均值偏移算法在目标跟踪中的应用研究
|
摘要
目标跟踪是指在视频序列中对感兴趣的目标进行有效的跟踪,一直都是计算机视觉领域中的一个典型问题。它在视频监控、智能交通、人机交互以及国防建设等诸多领域都有广泛的应用。在诸多的目标跟踪方法中,均值偏移(Mean Shift)算法是一种基于模板匹配的目标跟踪算法,它的核心思想是将特征的核密度估计漂移到局部极大值点,由于其计算简单、实时性好,对目标变形、部分遮挡具有一定鲁棒性,近年来成为目标跟踪领域的研究热点。本文主要的研究工作包括以下几个方面。
①对目标跟踪技术及均值偏移算法的研究现状进行了分析总结。根据对跟踪问题的处理思路不同,将现有的目标跟踪方法分为基于数据驱动的方法和基于模型驱动的方法,并分析了各类算法的优缺点,为今后目标跟踪技术的研究提供了一定的理论支持。
②深入研究了均值偏移算法及其在目标跟踪中的应用,对均值偏移目标跟踪算法的性能做了理论分析和实验验证。在跟踪过程中,通过手动方式在视频序列的第一帧中确定目标区域,然后在RGB颜色特征空间中建立目标模型及候选模型,利用Bhattacharyya系数衡量其相似度。以相似性最大为原则,均值偏移算法在后续帧中迭代地搜索出目标的真实位置。最后,选取几个有代表性的视频序列做为实验数据进行了仿真实验。在此基础上对算法的计算复杂度和优缺点进行了分析、总结,并指出了算法未来的主要研究方向。
③对均值偏移算法中的窗口带宽选择策略进行了深入研究。在均值偏移目标跟踪算法中,带宽的取值直接关系到参与计算的像素数量和背景噪声混入的程度,而大部分研究人员往往直接按照惯例采用传统的带宽选择策略,很少有人对此问题进行探讨。针对这个问题,本文通过实验分析了不同带宽选择策略对跟踪算法的影响,利用跟踪准确度和时间代价两个性能指标对其进行比较,说明了传统的基于半对角线的带宽选择策略在实际跟踪过程中效果较好。
④提出了一种基于相似椭圆的双带宽选择策略,并将其纳入到传统均值偏移目标跟踪算法中。通过构造一个与目标区域内接椭圆的相似椭圆,利用其长轴和短轴确定核函数带宽。该方法提高了窗宽中所包含的目标像素点比例,有效地减小了背景噪声对跟踪算法的影响。仿真实验结果表明,使用本文带宽选择策略的均值偏移跟踪算法,提高了跟踪算法的精度,有效地降低了算法的时间代价。
The efficient tracking of moving objects in video sequence is a typical issue in computer vision fields and is widely applied to visual surveillance, intelligent traffic,human-computer ineraction,and national defence.etc. Among many object tracking methods, Mean Shift is an efficient pattern matching algorithm,which move kernel density estimation of feature to local maximum. Because of simple Computation, very good real-time, robust for object distortion and partial occlusion, Mean Shift had become a hot topic of object tracking field. The main research tasks in this paper are described as follows.
①This paper does systematical analysis and summary on Target Tracking Techniques and the status quo of Mean Shift. According on different treatment consideration of Target tracking , existing Target tracking methods are divided into two kinds:methods based on Data-driven and methods based on Model-driven. Meanwhle, the advantage and disadvantage of each method are analyzed which provide some theory support for future study.
②The author has deeply researched the Mean Shift object tracking algorithm and its application in Target tracking. And, the performance of Mean Shift is analyzed and verigy. In the tracking process,the tartet area are usually determined in the first frame by user. Then object model and candidate model are established in RGB color space, and Bhattacharyya coefficient is used to measure similar. In the subsequent frames,the true place are searched iteratively by mean shft algorithm based on maximum similarity function. In the end, the author selected severral video frequency to carry on the simulation experiment. Based on the experimental results, the author has carried on the deeply analysis on condition on the computational complexity and advantages and disadvantages of algorithm. What's more important is the future research direction are also summarized and discussed.
③Band strategies of Mean Shift tracking algorithm are made a deep research. In Mean Shift tracking algorithm, band Value is directly related to the quantity of object Pixels which participation calculation and mixed level of backgrouns noise. But, there are few people Investigate this problem and most researchers usually select traditional band strategy according to the customs directly. To solve this problem, this paper give a analysis of the influence of different band strategies though experimental. By comparing these strategies on track accuracy and time cost, the traditional band strategy based on half diagonal has a better effect in practical application.
④A new double band strategy based on similar ellipse is put forward,and this band strategy is integrated into mean shift tracking algorithm.Through structuring an ellipse being similar with target window’s inscribes ellipse, it use major axis and minor axis to identify band. This way improve Proportion of object pixel in the window,and decrease the influence of background noise effectively. Experimental Results show that this strategy has a good effect on tracking target and reducing the time cost.
