智能监控系统算法研究
|
摘要
在视觉监视领域日益追求自动化的今天，设计一种能完全替代监控人员、自动对监视场景中运动目标(诸如人，汽车)进行实时、持续的主动监视的高智能化监控系统就显得非常迫切。这种高智能化监控系统一般要求具有运动检测、目标跟踪等自主功能。利用现有计算机强大的计算能力以及图像处理与分析、计算机视觉等多学科知识，完全有可能设计出这种无人值守下的高智能化监控系统。遗憾的是，目前国内尚无这种系统面诸于世。阻碍这种高智能化监控系统设计的最大难题是缺乏一个统观全局的算法设计，算法往往达不到前后模块链接紧密、复杂度不高、漏警误警率低等实用标准。本文就是在这样的背景下提出的，通过对现有多种算法的研究与分析，在借鉴前人研究的经验和教训的基础上，对运动检测和目标跟踪模块进行了算法上的精简与改进，使算法复杂度大为降低。整个算法的设计在做到实时性要求的同时尽量兼顾系统的鲁棒性，具有较为广阔的应用前景。
文章首先详细介绍了当前视觉监视理论与技术在国内外的研究现状和发展趋势，论文按功能划分为运动检测和目标跟踪两个模块分别进行深入探讨。
运动检测功能模块是智能化监控系统的首要环节，其算法设计的好坏是系统成败的关键，它也占据了整个系统大部分运算量，主要包括前处理子模块，运动检测子模块，后处理子模块。前处理子模块在分析传统滤波器优缺点的基础上，引入像素加权和像素分类的思想，构造出一个改进型多级中值滤波器，不仅具有良好的滤除性能，还较好的保护了图像的细节。运动检测子模块中对传统的算法进行了性能评析，指出各种传统算法的优缺点，重点提出了动态背景减除改进算法，将时域差分法和动态背景法的优点综合起来，构造一种动态更新的背景模型，达到了较好的效果。后处理子模块采用形态学滤波来去噪、补空洞，文章对传统的形态学后处理方案进行了改进。
目标跟踪功能模块是智能化监控系统的重要环节，其中的活动摄像机
浙江大学硕士学位论文
对运动目标的跟踪尤其具有实用价值。文章首先详细分析了固定背景下的
传统匹配跟踪算法。接着重点提出了一种运动目标的运动平滑性和尺寸连
续性等特征，在二值图像层面上对运动目标区域的外接矩形进行跟踪的计
算简单的算法。该算法充分考虑了互遮挡、目标短暂消失和永久移入、移
出问题，可用于多目标的跟踪，具有较强的鲁棒性。最后简要的分析了活
动摄像机下的目标跟踪问题。
文章的最后给出了实验的部分演示序列并对其加以了说明，分析了本
论文中算法上的不足，并指出尚待改进的方向。
In the modern time of pursuing for the high automatization in the field of visual surveillance, the desire for a smart surveillance system, which is the substitute of surveillance members and can monitoring the moving object(such as human, cars and so on) in some surveillance scape automatically and consecutively, is very strong. In most cases, this intelligent system has some functions such as motion detection and object tracking etc. It's very possible to design this system which no one needed to watch by full utilizing the powerful calculating capacity of modern computer and many subjects such as image processing and analyzing, computer vision etc. But regrettably this kind of system is not appeared in our country currently. The most difficult problem is lack of a device of global algorithm. Common algorithm often cannot reach the practical standards such as tight linking among modules, low algorithm complexity, low ratio of missing alarm and false alarm etc. This paper is put forward in such a background
. The algorithm complexity is more lowered by simplifying and improving the algorithm on the two modules of motion detection and object tracking on the basis of researching and analyzing many other person's algorithms and studying others' research fruits. The global algorithm synthetically considers the real-time property and robust performance, and it can be used in surveillance system.
Firstly this paper introduces the research status and tendency in the world in detail. The whole paper is divided by two modules of motion detection and object tracking based on its function, and the article elaborates them one by one.
Motion detection module is the most important part in the intelligent system. Its device is the key of the whole system and it also holds the most calculating task. This module is divided by three sub-modules of pre-processing, motion detection and post-processing. In the sub-module of pre-processing, the paper analyzes performance of traditional filters and points out its advantages and disadvantages, then presents a modified multilevel median filter by importing the idea of weight and classification. The proposed filter can not only suppress noise more effectively but also preserve the features of images. In the sub-module of motion detection, the article first analyzes performance
of traditional algorithm and points out some disadvantages, and then brings forward one improved algorithms: the modified dynamic background deduction algorithm, which puts the advantages of temporal differencing algorithm and dynamic background deduction algorithm together to present a dynamic background model. Experimental results are given to demonstrate its validity and practicality. The post-processing sub-module uses morphological filter to remove isolated noise and to fill the holes inside objects, and the traditional method of this sub-module is improved in the paper.
