt;% Suppress the undesired entries in the sgn. curve% ― Radii that correspond to short arcsSgnCv = SgnCv .* ( SgnCv_Cnt >= (pi/4 * [O:(numel(SgnCv_Cnt)-1 )]*));% ― Radii that are out of the given rangeSgnCv( 1 : (round(prm_r_range( 1))+1) ) = 0;SgnCv( (round(prm_r_range(2))+1) : end ) = 0;% Get rid of the zero radius entry in the arraySgnCv = SgnCv(2:end);% Smooth the sgn. curveSgnCv = filtfilt( fltr4SgnCv , [1] , SgnCv );% Get the maximum value in the sgn. curveSgnCv_max = max(SgnCv);if SgnCv_max <= 0,cirrad(k) = 0;continue;end% Find the local maxima in sgn. curve by 1st order derivatives% ― Mark the ascending edges in the sgn. curve as Is and% ― descending edges as OsSgnCv AscEdg = ( SgnCv(2:end) - SgnCv(l:(end-l)) ) > 0;% ― Mark the transition (ascending to descending) regionsSgnCv LMmask = [ 0; 0; SgnCv_AscEdg(l:(end-2)) ] & (〜SgnCv_AscEdg);SgnCv LMmask = SgnCvLMmask & [ SgnCv_LMmask(2:end); 0 ];% Incorporate the minimum value requirementSgnCvLMmask = SgnCvLMmask & ...(SgnCv(l:(end-l)) >= (prm_multirad * SgnCv_max));% Get the positions of the peaksSgnCv LMPos = sort( find(SgnCv_LMmask));% Save the detected radiiif isempty(SgnCvLMPos),cirrad(k) = 0;elsecirrad(k) = SgnCvLMPos(end);for i radii = (length(SgnCv LMPos) - 1) : -1 : 1,circen = [ circen; circen(k,:)];cirrad = [ cirrad; SgnCv_LMPos(i_radii)];endendend% Outputvarargout{l} = circen;varargout{2} = cirrad;if nargout > 3,varargout{3} = dbg_LMmask;endprograms：programs:2 function DrawCircle (x, y, r, nseg, S)% Draw a circle on the current figure using ploylines%% DrawCircle (x, y, r, nseg, S)% A simple function for drawing a circle on graph.%% INPUT: (x, y, r, nseg, S)% x, y: Center of the circle% r: Radius of the circle% nseg: Number of segments for the circle% S: Colors, plot symbols and line types%% OUTPUT: None%% BUG REPORT:% Please send your bug reports, comments and suggestions to% pengtao@glue. umd. edu . Thanks.