《计算机视觉.ppt》由会员分享，可在线阅读，更多相关《计算机视觉.ppt（26页珍藏版）》请在三一办公上搜索。
1、Sept.17,2010,计算机视觉 Computer Vision,Sept.17,2010,Outline,课程目标，资料来源，授课方式教材与参考书、作业、课程设计、考核方式Web sitesFTP sourcesTools(Intel OpenCV,IPL,)Demo相关学科与相关课程的联系OverviewIntroduction recommended：Forsyths introduction to CV,other related,Sept.17,2010,课程目标，资料，授课方式,目标全面了解计算机视觉研究领域，掌握基本原理，具有一定的实践能力。资料使用网上公开的资料，挑选、推荐
2、各种资料（ppt,pdf,codes,etc.），并直接使用我认为合适的讲稿进行讲解。授课方式课堂上全面介绍基本原理，引导资料阅读；课后需要学生阅读相关文献，消化理解。如有时间，建议课前预习。,学习方法的建议,根据自己学习本门课程的目的考虑投入时间一般了解、掌握、精通根据自己研究方向有所侧重有选择地专研部分内容将精读与一般了解相协调将建立基础知识与把握当前研究热点相结合结合自己的研究课题加深相关基础知识了解本方向的研究情况，研究趋势等。,Sept.17,2010,实践：课程设计、专题研究,实践非常重要，CV是实验学科，脱离实践，很难有所收获。课程设计（程序作业）可以使用OpenCV工具，可以使
3、用Mathlab工具，充分利用其他网上资源专题研究结合自己的课题，挑选论文精读，实现论文中的方法，分析存在的问题，如何解决，如何改进，,Sept.17,2010,Sept.17,2010,教材,英文原版：Szeliski,Richard,Computer Vision:Algorithms and Applications,Springer,Oct.,2010中文翻译版：艾海舟，兴军亮 等，计算机视觉：算法与应用，2011年 x月（预计2011年内出版）；清华大学出版社 英文影印版：DA Forsyth and J.Ponce,Computer Vision:A Modern Approach
4、,Prentice Hall.1st edition(August 14,2002);清华大学出版社 中文翻译版：林学言，王宏 等，计算机视觉：一种现代的方法，2004年6月；电子工业出版社,Sept.17,2010,参考书,马颂德，张正友，计算机视觉，科学出版社，北京，1998。R.Jain,R.Kasturi and B.G.Schunck,Machine Vision,McGraw-Hill companies,Inc.机械工业出版社，2003.8。L.G.Shapiro and G.C.Stockman,Computer Vision,Prentice Hall Inc,2001.M.
5、Sonka,V.Hlavac,and R.Boyle,Image processing,analysis,and machine vision,Chapman&Hall Computing,London,3rd Edition,THOMSON Learning,2008.M.Sonka,V.Hlavac,and R.Boyle,(艾海舟、苏延超 等译),图像处理、分析与机器视觉（第3版），清华大学出版社,2011.1。章毓晋，图象工程（第2版），清华大学出版社,2007.5,Sept.17,2010,参考书(模式识别与机器学习),Christopher M.Bishop,Pattern Rec
6、ognition and Machine Learning,Springer,2006.8R.O.Duda,P.E.Hart and D.G.Stork,Pattern Classification，机械工业出版社，2003.6。R.O.Duda,P.E.Hart and D.G.Stork(李宏东，姚天翔等译),模式分类，机械工业出版社，2003.9。S.Theodoridis and K.Koutroumbas,Pattern Recognition,机械工业出版社，2003.9。边肇祺，张学工 等，模式识别，清华大学出版社，2000.,Sept.17,2010,参考书(图像处理),R.C
7、.Gonzalez,R.E.Woods，Digital Image Processing,第三版(影印)电子工业出版社&Pearson Education,2010.R.C.Gonzalez,R.E.Woods,(阮秋琦、阮宇智等译)，数字图像处理，第二版，电子工业出版社&Prentice Hall,2003.章毓晋，图象工程（第2版），清华大学出版社,2007.5 郎锐，数字图像处理学Visual C+实现，北京希望电子出版社，2002.12周长发，精通Visual C+图像编程，电子工业出版社，2000.1,Sept.17,2010,作业、课程设计、考核方式,书面作业课程设计考核方式书面作
8、业 10%课程设计 60%期末考试 30%,Sept.17,2010,Web sites(1)-Search Engine,CVPapers-Computer Vision Resourcehttp:/search computer visionComputer vision homepageComputer vision onlineComputer vision source codesComputer vision test dataComputer vision.Paper search http:/,Sept.17,2010,Web sites(2)-Courses,计算机视觉 Sl
9、ides and lectures of Szeliskis books supplementary materialUW455:Undergraduate Computer Vision,http:/www.cs.washington.edu/education/courses/455/.UW576:Graduate Computer Vision,http:/www.cs.washington.edu/education/courses/576/.Stanford CS233B:Introduction to Computer Vision,http:/vision.stanford.ed
10、u/teaching/cs223b/.MIT 6.869:Advances in Computer Vision,http:/people.csail.mit.edu/torralba/courses/6869/putervision.htm.Berkeley CS 280:Computer Vision,http:/www.eecs.berkeley.edu/trevor/CS280.html.UNC COMP 776:Computer Vision,http:/www.cs.unc.edu/lazebnik/spring09/.Middlebury CS 453:Computer Visi
11、on,http:/www.cs.middlebury.edu/schar/courses/cs453-s10/.,Sept.17,2010,Web sites(3)-Course Ware,计算机视觉”课程的网上课件入口Computer Vision Education Digital Library Collectionhttp:/cved.org/Computer Visionhttp:/www.cs.washington.edu/education/courses/576/CurrentQtr/Introduction to Computer Vision http:/www.cse.p
12、su.edu/cg486/Learning and Inference in Visionwww.ai.mit.educourses6.899,Sept.17,2010,Web sites(4)-Codes,tutorial,etc.,KLT:An Implementation of the Kanade-Lucas-Tomasi Feature Trackerhttp:/www.ces.clemson.edu/stb/klt/installation.htmlEpipolar geometry,essential matrix,etc:online tutorialhttp:/homepag