Object tracking module is also an important part in the intelligent system. Among them the object tracking with an Active Camera has much more practical value. Firstly, the article introduces the traditional matching tracking algorithm detailedly in the fixed background. Secondly, a simple tracking algorithm integrated with many moving objects' characters such as smooth motion property and consecutive size property is brought forward. The simplified algorithm is just tracking the outer-rectangle of moving object area in-the binary image. Carefully considering the problems of the occlusion, temporary disappearance and being permanently moved in or moved out of objects detailedly, this algorithm can be used in the tracking of multi-object and its robust performance is also strong. Finally, the paper analyzes the object tracking with an active camera.
At the end of this paper, some demo sequences of experiment are given and elaborated. Finally the paper introduces shortcomings of proposed algorithms and aspects deserved to be strengthened.
文章首先详细介绍了当前视觉监视理论与技术在国内外的研究现状和发展趋势，论文按功能划分为运动检测和目标跟踪两个模块分别进行深入探讨。
运动检测功能模块是智能化监控系统的首要环节，其算法设计的好坏是系统成败的关键，它也占据了整个系统大部分运算量，主要包括前处理子模块，运动检测子模块，后处理子模块。前处理子模块在分析传统滤波器优缺点的基础上，引入像素加权和像素分类的思想，构造出一个改进型多级中值滤波器，不仅具有良好的滤除性能，还较好的保护了图像的细节。运动检测子模块中对传统的算法进行了性能评析，指出各种传统算法的优缺点，重点提出了动态背景减除改进算法，将时域差分法和动态背景法的优点综合起来，构造一种动态更新的背景模型，达到了较好的效果。后处理子模块采用形态学滤波来去噪、补空洞，文章对传统的形态学后处理方案进行了改进。
目标跟踪功能模块是智能化监控系统的重要环节，其中的活动摄像机
浙江大学硕士学位论文
对运动目标的跟踪尤其具有实用价值。文章首先详细分析了固定背景下的
传统匹配跟踪算法。接着重点提出了一种运动目标的运动平滑性和尺寸连
续性等特征，在二值图像层面上对运动目标区域的外接矩形进行跟踪的计
算简单的算法。该算法充分考虑了互遮挡、目标短暂消失和永久移入、移
出问题，可用于多目标的跟踪，具有较强的鲁棒性。最后简要的分析了活
动摄像机下的目标跟踪问题。
文章的最后给出了实验的部分演示序列并对其加以了说明，分析了本
论文中算法上的不足，并指出尚待改进的方向。
In the modern time of pursuing for the high automatization in the field of visual surveillance, the desire for a smart surveillance system, which is the substitute of surveillance members and can monitoring the moving object(such as human, cars and so on) in some surveillance scape automatically and consecutively, is very strong. In most cases, this intelligent system has some functions such as motion detection and object tracking etc. It's very possible to design this system which no one needed to watch by full utilizing the powerful calculating capacity of modern computer and many subjects such as image processing and analyzing, computer vision etc. But regrettably this kind of system is not appeared in our country currently. The most difficult problem is lack of a device of global algorithm. Common algorithm often cannot reach the practical standards such as tight linking among modules, low algorithm complexity, low ratio of missing alarm and false alarm etc. This paper is put forward in such a background
. The algorithm complexity is more lowered by simplifying and improving the algorithm on the two modules of motion detection and object tracking on the basis of researching and analyzing many other person's algorithms and studying others' research fruits. The global algorithm synthetically considers the real-time property and robust performance, and it can be used in surveillance system.
Firstly this paper introduces the research status and tendency in the world in detail. The whole paper is divided by two modules of motion detection and object tracking based on its function, and the article elaborates them one by one.