①对目标跟踪技术及均值偏移算法的研究现状进行了分析总结。根据对跟踪问题的处理思路不同,将现有的目标跟踪方法分为基于数据驱动的方法和基于模型驱动的方法,并分析了各类算法的优缺点,为今后目标跟踪技术的研究提供了一定的理论支持。
②深入研究了均值偏移算法及其在目标跟踪中的应用,对均值偏移目标跟踪算法的性能做了理论分析和实验验证。在跟踪过程中,通过手动方式在视频序列的第一帧中确定目标区域,然后在RGB颜色特征空间中建立目标模型及候选模型,利用Bhattacharyya系数衡量其相似度。以相似性最大为原则,均值偏移算法在后续帧中迭代地搜索出目标的真实位置。最后,选取几个有代表性的视频序列做为实验数据进行了仿真实验。在此基础上对算法的计算复杂度和优缺点进行了分析、总结,并指出了算法未来的主要研究方向。
③对均值偏移算法中的窗口带宽选择策略进行了深入研究。在均值偏移目标跟踪算法中,带宽的取值直接关系到参与计算的像素数量和背景噪声混入的程度,而大部分研究人员往往直接按照惯例采用传统的带宽选择策略,很少有人对此问题进行探讨。针对这个问题,本文通过实验分析了不同带宽选择策略对跟踪算法的影响,利用跟踪准确度和时间代价两个性能指标对其进行比较,说明了传统的基于半对角线的带宽选择策略在实际跟踪过程中效果较好。
④提出了一种基于相似椭圆的双带宽选择策略,并将其纳入到传统均值偏移目标跟踪算法中。通过构造一个与目标区域内接椭圆的相似椭圆,利用其长轴和短轴确定核函数带宽。该方法提高了窗宽中所包含的目标像素点比例,有效地减小了背景噪声对跟踪算法的影响。仿真实验结果表明,使用本文带宽选择策略的均值偏移跟踪算法,提高了跟踪算法的精度,有效地降低了算法的时间代价。
The efficient tracking of moving objects in video sequence is a typical issue in computer vision fields and is widely applied to visual surveillance, intelligent traffic,human-computer ineraction,and national defence.etc. Among many object tracking methods, Mean Shift is an efficient pattern matching algorithm,which move kernel density estimation of feature to local maximum. Because of simple Computation, very good real-time, robust for object distortion and partial occlusion, Mean Shift had become a hot topic of object tracking field. The main research tasks in this paper are described as follows.
①This paper does systematical analysis and summary on Target Tracking Techniques and the status quo of Mean Shift. According on different treatment consideration of Target tracking , existing Target tracking methods are divided into two kinds:methods based on Data-driven and methods based on Model-driven. Meanwhle, the advantage and disadvantage of each method are analyzed which provide some theory support for future study.
②The author has deeply researched the Mean Shift object tracking algorithm and its application in Target tracking. And, the performance of Mean Shift is analyzed and verigy. In the tracking process,the tartet area are usually determined in the first frame by user. Then object model and candidate model are established in RGB color space, and Bhattacharyya coefficient is used to measure similar. In the subsequent frames,the true place are searched iteratively by mean shft algorithm based on maximum similarity function. In the end, the author selected severral video frequency to carry on the simulation experiment. Based on the experimental results, the author has carried on the deeply analysis on condition on the computational complexity and advantages and disadvantages of algorithm. What's more important is the future research direction are also summarized and discussed.
③Band strategies of Mean Shift tracking algorithm are made a deep research. In Mean Shift tracking algorithm, band Value is directly related to the quantity of object Pixels which participation calculation and mixed level of backgrouns noise. But, there are few people Investigate this problem and most researchers usually select traditional band strategy according to the customs directly. To solve this problem, this paper give a analysis of the influence of different band strategies though experimental. By comparing these strategies on track accuracy and time cost, the traditional band strategy based on half diagonal has a better effect in practical application.
④A new double band strategy based on similar ellipse is put forward,and this band strategy is integrated into mean shift tracking algorithm.Through structuring an ellipse being similar with target window’s inscribes ellipse, it use major axis and minor axis to identify band. This way improve Proportion of object pixel in the window,and decrease the influence of background noise effectively. Experimental Results show that this strategy has a good effect on tracking target and reducing the time cost.
引文
[1]高文,陈熙霖.计算机视觉—算法与系统原理[M].北京:清华大学出版社,1998.
[2] Haritaoglu I, Harwood D, Davis L. W4: Real-time surveillance of people and their activities.[J]IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(8): 809-830.
[3] Pavlidis I, Morellas V, Tsiamyrtzis P, Harp S. Urban surveillance system: From the laboratory to the commercial world[J]. Proceedings of the IEEE, 2001, 89(10): 1478-1497.
[4] Collins R, Lipton A, Kanade T, Fujiyoshi H, Duggins D, Tsin Y, Tolliver D, Enomoto N,Hasegawa O, Burt P, Wixson L. A system for video surveillance and monitoring[M], VSAM final report,Carnegie Mellon University: Technical Report CMU-RI-TR-00-12, 2000.
[5] Collins R, Lipton A, Fujiyoshi H, Kanade T. Algorithms for cooperative multisensor surveillance[J]. Proceedings of the IEEE, 2001, 89(10): 1456-1477.
[6] Bobick, A.F. and A,D. Wilson. A state-based technique for the summarization ofrecognition of gesture[J]. in Proc. Int'1 Conf. Computer jrsion. 1995.
[7] Darell, T., B. Moghaddam, and A. Pentland. Active face tracking and pose estimation in an Interactive Room[J]. in Proc. IEEE Conf. Computer Tfsion and Pattern Recognition.1996.
[8] Gavrila, D. and L. Davis. Tracking humans in action: a 3D model-based approach[J], in Proc. Image Understanding Workshop. 1996.
[9] Dellaert,F.,D.Pomerleau, and C. Thorpe. Model-based car tracking integrated with a road follower[J]. in Int'd Conf. on Automation. 1998.
[10] Taylor, C.J., J. Malik, and J. Weber. A real-time approach to stereopsis and lane-finding in Proc[M]. IEEE Intelligent Vehicles Symposium. 1996.
[11] Collins, R.T. and A.J. Lipton, Introduction to the special section on video surveillance[J].IEEE Transactions on Patern Analysis and Machine Intelligence,2000. 22(8): 745-746.
[12] Sage, K. and S. Young, Security applications of computer vision[J]. IEEE Aerospace and Electronics Systems Magazine, 1999. 14(4): 19-29.
[13] Meyer, M, and T. Ohmacht, jrdeo surveillance applications using multiple views of ascene[J]. IEEE Aerospace and Electronics Systems Magazine, 1999.14(3): 13-18.
[14] Haritaoglu, L, D. Harwood, and Davis. W/sup 4/一Who? When? Where? What? a real time system for detecting and tracking people. in Proceedings of Third[M] IEEE International Conference onAutomatic Face and Gesture Recognition. 1998.