13、es.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/EPSRC_SSAZ/node18.htmlRANSAChttp:/homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/FISHER/RANSAC/.,Sept.17,2010,Tools(1),Intel OpenCV(Open source Computer Vision library)This library allows high level functions for computer vision and image processing.OpenCV
14、 offers many high-level data types such as sets,trees,graphs,matrices.OpenCV is open source to run on many computer platforms.High level functions such asCamera calibration(Zhang Zhengyous method)Face detection(a variation of Viola-Joness detector)Motion analysis and object trackingOptical flowLucas
15、-Kanade algorithmEstimatorsKalmanCondensation,Sept.17,2010,Tools(2),Intel IPP(Integrated Performance Primitives)It is a signal processing,image processing and matrix calculation library developed by Intel Corporation.IPP offers to programmers a wide range of low-level functions which are optimized w
16、hen used on an Intel processor(from Pentium to Itanium).It is really a good library for signal,image,video and sound processing with very good performances due to optimized instructions.IPP is not a free library,it comes under an Intel licensing policy which is explained at Intel website,Sept.17,201
17、0,demo,Face detectionObject contour trackingMotion object detection and trackingASM/AAM shape modelingPerceptual interface:smart roomVisual surveillanceRobotics vision3D modeling,face animation,Sept.17,2010,相关学科与相关课程的联系,数字图象处理,计算机视觉,模式识别,机器视觉,计算机图形学,线性代数,集合论,高级语言程序设计,数据结构,先后顺序,重叠量反应相关程度,基础知识,计算机视觉专题
18、（图象与视觉计算）,高等代数,最优化方法,。,信号与系统,计算几何,Sept.17,2010,Overview(1),计算机视觉的几何学基础摄像机模型单摄像机(pinhole model/perspective transformation)双摄像机(epipolar geometry:fundamental matrix/essential matrix)三摄像机及更多(multi-view geometry)运动估计对应点问题（correspondence problem）光流计算方法刚体运动参数估计（minimal projective reconstruction）2-view,7 p
19、oints in correspondence;(Faugeras)3-view,6 points in correspondence;(Quan Long)3-view,8 points with one missing in one of the three view.(Quan Long)几何重构（Geometry reconstruction）立体视觉(stereo vision)Shape from X(shading/motion/texture/contour/focus/de-focus/.),Sept.17,2010,Overview(2),计算机视觉的物理学基础摄像机及其成
20、像过程视点、光源、空间中光线、表面处的光线.明暗 shading、阴影 shadow光学/色彩 light/color辐射学，辐照率radiometry,物体表面特性漫反射表面（各向同性）Lambertian surfaceBDRF(bi-directional reflectance distribution fucntion),Sept.17,2010,Overview(3),计算机视觉的图像模型基础摄像机模型及其校准内参数、外参数图像特征边缘、角点、轮廓、纹理、形状图像序列特征(运动)对应点、光流,Sept.17,2010,Overview(4),计算机视觉的信号处理层次低层视觉处理单图
21、像：滤波/边缘检测/纹理多图像：几何/立体/从运动恢复仿射或透视结构 affine/perspective structure from motion中层视觉处理聚类分割/拟合线条、曲线、轮廓 clustering for segmentation,fitting line基于概率方法的聚类分割/拟合跟踪 tracking高层视觉处理匹配模式分类/关联模型识别 pattern classification/aspect graph recognition应用距离数据（range data）/图像数据检索/基于图像的绘制,Sept.17,2010,Overview(5),计算机视觉的数学基础摄影
22、几何、微分几何概率统计与随机过程数值计算与优化方法机器学习计算机视觉的基本的分析工具和数学模型Signal processing approach:FFT,filtering,wavelets,Subspace approach:PCA,LDA,CCA,ICA,Bayesian inference approach:EM,Condensation/SIS/,MCMC,.Machine learning approach:SVM/RVM/Kernel machine,Boosting/Adaboost,NN/Regression,HMM,BN/DBN,Gibbs,MRF,Sept.17,2010
23、,Overview(6),计算机视觉问题的特点高维数据的本质维数很低，使得模型化成为可能。High dimensional image/video data lie in a very low dimensional manifold.问题的不适定性 缺少约束的逆问题优化问题,Sept.17,2010,Introduction Recommended,Forsyths introduction to CVUllmans introduction to Computer and Human Vision(part 1)Ullmans introduction to Computer and Human Vision(part 2)Seitzs conclude on CV,Sept.17,2010,CV 文献,杂志 IJCV,PAMI,CVIU,PR,IVC等会议 ICCV,CVPR,ECCV,FG,ACCV,ICPR,ICIP 等,