Motion detection module is the most important part in the intelligent system. Its device is the key of the whole system and it also holds the most calculating task. This module is divided by three sub-modules of pre-processing, motion detection and post-processing. In the sub-module of pre-processing, the paper analyzes performance of traditional filters and points out its advantages and disadvantages, then presents a modified multilevel median filter by importing the idea of weight and classification. The proposed filter can not only suppress noise more effectively but also preserve the features of images. In the sub-module of motion detection, the article first analyzes performance
of traditional algorithm and points out some disadvantages, and then brings forward one improved algorithms: the modified dynamic background deduction algorithm, which puts the advantages of temporal differencing algorithm and dynamic background deduction algorithm together to present a dynamic background model. Experimental results are given to demonstrate its validity and practicality. The post-processing sub-module uses morphological filter to remove isolated noise and to fill the holes inside objects, and the traditional method of this sub-module is improved in the paper.
Object tracking module is also an important part in the intelligent system. Among them the object tracking with an Active Camera has much more practical value. Firstly, the article introduces the traditional matching tracking algorithm detailedly in the fixed background. Secondly, a simple tracking algorithm integrated with many moving objects' characters such as smooth motion property and consecutive size property is brought forward. The simplified algorithm is just tracking the outer-rectangle of moving object area in-the binary image. Carefully considering the problems of the occlusion, temporary disappearance and being permanently moved in or moved out of objects detailedly, this algorithm can be used in the tracking of multi-object and its robust performance is also strong. Finally, the paper analyzes the object tracking with an active camera.
At the end of this paper, some demo sequences of experiment are given and elaborated. Finally the paper introduces shortcomings of proposed algorithms and aspects deserved to be strengthened.
引文
[1]艾海舟，乐秀宇，面向视觉监视实时跟踪的动态背景更新方法，计算机工程与应用，2001，37(19)：104～106．
[2]雷晓峰，王耀南，远程图象监控系统的智能跟踪开发，吉首大学学报(自然科学版)，2001，22(3)：30～33．
[3]刘晓冬，苏光大，周全等，一种可视化智能户外监控系统，中国图象图形学报，2000，5(12)：1024～1029．
[4]王栓，艾海舟，何克忠，基于差分图象的多运动目标的检测与跟踪，中国图象图形学报，1999，4(6)：470～475．
[5]章毓晋，中国图象工程：1998，中国图象图形学报，1999，4(5)：427～438．
[6]章毓晋，中国图象工程：1999，中国图象图形学报，2000，5(5)：359～373．
[7]章毓晋，中国图象工程：2000，中国图象图形学报，2001，6(5)：409～424．
[8]Thomas B.Moeslund, Summaries of 107 Computer Vision-based Human Motion Capture Papers, Technical Report LIA 99-01, University of Aalborg, 1999.
[9]王亮，胡卫明，谭铁牛，人运动的视觉分析综述，计算机学报，2002，25(3)：225～237．
[10]刘亚，艾海舟，徐光佑，一种基于背景模型的运动目标检测与跟踪算法，信息与控制，2002，31(4)：315～319，328．
[11]Collins, A System for Video Surveillance and Monitoring:VSAM Final Report, Carnegie Mellon University, Technical Report, 2000.
[12]Robert T.Collins, Alan J.Lipton, Takeo Kanade, Hironobu Fujiyoshi, David Duggins, Yanghai Tsin, David Tolliver, Nobuyoshi Enomoto, Osamu Hasegawa, Peter Burr and Lambert
Wixson, A System for Video Surveillance and Monitoring, 2000.
[13]Ismail Haritaoglu, A Real Time System for Detecting and Tracking of People and Recognizing Their Activities, Computer Vision Laboratory University of Maryland College Park, 2000.
[14]P.Remagnino, T.Tan, B.Kaker, Multi-agent Visual Surveillance of Dynamic Scenes, Image and Vision Computing, 1998,16(8): 529～532.
[15]C.Maggioni and B.Kaemmerer, Oesture Computer-History, Design, and Applications, Vision in ManMachine Interfaces, Cambridge University Press, 1996.
[16]W.T.Freeman and C.D.Weissman, Television Control by Hand Oestures, Proceedings of the International Workshop on Automatic Face and Gesture Recognition, 1995,179～183.
[17]http://www.sinosurveillance.com, National Laboratory of Pattern Recognition(NLPR),Institute of Automation Chinese Academy of Sciences.