[15] Stringa, E. and C.S. Regazzoni, Real-time video-shot detection for scene surveillance applications[J]. IEEE Transactions on Image Processing, 2000. 9:69-79.
[16] Harrell, R.C., D.C. Slaughter, and PD. Adsit, A uit-tracking system for robotic harvesting. [J] Macne Vision and Applications, 1989. 2: 69-80.
[17] Reynard, D., et al. Learing dynamics of complex motions m image sequences[J]in Proc. European Conf. Computer lesion. 1996.
[18] Bardinet, E., L. Cohen, and N. Ayache. Tracking medical 3D data with a deformable parametric model[M]. in Proc. European Conf. Computer Trsion. 1996.
[19] Shi, P., et al. A model-based integrated approach to track myocardial deformation using displacement and velocity constraints[M]. in Proc. int'I Conf. Computer Irsion. 1995.
[20] Boyer, E., Object models from contour sequences. Proc. European Conf. Computer Vision[J], 1996. 2: 109-118.
[21] Shapiro,L., Acne analysis of image sequences[C]. Cambridge, England. 1995(2), 109-118.
[22] Tomasi, C. and T. Kanade, Shape and motion form image streams under orthography: A fractorization method[J]. Int’l J. Computer Vision, 1992. 9(2): 137-154.
[23] Dellaert, Collins, Fast Image-Based Tracking by Selective Pixel Integration[J]. In ICCV Workshop on France-Rate Vision, Corfu, Greece, 1999.
[24] Lipton, Local Application of Optic Flow to Analyse Rigid versus Non-Rigid Motion[J]. In ICCV Workshop on France-Rate Vision, Corfu, Greece, 1999.
[25] Collins, Tsin, Calibration of an Outdoor Active Camera System[M]. In IEEE Computer Vision and Pattern Recognition (CVPR99), Fort Collins, CO, 1999: 528-534.
[26] Bogaert M, Cheleq N, Cornez P, et al, The PASSWORDS project. In Proceedings of IEEE international Conference on Image Processing[M], Lansanne,Switzerland, 1996: 780-785.
[27] Zhu Z, Xu G, Yang B, et al. A real-time vision system for automatic traffic monitoring[J].Image and Vision Computer, 2000, 18(10): 781-794.
[28] Massoud O, Papanikolopoulos N P. A novel method for tracking and counting depestrians in real-time using a single camera[J]. IEEE Transactions on Vehicular Technology, 2001, 50(5):1267-1278.
[29] Hu W., Tan T., Wang L., et al. A survey on visual surveillance of object motion and behaviors[M]. IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews. 2004, 34 (3): 334-352.
[30]王亮,胡卫民,谭铁牛。人运动的视觉分析综述[J]。计算机学报。2002, 25 (3):1-16
[31]李和平,胡占义,吴毅红,et al。基于半监督学习的行为建模与异常检测[J]。软件学报。2007, 18 (3): 527-537.
[32] Li Y., Ai H., Yamashita T., et al. Tracking in Low Frame Rate Video: A Cascade Particle Filter with Discriminative Observers of Different Lifespans[C], Proceedings of Computer Vision and Pattern Recognition, 2007. CVPR '07. IEEE Conference on. 2007: 1-8.
[33]朱琳,周杰,宋靖燕。一种新的基于在线采样的多物体跟踪方法[J]。计算机学报.2008, 31 (1): 151-160.
[34]李平安,敬忠良,胡士强。基于自适应表而模型的概率视频跟踪算法[J]。控制与决策。2007, 22 (1): 53-58.
[35]田广,戚飞虎。移动摄像机环境卜基于特征变换和SVM的分级行人检测算法[J]。电子学报。2008, 36 (5): 1024-1028.
[36] Zhang W., Wu Q. M. J., Yang X., et al. Multilevel Framework to Detect and Handle Vehicle Occlusion. Intelligent Transportation Systems[J], IEEE Transactions on. 2008, 9 (1): 161-174.
[37]侯志强,韩崇昭,视觉跟踪技术综述A Survey of Visual Tracking[J]自动化学报, 2006年04期.
[38]朱胜利.Mean Shift及相关算法在视频跟踪中的研究[D]浙江大学博士论文.2006.
[39] Wang S, Ai H. Z, He K. Z, Difference-image-based multiple motion targets detection and tracking[J]. Chinese Journal of Image and Graphics, 1999,4(6): 470- 475.
[40] Hou Z Q, Han C Z. A background reconstruction algorithm based on pixel intensity classification[J]. Chinese Journal of Software, 2005,16(9): 1568-1576.
[41] BarornJ,Fleet D and Beauchemin S. Performance of optical flow techniques[J].International Journal of Computer Vision,1994,12(1): 42-77.
[42] Meyer F. G., Bouthemy P. Region-based tracking using affine motion models in long image sequences. CVGIP[J]: Image Understanding. 1994, 60 (2):119-140.
[43] Hsu C.-T., Hsieh S.-M. Region Tracking for Non-Rigid Video Objects in a Non-Parametric MAP Framework. Signal Processing[J]: Image Communication.2006, 21 (3): 235-251.
[44] Yanlin Q., Hsu S., Sawhney H. S., et al. Robust Object Matching for Persistent Tracking with Heterogeneous Features[M]. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2007, 29 (5): 824-839.
[45] Kass M., Witkin A., Terzopoulous D. Snakes: active contour models, Proceedings of the International Conference on Computer Vision[C]. London: IEEE Computer Society 1987: 259-268.
[46] Leymarie F., Levine M. D. Tracking deformable objects in the plane using an active contour model[M]. IEEE Transactions on Pattern Analysis and MachineIntelligence. 1993,15 (6): 617-634.
