摘要
移动目标视频跟踪是当前信息领域的前沿和热点方向,融合了计算机科学、自动控制、机器视觉、图像处理、模式识别、数学等多学科的先进技术。本文以智能视频监控作为主要线索,研究静止背静下运动目标的检测、跟踪等视频跟踪中的关键技术。
计算机智能视频监控是在不需要人为干预情况下,利用计算机视觉和视频分析的方法对摄像机拍录的图像序列进行自动分析,实现对动态场景中目标的定位、识别和跟踪,并在此基础上分析和判断目标的行为,得出对图像内容含义的理解以及对客观场景的解释,从而指导和规划行动。这种智能视频监控己在军事和工业上得到一些成功运用,但智能视频监控在理论和运用上都还存在很多难题。当前国内外很多学者投身该领域进行研究和探索,并取得了大量成果,本文是在这些成果的基础上进行的。
首先,本文系统地研究和总结了国内外运动目标检测的方法,分析了各方法的利弊、实用场合,在此基础上,重点研究了混合高斯模型,并将极大似然原理引入混合高斯模型,该方法较以往的基于经验值的混合高斯相比,有严密的数学理论作支撑,并且在此方法下的检测效果也能满足固定场合的要求。
其次,本文对图像增强、图像去噪和图像分割也作了较系统的总结,并给出了部分实验结果;在此基础上总结出一种图像点、线增强的一般方法。
最后,在跟踪方面使用经典的运动分析理论—卡尔曼滤波,对运动目标的下帧所在区域作估计,进而缩小搜索空间,提高检测和跟踪速度。
Video tracking of moving object is a hot and latest direction in today's information technology field. It refers to lots of subjects, including computer science, automation control, computer vision, image processing, pattern recognition and math. The thesis put the intelligent surveillance system as the thread, focused on moving object detection and moving object tracking in stable background.
The intelligent surveillance system can automatically analyze the sequence of images by the methods of computer vision and video analysis. The system can position, recognize and track objects in a moving environment. Furthermore, it can also analyze and judge the movement of objects. The aims of this system are to understand the meanings of video stream and to explain the scenes comprehensibly, hence to guide actions and make decisions. Some of the intelligent surveillance systems have successfully been applied to army and industry, but the project still faces lots of challenges. A large number of researchers have been studying the subject and achieved many progresses. This paper was based on these achievements.
Firstly, the different kinds of detection methods with their advantages and disadvantages were concluded. We mainly researched the Gauss Mixture Model, and then put the maximizing likelihood idea into the Gauss Mixture Model. The new method was more rigorous, and the detection results can meet the requirements.
Secondly, the methods about image enhancement, image filter and image segmentation were concluded, and then some experiment results were given. We extended the image enhancement method, which can used to enhance some dots and threads.
Finally, we used the Kalman filter to forecast the position of the moving object in the next frame. The method can reduce the search space in order to improve the efficiency.
引文
[1]www.esmchina.com/ART_8800080968_1100_2704_0_0_c09c2c0e.HTM(国际电子商情网)
[2]王长军.基于视频的目标检测与跟踪技术研究:[博士学位论文].杭州:浙江大学,2006,15-15
[3]赵建云,郑晓势,周伟,等.基于视频的车辆检测与跟踪技术综述.计算机与信息技术,2007,9(10):30-33
[4]杨建全,梁华,王成友.视频监控技术的发展与现状.现代电子技术,2006,21:84-88
[5]王素玉,沈兰荪.智能视觉监控技术研究进展.中国图像图形学报,2007,12(9):1507-1512
[6]牛少平.计算机视频监控技术研究:[硕士学位论文].西安:西北工业大学,2006,6-10
[7]Kenneth R.Castleman.数字图像处理,朱志刚译.北京:电子工业出版社,2002,204-206
[8]William K.Part..数字图像处理,邓鲁华译.北京:机械工业出版社,2005,269-295
[9]陆宗骐.C/C++图像处理编程.北京:清华大学出版社,2005,82-83
[10]Iain E.G Richardson.H.264 and MPEG-4 Video Compression.USA:WILLEY,2003,17-27
[11]C Stauffer,WEL Grimson.Learning Patterns of Activity Using Real-Time Tracking.IEEE Trans.PAMI,2000,22(8):747-757
[12]I.Haritaoglu,D Harwood,L S Davis.W4:Rea-Time Surveillance of People and Their Activities.IEEE Trans.PAMI,2000,22(8):809-830
[13]A.Elgammal,R.Duraiswami,D.Harwood,et al.Background and foreground modeling using nonparametric kernel density estimation for visual surveillance.Proc.Of the IEEE,2002,vol.90:1151-1163
[14]P.Msssimo,Background subtraction techniques:a review.2004 IEEE International Conference on Systems,Man and Cybernetics,2004,3099-3104
[15]张玲,叶海炳,何伟.一种基于边缘信息的改进车辆检测方法.重庆大学学报(自然科学版),2000,27(11):56-58
[16]王建中,姜礼明.基于块匹配的运动对像检测方法.微电子信息,2005,22(1):204-204
[17]朱江,宣国容.基于视频动态投影的实时车辆流量检测系统.计算机工程,2001,27(11):25-26