% Author: Tao Peng% Department of Mechanical Engineering% University of Maryland, College Park, Maryland 20742, USA % pengtao@glue. umd. edu% Version: alpha Revision: Jan. 10, 2006theta = 0 : (2 * pi / nseg) : (2 * pi);pline_x 二r * cos(theta) + x;pline_y 二r * sin(theta) + y;plot (pline_x, pline_y, S);3function testiml二imread (' image 1. jpg');% rawimg = imread(，TestImg_CHT_c3. bmp J);rawimg=rgb2gray(iml);tic;[accum, circen, cirrad] = CircularHough_Grd(rawimg, [20 30], 5,50);circentoe;figure(1) ; imagesc(accum); axis image;title (J Accumulation Array from Circular Hough Transform，); figure (2) ; imagesc (rawimg) ; colormap (J gray，) ; axis image; hold on;plot (circen(:, 1), circen(:, 2), ' r+');for k = 1 : size (circen, 1),DrawCircle (circen (k, 1), circen (k, 2), cirrad (k), 32, ' b-'); end hold off; title(f Raw Image with Circles Detected ...'(center positions and radii marked)']);figure (3); surf(accum, ' EdgeColor，, ' none5); axis ij; title (J 3-D View of the Accumulation Array');附带图像image 1. jpg直接运行test.m即可得到上方的结果！当然方法是活的，只要合理即可行。
- 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)。
– 用有监督的方式训练Hough森林。训练包括建树、叶子节点信 息的分配；测试时样本从每个树开始传递，输出是到达的所 有叶子节点分布的平均。 – 训练数据和叶子信息：训练的patch表示 训练数据和叶子信息 数据和叶子信息： 为： ，分别代表patch的特征、patch的类 别（物体或是背景）、patch中心到bounding box中心的偏移 量。叶子节点存储总的类别构成和偏移量，形成码本。
上面两项都可以直接求得。
Hough Forests （7）——算法
对于一棵树，用核密度估计：
对整个森林，取各个树的平均。每个位置的得分是来自 bounding box内各个patch投票的总和。所以检测输出最大得分的位 置及其置信度。 注：为了提高速度，文中提到简化的策略。
处理不同尺度。 处理不同尺度。将测试图像按尺度缩放，共S组。各个尺度 内进行投票，最后得到得分最高的位置。
位置投票，选取最大值作为物体的中心。
创新点 创新点：Hough森林的引入。Hough森林的建立、在物体检测中的应用 是本文的重点。 对比： 对比：ISM和BFM用产生式的码本，而Hough森林的方法是判别式，所以 后者速度快。
Hough Forests （2）——算法
Hough森林的特点。 Hough森林的特点。 森林的特点
的部分投票，本文只用物体的边缘投票。
BFM（2）——算法
学习轮廓片段。 学习轮廓片段。
– 主要涉及对轮廓片段打分和选择问题。 – 需要训练集（已用bounding box划定了物体）和验证集（标 注了物体是否存在和物体的中心，但没用bounding box）。
BFM（3）——算法
对轮廓片段打分
候选的轮廓片段需要满足：（i）匹配的边缘链经常在正例 样本而不在负例样本；（ii）能较好的预测正例中的物体中心。根 据以上两点对轮廓片段打分。轮廓片段的代价函数为：
– 叶子节点存的是具有判别性的码本（一个patch是来自物体还 是来自背景，物体中心距离当前patch中心的位置）。 – 建立Hough森林可以优化投票性能，即叶子节点投票时的不确 定度将降低。 – 用有监督的方法建立树，即一个patch是来自物体还是来自背 景， patch来自物体的哪个部分。
Hough Forests （3）——算法
Hough Forests （4）——算法
– Patch的特征和二值测试：一个patch的C种特 Patch的特征和二值测试： 的特征和二值测试 征 ，同一种特征来自两个位置的二值测试定 义为：
– 树的建立：建树时，每个节点得到很多patch，如果达到成为 的建立： 叶子节点的标准，该节点作为叶子节点。否则，需要拆分该 节点。 – 一个好的二值测试，应该能使得后继结点的数据尽可能的 一个好的二值测试， 也就是说，越往叶子节点， “纯”。也就是说，越往叶子节点，类别和偏移量的不确定 性应该越低。 性应该越低。分别用类别不确定性和偏移量不确定性来度量 节点的不纯度，定义为： 节点的不纯度，定义为：
相关3 Hough voting 相关3篇文献
淮静 2011.11.11