[47] Karaulova 1, Hall P, Marshall A. A hierarchical model of dynamics for tracking people with asingle video camera [A]. Proceedings of British Machine Vision Conference [C], Bristol,U.K,2000:352-361.
[48] Ju S, Black M, Yaccpb Y. Cardboard people: A parameterized model of articulated image motion[A].IEEE International Conference on Automatic Faceand Gesture Recognition[C],Killington, Vermont USA, 1996:38-44.
[49] Wachter S, Nagel H-H. Tracking persons in monocular image sequences [J].Computer Vision and Image Understanding.l999, 74(5):174-192.
[50] Welch G, Bishop G An introduction to the Kalman filter[A]. 2004.
[51] Ernst Dieter D., Volker G. Applications of dynamic monocular machine vision.Machine Vision and Applications[J]. 1988, 1 (4): 241-261.
[52] Donald B. G. Visual of known three-dimensional[J] objects.International Journal of Computer Vision. 1992, 7 (3): 243-270.
[53] Matthies L. H., Kanade T., Szeliski R. Kalman Filter-based Algorithms for Estimating Depth from Image Sequences[M]. International Journal of Computer Vision. 1989, 3 (3): 209-238.
[54] Gordon N. J., Salmond D. J., Smith A. F. M. Novel approach to nonlinear/non-Gaussian Bayesian state estimation[J]. IEE Proceedings on Radar and Signal Processing. 1993, 140(2): 107-113.
[55] Arulampalam M. S., Maskell S., Gordon N., et al. A tutorial on particle filters for online nonlinear/non-gaussian bayesian tracking[J]. IEEE Transactions on Signal Processing. 2002, 50 (2): 174-188.
[56] Isard M., Blake A. CONDENSATION-conditional density propagation for visual tracking[J]. International Journal of Computer Vision. 1998, 29 (1): 5-28.
[57] Katja Nummiaro E. K.-M., and Luc Van Cool. An adaptive color-based filter particle filter[J]. Image and Vision Computing. 2002, 21 (1): 99-110.
[58] Chang C., Ansari R. Kernel particle filter for visual tracking[J]. IEEE Signal Processing Letters. 2005, 12 (3): 242-245.
[59] Zhu G, Liang D., Liu Y., et al. Improving particle filter with support vector regression for efficient visual tracking[J], Proceedings of International Conference on Image Processing. Genova: IEEE, 2005: 422-425.
[60] K. Fukunaga and L.D. Hostetler.The Estimation of the Gradient of a Density Function, with Applications in Pattern Recognition[J]," IEEE Trans. Information Theory, 1975,21(1), 32-40.
[61] Yizong Cheng, Mean Shift, Mode Seeking, and Clustering[J], IEEE Transactions on Pattern Analysis and Machine Intelligence, 1995,17 (8):790-799.
[62] Comaniciu D, Ramesh V, Meer P. Real-Time tracking of non-rigid objects using meanshift[C]. In:Proc. of the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR).2000: 142-149.
[63] C.Yang,R. Duraiswami, D. DeMenthon and L. Davis. Mean-Shift Analysis Using Quasi-Newton Methods[C]. In IEEE International Conference on Image Processing. 2003(3):447-450.
[64] M. Fashing and C. Tomasi.Mean shift is a bound optimization[J]. IEEE Trans. Pattern Anal. Machine Intell. 2005, 25 (3):471-474.
[65]李乡儒,吴福朝,吴战义.均值漂移算法的收敛性[J].软件学报.2005, 16(3):365-374.
[66] Comaniciu D, Ramesh V, Meer P. Real-Time tracking of non-rigid objects using mean shift[J]. In:Proc. of the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR).2000: 142-149.
[67] Comaniciu D, Ramesh V. Mean shift and optimal prediction for efficient object tracking[J]. In:Mojsilovic A, Hu J, eds. Proc. of the IEEE Int’1 Conf. on Image Processing (ICIP).2000:70-73.
[68] Comaniciu D, Ramesh V, Meer P. Kernel-Based object tracking[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence.2003,25 (5):564-575.
[69] Zhou XS, Comaniciu D, Krishnan S. An information fusion framework for robust shape tracking[J].Workshop on Statistical and Computational Theories of Vision SCTV.2003:1-24.
[70] Comaniciu D. An Algorithm for Data-Driven Bandwidth Selection[J], IEEE Trans. Pattern Analysis and Machine Intelligence, 2003, 25(2):1-8.
[71]彭宁篙,杨杰等.Mean-Shift跟踪算法中核函数窗宽的自动选取[J],软件学报,2005, 16(9):1542-1550.
[72] Peng, N S,Yang J, Liu Z. Mean shift blob tracking with kernel histogram filtering and hypothesis testing[J], Pattern Recotion Letters, 2005, 26(5): 605-614.
[73] Bradski G R, Clara S. Computer vision face tracking for use in a perceptual user interface[J]. Intelligence Technology Journal,1998,2(2): 1-15.
[74] Bradski GR. Computer vision face tracking for use in a perceptual user interface. In:Regina Spencer Sipple[J].IEEE Workshop onApplications of Computer Vision. Stoughton:Printing House. 1998:214-219 .
[75] Zhu Z, Ji Q, Fujimura K, et al. Combining Kalman Filtering and Mean Shift for Real Time Eye Tracking Under Active IR Illumination[J], In Proc. of 16th Int. Conf. on Pattern Recognition, New York: IEEE press, 2002, 4:318-321.
[76] C. Yang, R. Duraiswami, and L.Davis. Similarity Measure for Nonparametric Kernel Density Based Object Tracking. Eighteenth Annual[C] Conference Neural Information ProcessingSystems. 2005:1561-1568.
[77] C. Yang, R. Duraiswami, A. Elgammal and L. Davis. Real-Time Kernel-Based Tracking in Joint Feature-Spatial Spaces[J] Technical Report CS-TR-4567. University of Maryland. 2003.
[78]朱胜利,朱善安,李旭超.快速运动目标的Mean shift跟踪算法[J].光电工程.2006,33 (5):66-70.