[18]关海英，阮秋琦，基于Hausdorff距离的非刚体目标自适应轮廓跟踪，通信学报，1998，19(11)：38～43．
[19]关海英，阮秋琦，序列图象中运动目标的检测，铁道学报，1996，18(6)：32～35．
[20]刘亚，艾海舟，徐光佑，基于主运动分析的野外视觉侦察系统——运动目标检测、跟踪及全景图的生成，机器人，2001，23(3)：250～255．
[21]艾海舟，王栓，何克忠，基于差分图象的人脸检测，中国图象图形学报，1998，3(12)：987～992．
[22]范勇，游志胜，张建州等，一种快速运动目标检测与跟踪算法，光电工程，2000，27(6)：30～33．
[23]潘锦辉，廖庆敏，林行刚，视频序列中运动目标的自动提取，清华大学学报(自然科学版)，2001，41(4)：190～193．
[24]罗涛，头肩视频图象的运动物体自动提取，北京大学学报(自然科学版)，2000，36(5)：599～607．
[25]D.Murray, A.Basu, Motion Tracking with an Active Camera, IEEE Transactions on Pattern Analysis And Machine Intelligence, 1994,16(5):449～459.
[26]刘明宝，高文等，彩色图象的实时人脸跟踪方法，计算机学报，1998，21(6)：527～532．
[27]付永会，张风超，张宪民，一种改进的基于颜色直方图的实时目标跟踪算法，数据采集与处理，2001，16(3)：309～314．
[28]王延江，袁保宗，唐晓芳，基于遗传算法的人脸跟踪研究，石油大学学报(自然科学版)，2002，26(4)：92～94，100．
[29]Huazhong Ning, Liang Wang, Weiming Hu and Tieniu Tan, Tracking People Based on Human Body Model and Kinematics, National Laboratory of Pattern Recognition Institute of Automation, Chinese Academy of Sciences, Beijing, P.R. China, 100080.
[30]Jia-Ching Cheng and Jose M.F.Moura, Capture and Representation of Human Walking in Live Video Sequences, IEEE Transactions on Multimedia, 1999,1(2):144～156.
[31]王虎元，刘永昌，实时双模图象跟踪系统技术研究，中国图象图形学报，1997，2(2)：113～118．
[32]张志勇，黄爱民，刘建平，视频智能交通系统，计算机工程与应用，2001，37(6)：113～116．
[33]王春波，张卫东，张文渊等，复杂交通环境中车辆的视觉检测，上海交通大学学报，2000，34(12)．
[34]胡汉南，水运交通电视监控图象目标识别技术研究，交通部上海船舶运输科学研究所学报，2001，24(1)：3～9．
[35]赵亚男，达庆东，杨群等，智能交通安全系统的研究，中国安全科学学报，2001，11(3)：27～33．
[36]Sun T, Neuvo Y, Detail-preserving Median Based Filters in Image
Processing, Pattern Recognition, 1994,15:341～347.
[37]Wang Zhou, Zhang David, Progressive Switching Median Filter for the Removal of Impulse Noise From Highly Corrupted Images, IEEE Transactions on Circuits and Systems Ⅱ: Analog and Digital Signal Porcessing, 1999,46(1):78～80.
[38]Wang Jung-Hua, Lin Lian-Da, Improved Median Filter Using Minmax Algorithm for Image Processing, Electronics Letters, 1997, 33 (16): 134～143.
[39]Brownrigg DRK, The Weighted Median Filter, Commun. Ass. Comput, Mach, 1984,27(8):807～818.
[40]张嫣，姚耀文，唐华松，运动图象序列中多目标跟踪的研究与实现，计算机应用研究，2002，19(1)：74～76．
[41]李强，杨涛，形心和匹配跟踪算法的改进，光电工程，2000，27(2)：28～30．
[42]胡俊，苏祥芳，刘立海 等，图象序列运动检测算法的研究及其应用，武汉大学学报(自然科学版)，2000，46(5)：613～616．
[43]张桂林，张天序，基于边缘特征的运动目标的提取与跟踪，华中理工大学学报，1994，22(5)：42～45．
[44]庄涛，王人成，一种运动图象标志点识别跟踪方法的研究，系统工程与电子技术，2002，24(4)：89～92．
[45]刘小军，江林华，一种基于投影算法的运动目标自动跟踪系统，华北工学院测试技术学报，1999，13(4)80～82．
[46]刘珂，张宪民，付永会，一种改进的Hausdorff距离目标跟踪算法，上海交通大学学报，2001，35(2)：223～227．
[47]杨杨，张田文，基于多分辨率方法的主动轮廓线路跟踪算法，计算机学报，1998，21(3)：210～216．
[48]Thomas Meier, King N.Ngan, Automatic Segmentation of Moving Objects for Video Object Plane Generation, IEEE Transactions on Circuits and Systems for Video Technology, 1998,8(5):525～
538．
[49]史立，张兆扬，马然，基于运动跟踪匹配技术的视频对象提取，通信学报，2001，22(11)：77～85．
[50]汪亚明，基于动态图象序列的自动扶梯客流量的测量，华北工学院测试技术学报，2002，16(1)：14～16．
[51]徐建华，图象处理与分析，1992年，科学出版社．
[52]汪亚明，楼正国，卞昕 等，一种非刚体运动图象序列的特征点对应方法，中国图象图形学报，2000，5(3)：232～236．
[53]王春平，朱元昌，黄允华，基于图象信息的跟踪算法分析，火力与指挥控制，2000，25(1)：12～15．
[54]Chia-Wen Lin, Yao-Ten Chang, Chih-Ming Wang, etc, A Standard Compliant Virtual Meeting System With Active Video Object Tacking, Applied Signal Processing, 2002.