Outline
文献1：Combined Object Categorization and Segmentation with an Implicit Shape Model，ECCV 04 Workshop。 文献2：A Boundary-Fragment-Model for Object Detection，ECCV 06。 文献3： Class-Specific Hough Forests for Object Detection，CVPR 09。
步骤：提取边缘——用强分类器匹配——每个弱分类器投票， 得到候选点——在候选点附件窗口内，用mean-shift的到新的中 心——把对中心投票过的弱分类器重建出来——检测物体——分割。
Hough Forests （1）——简介
内容简介： 内容简介：通过训练特定类的Hough森林，由各个部分对物体的中心
ISM（1）——简介
主要内容： 主要内容：
– 用隐形状模型 隐形状模型（Implicit Shape Model, ISM）把物体的检测 隐形状模型 和分割结合起来。
创新点： 创新点：
– 用隐形状模型把物体的识别和分割结合起来，整合到同一个 概率框架下。 – 用基于最小描述长度（Minimaห้องสมุดไป่ตู้ Description Length, MDL） 的标准来做多个物体检测问题。
基于最小描述长度的规则做多个物体场景分析： 基于最小描述长度的规则做多个物体场景分析：
最小描述长度的规则要求最小化图像、模型和错误的总描述长 度。提出savings的概念：
指可以被假设h解释的像个数； 指模型复杂度。 两个假说时：
指描述误的代价；
BFM（1）——简介
内容简介： 内容简介：
– 提出了“Boundary-Fragment-Model”（BFM）检测器，用物 体的轮廓来做检测。
– 针对各自不太奏效的样本
综合以后的方法如何评价好坏？跟哪些方法做比较？
ISM（4）——算法
概率形式化： 概率形式化：
第一项是给定物体类别、patch的解释后，hough投票物体位置的概 率；第二项是码本匹配的是物体而不是背景的概率；第三项是patch和码本 匹配的概率。 一个中心作为特定物体的中心的得分为：
ISM（5）——算法
物体分割： 物体分割：
ISM（6）——算法
创新点 创新点：
– 学习轮廓片段码本的方式。片段不但具有较高的类别信息， 并且能够稳定的预测物体的中心。 – 用Boosting方法把一系列基于边缘片段的弱检测器构造成强 检测器。有这些检测器，检测物体时不再需要滑动窗对物体 定位.。
对比： 对比：与ISM（implicit shape model）不同在于，ISM用所有
– 弱分类器 弱分类器：几个轮廓片段的组合就有了较好的判别信息，可 以组成弱分类器。如下图：如果几个轮廓片段能匹配图像的 边缘链，他们预测的中心比较靠近，并且与真正的物体中心 接近。则形成弱分类器。 – 强分类器 强分类器：用Adaboost的方 法由弱分类器得到强分类器。
BFM（6）——算法
物体检测。 物体检测。
ISM（2）——算法
码本： 码本：Harr检测器提取感兴趣点，提取patch，合并相似的
patch，将类中心存入码本。 类相似度定义：
其中：
ISM（3）——算法
物体识别的过程： 物体识别的过程：Harr检测器提取感兴趣点——提取patch——与
码本匹配——对物体中心投票——得到最大得分——搜集相关patch—— 提取相关patch附件的patch。
Hough Forests （8）——实验
将图像设置为相同的高度。 训练时先随机选取正反例建立5棵树，再用更难区分的 正反例建立5棵树，直到有15棵树时停止，Hough森林 建立成功。 特征用颜色、梯度、HOG。
Discussion
Hough voting 方法的特点是什么？有没有什么缺点？ 如何与显著性的方法综合？
Hough森林做物体检测。 Hough森林做物体检测。 森林做物体检测
假设特定物体类别的bounding box大小固定，只要找到 bounding box的中心即可。 E(x)代表随机事件——物体中心位于x。给定一个 patch ，,当patch的中心在 bounding box内时, 物体中心位于x的概率为：
Hough Forests （5）——算法
– 二值测试过程为： 二值测试过程为： 给定一些patch，均匀采样得到一系列测试像素；随机选择最下 化类别不确定性还是偏移量不确定性；对不确定性求和：
随机选取可以保证叶子结点的类别不确定性和偏移量不确定 性都比较低。
Hough Forests （6）——算法
其中：
BFM（4）——算法
选择轮廓片段 选择轮廓片段
步骤：在轮廓片段上随机撒一些种子——让种子在轮廓片段 上生长，并随时计算在验证集的代价 ——选择最佳的轮廓片段，得 到码本。码本中有轮廓片段相对于中心的几何信息。——对码本进 行合并，降低冗余。
BFM（5）——算法
用boosting的方法训练物体检测器。 boosting的方法训练物体检测器。 的方法训练物体检测器