[79]贾静平,张艳宁,柴艳妹,赵荣椿.目标多自由度Mean Shift序列图像跟踪算法[J].西北工业大学学报.2005, 23 (5) :618-622.
[80]李培华一种改进的mean shift跟踪算法[J].自动化学报.2007, 33 (4 ):347-354.
[81]程伟,杨杰一种基于均值移位的红外目标跟踪新方法[J].红外与毫米波学报.2005,24 (3):231-235.
[82] K. Deguchi, O.Kawanaka and T.Okatani. Object tracking by the mean-shift of regional colour distribution combined with the particle-filter algorithm[J]. Proc. ICPR. 2004:506-509.
[83] Comaniciu D, Meer P. Mean shift:A robust approach toward feature space analysis[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence. 2002, 24(5):603-619.
[84] Comaniciu D, Meer P. Mean shift analysis and applications[J]. In: Proc. of the IEEE Int’1 Conf. on Computer Vision (ICCV).1999:1197-1203.
[85] A. Bandera, J. M. Perez-Lorenzo et al. Mean shift based clustering of Hough domain for fast line segment detection[J]. Pattern Recognition Letters.2006(27):578-586.
[86] Yuzhong Wang etal.Unsupervised color-texture segmentation based on soft criterionwith adaptive mean-shift clustering[J]. PRL. 2006, 27 (5):386-392.
[87] Wang J, Thiesson B, Xu YQ, Cohen M. Image and video segmentation by anisotropic kernel mean shift[J]. In:Proc. of the European Conf. on Computer Vision(ECCV).2004.
[88] Barash D, Comaniciu D. A common framework for nonlinear diffusion adaptive smoothing, bilateral filtering and mean shift. Image and Video Computing[M]. 2003, 22 (1):73-81.
[89] Comaniciu D, Ramesh V, Bue AD. Multivariate saddle point detection for statistical clustering[M]. In:Proc. of the European Conf. Computer Vision (ECCV).2002:561-576.
[90] M.A.Carreira-Perpinbn.Fast Nonparametric Clustering with Gaussian Blurring Mean-Shift[A],Proc. 23rd In t'1 Conf. Machine Learning,2006.
[91] Georgescu, B. Shimshoni, Meer P. Mean shift based clustering in high dimensions:a texture classification example[C]. In:IEEE Int. Conf. on Computer Vision. 2003 (2):456-463.
[92]薛孝琴.基于改进Mean-Shift算法的粒子滤波跟踪的应用研究.武汉科技大学硕士学位论文.2009.
[93] Comaniciu D,Ramesh V,Meer P.Kernel-Based object tracking[J].IEEE Transaction on Pattern Analysis and Machine Intelligence,2003,(25):564-575.
[94] Collins RT.Mean-Shift blob tracking through scale space[A].In:IEEE International Conference on Computer Vision and Pattern Recognition [C],Baltimore:Victor Graphics,2003(2):234-240.
[95]钱惠敏,茅耀斌,王执铨.自动选择跟踪窗尺度的Mean-Shift算法[J].中国图象图形学报,2007,13(2):245-250.
[96] Yang C, Duraiswami R, Davis L. Efficient Mean-Shift Tracking via a New Similarity Measure[J], In Proc. of the 2005 IEEE Computer Society Conf. on Computer Vision and Pattern Recognition, Washington DC: IEEE Computer Society, 2005,1:176-183.
[97] Nguyen H T , Worring M , Rein van den Boomgarrd. Occlusion robust adaptive template t racking [C] .Proceedings of Eighth International Conference on Computer Vision ( ICCV’01) , Vancouver , British Columbia , Canada , 2001 , (1) :678-683.
[98]朱胜利,朱善安.基于卡尔曼滤波器组的meanshift模板更新算法[J].中国图像图形学报, 2007 ,12 (3) :460-465.
[99]彭宁嵩,杨杰,刘志,张凤超. Mean Shift跟踪算法中核函数窗宽的自动选取[J].软件学报,2005 ,16 (9) :1542-1550.
[100]朱胜利,朱善安.核函数带宽自适应的Mean shift目标跟踪算法[J].光电工程,2006,33 (8) :11-16.
[101] D. Comaniciu, V. Ramesh, and P. Meer, "Real-Time Tracking of Non-Rigid Objects Using Mean Shift," [J] IEEE Computer Vision and Pattern Recognition, 2000,2: 142-149.
[102] D Comaniciu, P Meer, Mean shift: a robust approach toward feature space analysis, [J] IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, 24(5):603-619.
[103] Collins RT. Mean-Shift blob tracking through scale space[C]. Proceeding of the conference on Computer Vision and Pattern Recognition, 2003:234~240.
[104]左军毅,梁彦,赵春晖,潘泉,张洪才等. Mean Shift跟踪算法中尺度自适应策略的研究[J].中国图象图形学报.2008,13(9): 1750-1757.
[105]高建坡,王煜坚,杨浩,吴镇扬.基于均值移动和椭圆拟合的人脸跟踪算法[J].东南大学学报(自然科学版),2006,36(6):897-902.
[2] Haritaoglu I, Harwood D, Davis L. W4: Real-time surveillance of people and their activities.[J]IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(8): 809-830.
[3] Pavlidis I, Morellas V, Tsiamyrtzis P, Harp S. Urban surveillance system: From the laboratory to the commercial world[J]. Proceedings of the IEEE, 2001, 89(10): 1478-1497.
[4] Collins R, Lipton A, Kanade T, Fujiyoshi H, Duggins D, Tsin Y, Tolliver D, Enomoto N,Hasegawa O, Burt P, Wixson L. A system for video surveillance and monitoring[M], VSAM final report,Carnegie Mellon University: Technical Report CMU-RI-TR-00-12, 2000.