[55]艾海舟，吕风军，刘伟等，面向视觉监视的运动检测与分割，计算机工程与应用，2001，37(5)：75～77．
[56]章毓晋等，图象分割，2000，科学出版社．
[57]章毓晋，图象工程(上册)：图象处理和分析，1999年，清华大学出版社．
[58]周兵，姜东，毕波，摄像机运动下的快速运动对象检测，第一届全国智能视觉监控学术会议论文，2002年5月．
[59]章毓晋，图象工程(下册)：图象理解与计算机视觉，2000年，清华大学出版社．
[60]李海明，陈新，吴芳等，复杂背景下运动目标的光流区域提取方法，福州大学学报(自然科学版)，2001，29(4)：101～103，106．
[61]杨杨，张田文，一种基于特征光流的运动目标跟踪方法，宇航学报，2000，21(2)：8～15．
[62]Barron J, Fleet D, BeaucheminS, Performance of Optical Flow Techniques, International Journal of Computer Vision, 1994, 12(1):42～77.
[63]杨威，张田文，复杂景物环境下运动目标检测的新方法，计算机研究与发展，1998，35(8)：724～728．
[64]C.Stauffer, W.Grimson, Adaptive Background Mixture Models for Real-time Tracking, Computer Vision and Pattern Recognition, 1999,2(2):246～252.
[65]McKenna Setal, Tracking Groups of People, Computer Vision and Image Understanding, 2000,80(1):42～56.
[66]K.Karmann, A.Brandt, Moving Object Recognition Using an Adaptive Background Memory, Time-varying Image Processing and Moving Object Recognition,2, Elsevier,Amsterdam, The Netherlands, 1990.
[67]Aach Til, A.Kaup, R.Mester, Statistical Model-based Change Detection in Moving Video, [J].Signal Processing, 1993, 31:165～180.
[68]章毓晋，数字图象处理，2002年，清华大学出版社．
[69]阮秋琦，数字图象处理学，2001年，电子工业出版社．
[70]张桂林等，基于跑长码的连通区域标记算法，华中理工大学学报，1994，22(5)：11～14．
[71]杨威，张田文，师海峰，一种用于二值图象分割的快速聚类算法，计算机研究与发展，1998，35(8)：719～723．
[72]吴健康，数字图象分析，1987年，人民邮电出版社．
[73]周志宇，汪亚明，黄文清，基于动态图象序列的运动目标跟踪，浙江工程学院学报，2002，19(3)：165～170．
[74]D.P.Huttenlocher, G.A.Klanderman, W.J.Rucklidge, Comparing Images Using the Hausdorff Distance, IEEE Transactions on Pattern Analysis and Machine Intelligence, 1993,15:850～863.
[75]D.P.Huttenlocher, J.J.Noh and W.J.Rucklidge, Tracking Nonrigid Objects in Complex Scenes, The 4th International Conference on Computer Vision, Berlin, Germany, 1993,93～101.
[76]Jianguang Lou, Hao Yang, Weiming Hu etc, An Illumination Invariant Change Detection Algorithm, The 5th Asian Conference on Computer Vision, 2002,23～25.
[77]I.Haritaoglu, D. Harwood and L.Davis, W4:Real-time Surveillance of People and Their Activities, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000,22(8):809～830.
[78]C.Stauffer and W.E.L.Grimson, idaptive Background Mixture Models for Real-time Tracking, Computer Vision and Pattern Recognition, 1999,2:246～252.