[18]V Caselles,R Kimmel,G.Sapiro.Geodesic active contours.International Journal of Computer Vision,1997,22(1):61-79
[19]Ma S D.A self-calibration technique for active vision system.IEEE TransRobot Automation,1996,12(1):114-120
[20]A.Canny.A computational approach to edge detection.IEEE Transactions on Pattern Analysis and Machine Intelligence,1986,vol.8:679-698
[21]皮文凯,刘宏,查红彬.基于自适应背景模型的全方位视觉人体运动检测.北京大学学报(自然科学版),2004,40(3):458-464
[22]于长官.现代控制理论.哈尔滨:哈尔滨工业大学出版社,1997,163-172
[23]侯志强,韩崇昭.基于像素灰度归类的背景重构算法.软件学报,2005,16(9):1568-1576
[24]刘富强.数字视频监控系统开发及应用.北京:机械工业出版社,2003,88-94
[25]Remagnino P,Tan T,Baker K.Multi-agent visual surveillance of dynamic scenes.Image and Vision Computing,1998,16(8):529-532
[26]Verri A,Uras S,DeMicheli E.Motion segmentation from optical flow.Proc the 5th Alvey Vision Conference,Brighton,UK,1989,209-214
[27]Barron J,Fleet D,Beauchemin S.Performance of optical flow techniques.International journal of computer vision,1994:12(1):42-77
[28]MaeY,Shirai Y,Miura J,Kuno Y.Object tracking in cluttered background based on optical flow and edges.Proc the 13th International Conference on Pattern Recognition,1996,196-200
[29]Stocker A.A.An improved 2D optical flow sensor for motion segmentation.IEEE International Symposium on Circuits and Systems,2002,vol.2:26-29
[30]Kass M,Witkin A,Terzopoulous D.Snakes:active contour models.International Journal of Computer Vision,1987,1(4):321-331
[31]宋新,罗军,王鲁平.基于GVF Snake的运动目标跟踪方法.红外与激光工程,2007,36(2),226-228
[32]C E Daniell,D J Kemsley,W P Lincoln,et al.Artificial neural networks for automatic target recognition.Optical Engineering,1992,31(12):2521-2530
[33]S K Rogers,J M Colombi,C E Martin,et al.Neural networks for automatic target recognition.Neural Networks,Optical Engineering,1995,18(7):1153-1184
[34]W.A.Thoet,T.G.Rainey,D WBrettle,et al.Neural network program for three-dimensional automatic target recognition.Optical Engineering,1992,31(12):2532-2539
[35]王超,侯丽敏.一种新的高斯混合模型参数估计算法.上海大学学报,2005,11(5):475-479
[36]林洪文,涂丹,李国辉.基于统计背景模型的运动目标检测方法.计算机工程,2003,29(16):97-99
[37]王陈阳,周明全.基于自适应背景模型运动目标检测.计算机技术与发展,2007,17(4):21-26
[38]罗盈盈,陈川,毛云芳.基于传感器网络的K-均值聚类算法研究.计算机工程与设计,2007,28(6):1349-1351
[39]岳佳,王士同.高斯混合模型聚类中EM算法及初试化的研究.软件时空,2006,22(11):244-248
[40]P.S.BRADLEY,O.L.MANGASARIAN.k-Plane Clustering.Journal of Global Optimization,2000,vol.16:23-32
[41]郑南宁.计算机视觉与模式识别.北京:国防工业出版社,1998,66-67
[42]Highnam R.,Brady M.Model-based image enhancement of far infrared images.IEEE Transactions on Pattern Analysis and Machine Intelligence,1997,19(4):410-415
[43]Arad N.,Gotsman C.Enhancement by image-dependent warping.IEEE Transactions on Image Processing,1999,8(8):1063-1074
[44]林晓春,李存志.一种基于图像融合的红外图像增强的新方法.西安电子科技大学学报,2005,32(2):189-192
[45]万琴,王耀南.基于卡尔曼滤波器的运动目标检测与跟踪.湖南大学学报(自然科学版),2007,34(3):36-40
[46]王纪,叶宇程.卡尔曼滤波在视觉伺服机器人控制中的运用.河南大学学报(自然科学版),2007,28(3):35-38
[47]S.Gillijns,O.Barrero Mendoza.What is the Ensemble Kalman Filter and How Well Does it work.Proceeding of the 2006 American Control Conference,USA,2006:14-16
[48]Ribaric S.G,Adrinek,Segvic S.Real-time active visual tracking system.IEEE Proceedings Mediterranean Electrotechnical Conference,2004,vol.1:231-234
[49]陆光华,彭学愚,张林让等.随机信号处理.西安:西安电子科技大学出版社,2002:48-57.
[50]沈晔青,龚华军.自适应卡尔曼滤波在目标跟踪系统中的运用.计算机仿真,2007,24(11):210-213
[51]张贝贝,肖国强,江健民.基于区域行心及其矢量方向的视频对象跟踪.计算机运用,2007,27(6):1502-1505
[52]Ribaric S.G,Adrinek,Segvic S.Real-time active visual tracking system.IEEE Proceedings Mediterranean Electrotechnical Conference.2004,vol.1:231-234
[53]S.Umesh Kumar,C.S.kumar.Sensor Networks for Tracking a Moving Object using Kalman Filering.IEEE International Conference on Industrial Technology,2006:1077-1082
[54]A.Murat Tekalp.Digital video processing.Signal:PRENTICE HALL SIGNAL PROCESSING SERIES,1995:518-521