[5] Collins R, Lipton A, Fujiyoshi H, Kanade T. Algorithms for cooperative multisensor surveillance[J]. Proceedings of the IEEE, 2001, 89(10): 1456-1477.
[6] Bobick, A.F. and A,D. Wilson. A state-based technique for the summarization ofrecognition of gesture[J]. in Proc. Int'1 Conf. Computer jrsion. 1995.
[7] Darell, T., B. Moghaddam, and A. Pentland. Active face tracking and pose estimation in an Interactive Room[J]. in Proc. IEEE Conf. Computer Tfsion and Pattern Recognition.1996.
[8] Gavrila, D. and L. Davis. Tracking humans in action: a 3D model-based approach[J], in Proc. Image Understanding Workshop. 1996.
[9] Dellaert,F.,D.Pomerleau, and C. Thorpe. Model-based car tracking integrated with a road follower[J]. in Int'd Conf. on Automation. 1998.
[10] Taylor, C.J., J. Malik, and J. Weber. A real-time approach to stereopsis and lane-finding in Proc[M]. IEEE Intelligent Vehicles Symposium. 1996.
[11] Collins, R.T. and A.J. Lipton, Introduction to the special section on video surveillance[J].IEEE Transactions on Patern Analysis and Machine Intelligence,2000. 22(8): 745-746.
[12] Sage, K. and S. Young, Security applications of computer vision[J]. IEEE Aerospace and Electronics Systems Magazine, 1999. 14(4): 19-29.
[13] Meyer, M, and T. Ohmacht, jrdeo surveillance applications using multiple views of ascene[J]. IEEE Aerospace and Electronics Systems Magazine, 1999.14(3): 13-18.
[14] Haritaoglu, L, D. Harwood, and Davis. W/sup 4/一Who? When? Where? What? a real time system for detecting and tracking people. in Proceedings of Third[M] IEEE International Conference onAutomatic Face and Gesture Recognition. 1998.
[15] Stringa, E. and C.S. Regazzoni, Real-time video-shot detection for scene surveillance applications[J]. IEEE Transactions on Image Processing, 2000. 9:69-79.
[16] Harrell, R.C., D.C. Slaughter, and PD. Adsit, A uit-tracking system for robotic harvesting. [J] Macne Vision and Applications, 1989. 2: 69-80.
[17] Reynard, D., et al. Learing dynamics of complex motions m image sequences[J]in Proc. European Conf. Computer lesion. 1996.
[18] Bardinet, E., L. Cohen, and N. Ayache. Tracking medical 3D data with a deformable parametric model[M]. in Proc. European Conf. Computer Trsion. 1996.
[19] Shi, P., et al. A model-based integrated approach to track myocardial deformation using displacement and velocity constraints[M]. in Proc. int'I Conf. Computer Irsion. 1995.
[20] Boyer, E., Object models from contour sequences. Proc. European Conf. Computer Vision[J], 1996. 2: 109-118.
[21] Shapiro,L., Acne analysis of image sequences[C]. Cambridge, England. 1995(2), 109-118.
[22] Tomasi, C. and T. Kanade, Shape and motion form image streams under orthography: A fractorization method[J]. Int’l J. Computer Vision, 1992. 9(2): 137-154.
[23] Dellaert, Collins, Fast Image-Based Tracking by Selective Pixel Integration[J]. In ICCV Workshop on France-Rate Vision, Corfu, Greece, 1999.
[24] Lipton, Local Application of Optic Flow to Analyse Rigid versus Non-Rigid Motion[J]. In ICCV Workshop on France-Rate Vision, Corfu, Greece, 1999.
[25] Collins, Tsin, Calibration of an Outdoor Active Camera System[M]. In IEEE Computer Vision and Pattern Recognition (CVPR99), Fort Collins, CO, 1999: 528-534.
[26] Bogaert M, Cheleq N, Cornez P, et al, The PASSWORDS project. In Proceedings of IEEE international Conference on Image Processing[M], Lansanne,Switzerland, 1996: 780-785.
[27] Zhu Z, Xu G, Yang B, et al. A real-time vision system for automatic traffic monitoring[J].Image and Vision Computer, 2000, 18(10): 781-794.
[28] Massoud O, Papanikolopoulos N P. A novel method for tracking and counting depestrians in real-time using a single camera[J]. IEEE Transactions on Vehicular Technology, 2001, 50(5):1267-1278.
[29] Hu W., Tan T., Wang L., et al. A survey on visual surveillance of object motion and behaviors[M]. IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews. 2004, 34 (3): 334-352.
[30]王亮,胡卫民,谭铁牛。人运动的视觉分析综述[J]。计算机学报。2002, 25 (3):1-16
[31]李和平,胡占义,吴毅红,et al。基于半监督学习的行为建模与异常检测[J]。软件学报。2007, 18 (3): 527-537.
[32] Li Y., Ai H., Yamashita T., et al. Tracking in Low Frame Rate Video: A Cascade Particle Filter with Discriminative Observers of Different Lifespans[C], Proceedings of Computer Vision and Pattern Recognition, 2007. CVPR '07. IEEE Conference on. 2007: 1-8.
[33]朱琳,周杰,宋靖燕。一种新的基于在线采样的多物体跟踪方法[J]。计算机学报.2008, 31 (1): 151-160.
[34]李平安,敬忠良,胡士强。基于自适应表而模型的概率视频跟踪算法[J]。控制与决策。2007, 22 (1): 53-58.
[35]田广,戚飞虎。移动摄像机环境卜基于特征变换和SVM的分级行人检测算法[J]。电子学报。2008, 36 (5): 1024-1028.
[36] Zhang W., Wu Q. M. J., Yang X., et al. Multilevel Framework to Detect and Handle Vehicle Occlusion. Intelligent Transportation Systems[J], IEEE Transactions on. 2008, 9 (1): 161-174.
[37]侯志强,韩崇昭,视觉跟踪技术综述A Survey of Visual Tracking[J]自动化学报, 2006年04期.