[79]G.L.Foresti, A Real-Time System for Video Surveillance of Unattended Outdoor Environments, IEEE Transactions on Circuits and Systems for Video Technology, 1998,8(6):697～704.
[80]M.Kilger, A Shadow Handler in a Video-based Real-time Traffic Monitoring System, IEEE Workshop on Applications of Computer Vision, 1992,11～18.
[81]刘小明，庄越挺，潘云鹤，基于模型的人体运动跟踪，计算机研究与发展，1999，36(10)：1268～1273．
[82]李梦东，阮秋琦，利用三维网格模型的头部运动估计，铁道学报，2002，24(2)：117～120．
[83]孙怡，王恩亮，安宏杰，胡家升，基于2-D模型的人体运动跟踪，中国图象图形学报，2002，7(7)：625～632．
[2]雷晓峰，王耀南，远程图象监控系统的智能跟踪开发，吉首大学学报(自然科学版)，2001，22(3)：30～33．
[3]刘晓冬，苏光大，周全等，一种可视化智能户外监控系统，中国图象图形学报，2000，5(12)：1024～1029．
[4]王栓，艾海舟，何克忠，基于差分图象的多运动目标的检测与跟踪，中国图象图形学报，1999，4(6)：470～475．
[5]章毓晋，中国图象工程：1998，中国图象图形学报，1999，4(5)：427～438．
[6]章毓晋，中国图象工程：1999，中国图象图形学报，2000，5(5)：359～373．
[7]章毓晋，中国图象工程：2000，中国图象图形学报，2001，6(5)：409～424．
[8]Thomas B.Moeslund, Summaries of 107 Computer Vision-based Human Motion Capture Papers, Technical Report LIA 99-01, University of Aalborg, 1999.
[9]王亮，胡卫明，谭铁牛，人运动的视觉分析综述，计算机学报，2002，25(3)：225～237．
[10]刘亚，艾海舟，徐光佑，一种基于背景模型的运动目标检测与跟踪算法，信息与控制，2002，31(4)：315～319，328．
[11]Collins, A System for Video Surveillance and Monitoring:VSAM Final Report, Carnegie Mellon University, Technical Report, 2000.
[12]Robert T.Collins, Alan J.Lipton, Takeo Kanade, Hironobu Fujiyoshi, David Duggins, Yanghai Tsin, David Tolliver, Nobuyoshi Enomoto, Osamu Hasegawa, Peter Burr and Lambert
Wixson, A System for Video Surveillance and Monitoring, 2000.
[13]Ismail Haritaoglu, A Real Time System for Detecting and Tracking of People and Recognizing Their Activities, Computer Vision Laboratory University of Maryland College Park, 2000.
[14]P.Remagnino, T.Tan, B.Kaker, Multi-agent Visual Surveillance of Dynamic Scenes, Image and Vision Computing, 1998,16(8): 529～532.
[15]C.Maggioni and B.Kaemmerer, Oesture Computer-History, Design, and Applications, Vision in ManMachine Interfaces, Cambridge University Press, 1996.
[16]W.T.Freeman and C.D.Weissman, Television Control by Hand Oestures, Proceedings of the International Workshop on Automatic Face and Gesture Recognition, 1995,179～183.
[17]http://www.sinosurveillance.com, National Laboratory of Pattern Recognition(NLPR),Institute of Automation Chinese Academy of Sciences.
[18]关海英，阮秋琦，基于Hausdorff距离的非刚体目标自适应轮廓跟踪，通信学报，1998，19(11)：38～43．
[19]关海英，阮秋琦，序列图象中运动目标的检测，铁道学报，1996，18(6)：32～35．
[20]刘亚，艾海舟，徐光佑，基于主运动分析的野外视觉侦察系统——运动目标检测、跟踪及全景图的生成，机器人，2001，23(3)：250～255．
[21]艾海舟，王栓，何克忠，基于差分图象的人脸检测，中国图象图形学报，1998，3(12)：987～992．
[22]范勇，游志胜，张建州等，一种快速运动目标检测与跟踪算法，光电工程，2000，27(6)：30～33．
[23]潘锦辉，廖庆敏，林行刚，视频序列中运动目标的自动提取，清华大学学报(自然科学版)，2001，41(4)：190～193．
[24]罗涛，头肩视频图象的运动物体自动提取，北京大学学报(自然科学版)，2000，36(5)：599～607．
[25]D.Murray, A.Basu, Motion Tracking with an Active Camera, IEEE Transactions on Pattern Analysis And Machine Intelligence, 1994,16(5):449～459.