[38]朱胜利.Mean Shift及相关算法在视频跟踪中的研究[D]浙江大学博士论文.2006.
[39] Wang S, Ai H. Z, He K. Z, Difference-image-based multiple motion targets detection and tracking[J]. Chinese Journal of Image and Graphics, 1999,4(6): 470- 475.
[40] Hou Z Q, Han C Z. A background reconstruction algorithm based on pixel intensity classification[J]. Chinese Journal of Software, 2005,16(9): 1568-1576.
[41] BarornJ,Fleet D and Beauchemin S. Performance of optical flow techniques[J].International Journal of Computer Vision,1994,12(1): 42-77.
[42] Meyer F. G., Bouthemy P. Region-based tracking using affine motion models in long image sequences. CVGIP[J]: Image Understanding. 1994, 60 (2):119-140.
[43] Hsu C.-T., Hsieh S.-M. Region Tracking for Non-Rigid Video Objects in a Non-Parametric MAP Framework. Signal Processing[J]: Image Communication.2006, 21 (3): 235-251.
[44] Yanlin Q., Hsu S., Sawhney H. S., et al. Robust Object Matching for Persistent Tracking with Heterogeneous Features[M]. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2007, 29 (5): 824-839.
[45] Kass M., Witkin A., Terzopoulous D. Snakes: active contour models, Proceedings of the International Conference on Computer Vision[C]. London: IEEE Computer Society 1987: 259-268.
[46] Leymarie F., Levine M. D. Tracking deformable objects in the plane using an active contour model[M]. IEEE Transactions on Pattern Analysis and MachineIntelligence. 1993,15 (6): 617-634.
[47] Karaulova 1, Hall P, Marshall A. A hierarchical model of dynamics for tracking people with asingle video camera [A]. Proceedings of British Machine Vision Conference [C], Bristol,U.K,2000:352-361.
[48] Ju S, Black M, Yaccpb Y. Cardboard people: A parameterized model of articulated image motion[A].IEEE International Conference on Automatic Faceand Gesture Recognition[C],Killington, Vermont USA, 1996:38-44.
[49] Wachter S, Nagel H-H. Tracking persons in monocular image sequences [J].Computer Vision and Image Understanding.l999, 74(5):174-192.
[50] Welch G, Bishop G An introduction to the Kalman filter[A]. 2004.
[51] Ernst Dieter D., Volker G. Applications of dynamic monocular machine vision.Machine Vision and Applications[J]. 1988, 1 (4): 241-261.
[52] Donald B. G. Visual of known three-dimensional[J] objects.International Journal of Computer Vision. 1992, 7 (3): 243-270.
[53] Matthies L. H., Kanade T., Szeliski R. Kalman Filter-based Algorithms for Estimating Depth from Image Sequences[M]. International Journal of Computer Vision. 1989, 3 (3): 209-238.
[54] Gordon N. J., Salmond D. J., Smith A. F. M. Novel approach to nonlinear/non-Gaussian Bayesian state estimation[J]. IEE Proceedings on Radar and Signal Processing. 1993, 140(2): 107-113.
[55] Arulampalam M. S., Maskell S., Gordon N., et al. A tutorial on particle filters for online nonlinear/non-gaussian bayesian tracking[J]. IEEE Transactions on Signal Processing. 2002, 50 (2): 174-188.
[56] Isard M., Blake A. CONDENSATION-conditional density propagation for visual tracking[J]. International Journal of Computer Vision. 1998, 29 (1): 5-28.
[57] Katja Nummiaro E. K.-M., and Luc Van Cool. An adaptive color-based filter particle filter[J]. Image and Vision Computing. 2002, 21 (1): 99-110.
[58] Chang C., Ansari R. Kernel particle filter for visual tracking[J]. IEEE Signal Processing Letters. 2005, 12 (3): 242-245.
[59] Zhu G, Liang D., Liu Y., et al. Improving particle filter with support vector regression for efficient visual tracking[J], Proceedings of International Conference on Image Processing. Genova: IEEE, 2005: 422-425.
[60] K. Fukunaga and L.D. Hostetler.The Estimation of the Gradient of a Density Function, with Applications in Pattern Recognition[J]," IEEE Trans. Information Theory, 1975,21(1), 32-40.
[61] Yizong Cheng, Mean Shift, Mode Seeking, and Clustering[J], IEEE Transactions on Pattern Analysis and Machine Intelligence, 1995,17 (8):790-799.
[62] Comaniciu D, Ramesh V, Meer P. Real-Time tracking of non-rigid objects using meanshift[C]. In:Proc. of the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR).2000: 142-149.
[63] C.Yang,R. Duraiswami, D. DeMenthon and L. Davis. Mean-Shift Analysis Using Quasi-Newton Methods[C]. In IEEE International Conference on Image Processing. 2003(3):447-450.
[64] M. Fashing and C. Tomasi.Mean shift is a bound optimization[J]. IEEE Trans. Pattern Anal. Machine Intell. 2005, 25 (3):471-474.
[65]李乡儒,吴福朝,吴战义.均值漂移算法的收敛性[J].软件学报.2005, 16(3):365-374.
[66] Comaniciu D, Ramesh V, Meer P. Real-Time tracking of non-rigid objects using mean shift[J]. In:Proc. of the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR).2000: 142-149.
[67] Comaniciu D, Ramesh V. Mean shift and optimal prediction for efficient object tracking[J]. In:Mojsilovic A, Hu J, eds. Proc. of the IEEE Int’1 Conf. on Image Processing (ICIP).2000:70-73.
[68] Comaniciu D, Ramesh V, Meer P. Kernel-Based object tracking[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence.2003,25 (5):564-575.
[69] Zhou XS, Comaniciu D, Krishnan S. An information fusion framework for robust shape tracking[J].Workshop on Statistical and Computational Theories of Vision SCTV.2003:1-24.