[26]刘明宝，高文等，彩色图象的实时人脸跟踪方法，计算机学报，1998，21(6)：527～532．
[27]付永会，张风超，张宪民，一种改进的基于颜色直方图的实时目标跟踪算法，数据采集与处理，2001，16(3)：309～314．
[28]王延江，袁保宗，唐晓芳，基于遗传算法的人脸跟踪研究，石油大学学报(自然科学版)，2002，26(4)：92～94，100．
[29]Huazhong Ning, Liang Wang, Weiming Hu and Tieniu Tan, Tracking People Based on Human Body Model and Kinematics, National Laboratory of Pattern Recognition Institute of Automation, Chinese Academy of Sciences, Beijing, P.R. China, 100080.
[30]Jia-Ching Cheng and Jose M.F.Moura, Capture and Representation of Human Walking in Live Video Sequences, IEEE Transactions on Multimedia, 1999,1(2):144～156.
[31]王虎元，刘永昌，实时双模图象跟踪系统技术研究，中国图象图形学报，1997，2(2)：113～118．
[32]张志勇，黄爱民，刘建平，视频智能交通系统，计算机工程与应用，2001，37(6)：113～116．
[33]王春波，张卫东，张文渊等，复杂交通环境中车辆的视觉检测，上海交通大学学报，2000，34(12)．
[34]胡汉南，水运交通电视监控图象目标识别技术研究，交通部上海船舶运输科学研究所学报，2001，24(1)：3～9．
[35]赵亚男，达庆东，杨群等，智能交通安全系统的研究，中国安全科学学报，2001，11(3)：27～33．
[36]Sun T, Neuvo Y, Detail-preserving Median Based Filters in Image
Processing, Pattern Recognition, 1994,15:341～347.
[37]Wang Zhou, Zhang David, Progressive Switching Median Filter for the Removal of Impulse Noise From Highly Corrupted Images, IEEE Transactions on Circuits and Systems Ⅱ: Analog and Digital Signal Porcessing, 1999,46(1):78～80.
[38]Wang Jung-Hua, Lin Lian-Da, Improved Median Filter Using Minmax Algorithm for Image Processing, Electronics Letters, 1997, 33 (16): 134～143.
[39]Brownrigg DRK, The Weighted Median Filter, Commun. Ass. Comput, Mach, 1984,27(8):807～818.
[40]张嫣，姚耀文，唐华松，运动图象序列中多目标跟踪的研究与实现，计算机应用研究，2002，19(1)：74～76．
[41]李强，杨涛，形心和匹配跟踪算法的改进，光电工程，2000，27(2)：28～30．
[42]胡俊，苏祥芳，刘立海 等，图象序列运动检测算法的研究及其应用，武汉大学学报(自然科学版)，2000，46(5)：613～616．
[43]张桂林，张天序，基于边缘特征的运动目标的提取与跟踪，华中理工大学学报，1994，22(5)：42～45．
[44]庄涛，王人成，一种运动图象标志点识别跟踪方法的研究，系统工程与电子技术，2002，24(4)：89～92．
[45]刘小军，江林华，一种基于投影算法的运动目标自动跟踪系统，华北工学院测试技术学报，1999，13(4)80～82．
[46]刘珂，张宪民，付永会，一种改进的Hausdorff距离目标跟踪算法，上海交通大学学报，2001，35(2)：223～227．
[47]杨杨，张田文，基于多分辨率方法的主动轮廓线路跟踪算法，计算机学报，1998，21(3)：210～216．
[48]Thomas Meier, King N.Ngan, Automatic Segmentation of Moving Objects for Video Object Plane Generation, IEEE Transactions on Circuits and Systems for Video Technology, 1998,8(5):525～
538．
[49]史立，张兆扬，马然，基于运动跟踪匹配技术的视频对象提取，通信学报，2001，22(11)：77～85．
[50]汪亚明，基于动态图象序列的自动扶梯客流量的测量，华北工学院测试技术学报，2002，16(1)：14～16．
[51]徐建华，图象处理与分析，1992年，科学出版社．
[52]汪亚明，楼正国，卞昕 等，一种非刚体运动图象序列的特征点对应方法，中国图象图形学报，2000，5(3)：232～236．
[53]王春平，朱元昌，黄允华，基于图象信息的跟踪算法分析，火力与指挥控制，2000，25(1)：12～15．
[54]Chia-Wen Lin, Yao-Ten Chang, Chih-Ming Wang, etc, A Standard Compliant Virtual Meeting System With Active Video Object Tacking, Applied Signal Processing, 2002.
[55]艾海舟，吕风军，刘伟等，面向视觉监视的运动检测与分割，计算机工程与应用，2001，37(5)：75～77．
[56]章毓晋等，图象分割，2000，科学出版社．
[57]章毓晋，图象工程(上册)：图象处理和分析，1999年，清华大学出版社．
[58]周兵，姜东，毕波，摄像机运动下的快速运动对象检测，第一届全国智能视觉监控学术会议论文，2002年5月．
[59]章毓晋，图象工程(下册)：图象理解与计算机视觉，2000年，清华大学出版社．
[60]李海明，陈新，吴芳等，复杂背景下运动目标的光流区域提取方法，福州大学学报(自然科学版)，2001，29(4)：101～103，106．
[61]杨杨，张田文，一种基于特征光流的运动目标跟踪方法，宇航学报，2000，21(2)：8～15．
[62]Barron J, Fleet D, BeaucheminS, Performance of Optical Flow Techniques, International Journal of Computer Vision, 1994, 12(1):42～77.
[63]杨威，张田文，复杂景物环境下运动目标检测的新方法，计算机研究与发展，1998，35(8)：724～728．
[64]C.Stauffer, W.Grimson, Adaptive Background Mixture Models for Real-time Tracking, Computer Vision and Pattern Recognition, 1999,2(2):246～252.
[65]McKenna Setal, Tracking Groups of People, Computer Vision and Image Understanding, 2000,80(1):42～56.
[66]K.Karmann, A.Brandt, Moving Object Recognition Using an Adaptive Background Memory, Time-varying Image Processing and Moving Object Recognition,2, Elsevier,Amsterdam, The Netherlands, 1990.
[67]Aach Til, A.Kaup, R.Mester, Statistical Model-based Change Detection in Moving Video, [J].Signal Processing, 1993, 31:165～180.
[68]章毓晋，数字图象处理，2002年，清华大学出版社．
[69]阮秋琦，数字图象处理学，2001年，电子工业出版社．
[70]张桂林等，基于跑长码的连通区域标记算法，华中理工大学学报，1994，22(5)：11～14．
[71]杨威，张田文，师海峰，一种用于二值图象分割的快速聚类算法，计算机研究与发展，1998，35(8)：719～723．
[72]吴健康，数字图象分析，1987年，人民邮电出版社．
[73]周志宇，汪亚明，黄文清，基于动态图象序列的运动目标跟踪，浙江工程学院学报，2002，19(3)：165～170．
[74]D.P.Huttenlocher, G.A.Klanderman, W.J.Rucklidge, Comparing Images Using the Hausdorff Distance, IEEE Transactions on Pattern Analysis and Machine Intelligence, 1993,15:850～863.
[75]D.P.Huttenlocher, J.J.Noh and W.J.Rucklidge, Tracking Nonrigid Objects in Complex Scenes, The 4th International Conference on Computer Vision, Berlin, Germany, 1993,93～101.
[76]Jianguang Lou, Hao Yang, Weiming Hu etc, An Illumination Invariant Change Detection Algorithm, The 5th Asian Conference on Computer Vision, 2002,23～25.
[77]I.Haritaoglu, D. Harwood and L.Davis, W4:Real-time Surveillance of People and Their Activities, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000,22(8):809～830.
[78]C.Stauffer and W.E.L.Grimson, idaptive Background Mixture Models for Real-time Tracking, Computer Vision and Pattern Recognition, 1999,2:246～252.
[79]G.L.Foresti, A Real-Time System for Video Surveillance of Unattended Outdoor Environments, IEEE Transactions on Circuits and Systems for Video Technology, 1998,8(6):697～704.
[80]M.Kilger, A Shadow Handler in a Video-based Real-time Traffic Monitoring System, IEEE Workshop on Applications of Computer Vision, 1992,11～18.
[81]刘小明，庄越挺，潘云鹤，基于模型的人体运动跟踪，计算机研究与发展，1999，36(10)：1268～1273．
[82]李梦东，阮秋琦，利用三维网格模型的头部运动估计，铁道学报，2002，24(2)：117～120．
[83]孙怡，王恩亮，安宏杰，胡家升，基于2-D模型的人体运动跟踪，中国图象图形学报，2002，7(7)：625～632．