[70] Comaniciu D. An Algorithm for Data-Driven Bandwidth Selection[J], IEEE Trans. Pattern Analysis and Machine Intelligence, 2003, 25(2):1-8.
[71]彭宁篙,杨杰等.Mean-Shift跟踪算法中核函数窗宽的自动选取[J],软件学报,2005, 16(9):1542-1550.
[72] Peng, N S,Yang J, Liu Z. Mean shift blob tracking with kernel histogram filtering and hypothesis testing[J], Pattern Recotion Letters, 2005, 26(5): 605-614.
[73] Bradski G R, Clara S. Computer vision face tracking for use in a perceptual user interface[J]. Intelligence Technology Journal,1998,2(2): 1-15.
[74] Bradski GR. Computer vision face tracking for use in a perceptual user interface. In:Regina Spencer Sipple[J].IEEE Workshop onApplications of Computer Vision. Stoughton:Printing House. 1998:214-219 .
[75] Zhu Z, Ji Q, Fujimura K, et al. Combining Kalman Filtering and Mean Shift for Real Time Eye Tracking Under Active IR Illumination[J], In Proc. of 16th Int. Conf. on Pattern Recognition, New York: IEEE press, 2002, 4:318-321.
[76] C. Yang, R. Duraiswami, and L.Davis. Similarity Measure for Nonparametric Kernel Density Based Object Tracking. Eighteenth Annual[C] Conference Neural Information ProcessingSystems. 2005:1561-1568.
[77] C. Yang, R. Duraiswami, A. Elgammal and L. Davis. Real-Time Kernel-Based Tracking in Joint Feature-Spatial Spaces[J] Technical Report CS-TR-4567. University of Maryland. 2003.
[78]朱胜利,朱善安,李旭超.快速运动目标的Mean shift跟踪算法[J].光电工程.2006,33 (5):66-70.
[79]贾静平,张艳宁,柴艳妹,赵荣椿.目标多自由度Mean Shift序列图像跟踪算法[J].西北工业大学学报.2005, 23 (5) :618-622.
[80]李培华一种改进的mean shift跟踪算法[J].自动化学报.2007, 33 (4 ):347-354.
[81]程伟,杨杰一种基于均值移位的红外目标跟踪新方法[J].红外与毫米波学报.2005,24 (3):231-235.
[82] K. Deguchi, O.Kawanaka and T.Okatani. Object tracking by the mean-shift of regional colour distribution combined with the particle-filter algorithm[J]. Proc. ICPR. 2004:506-509.
[83] Comaniciu D, Meer P. Mean shift:A robust approach toward feature space analysis[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence. 2002, 24(5):603-619.
[84] Comaniciu D, Meer P. Mean shift analysis and applications[J]. In: Proc. of the IEEE Int’1 Conf. on Computer Vision (ICCV).1999:1197-1203.
[85] A. Bandera, J. M. Perez-Lorenzo et al. Mean shift based clustering of Hough domain for fast line segment detection[J]. Pattern Recognition Letters.2006(27):578-586.
[86] Yuzhong Wang etal.Unsupervised color-texture segmentation based on soft criterionwith adaptive mean-shift clustering[J]. PRL. 2006, 27 (5):386-392.
[87] Wang J, Thiesson B, Xu YQ, Cohen M. Image and video segmentation by anisotropic kernel mean shift[J]. In:Proc. of the European Conf. on Computer Vision(ECCV).2004.
[88] Barash D, Comaniciu D. A common framework for nonlinear diffusion adaptive smoothing, bilateral filtering and mean shift. Image and Video Computing[M]. 2003, 22 (1):73-81.
[89] Comaniciu D, Ramesh V, Bue AD. Multivariate saddle point detection for statistical clustering[M]. In:Proc. of the European Conf. Computer Vision (ECCV).2002:561-576.
[90] M.A.Carreira-Perpinbn.Fast Nonparametric Clustering with Gaussian Blurring Mean-Shift[A],Proc. 23rd In t'1 Conf. Machine Learning,2006.
[91] Georgescu, B. Shimshoni, Meer P. Mean shift based clustering in high dimensions:a texture classification example[C]. In:IEEE Int. Conf. on Computer Vision. 2003 (2):456-463.
[92]薛孝琴.基于改进Mean-Shift算法的粒子滤波跟踪的应用研究.武汉科技大学硕士学位论文.2009.
[93] Comaniciu D,Ramesh V,Meer P.Kernel-Based object tracking[J].IEEE Transaction on Pattern Analysis and Machine Intelligence,2003,(25):564-575.
[94] Collins RT.Mean-Shift blob tracking through scale space[A].In:IEEE International Conference on Computer Vision and Pattern Recognition [C],Baltimore:Victor Graphics,2003(2):234-240.
[95]钱惠敏,茅耀斌,王执铨.自动选择跟踪窗尺度的Mean-Shift算法[J].中国图象图形学报,2007,13(2):245-250.
[96] Yang C, Duraiswami R, Davis L. Efficient Mean-Shift Tracking via a New Similarity Measure[J], In Proc. of the 2005 IEEE Computer Society Conf. on Computer Vision and Pattern Recognition, Washington DC: IEEE Computer Society, 2005,1:176-183.
[97] Nguyen H T , Worring M , Rein van den Boomgarrd. Occlusion robust adaptive template t racking [C] .Proceedings of Eighth International Conference on Computer Vision ( ICCV’01) , Vancouver , British Columbia , Canada , 2001 , (1) :678-683.
[98]朱胜利,朱善安.基于卡尔曼滤波器组的meanshift模板更新算法[J].中国图像图形学报, 2007 ,12 (3) :460-465.
[99]彭宁嵩,杨杰,刘志,张凤超. Mean Shift跟踪算法中核函数窗宽的自动选取[J].软件学报,2005 ,16 (9) :1542-1550.
[100]朱胜利,朱善安.核函数带宽自适应的Mean shift目标跟踪算法[J].光电工程,2006,33 (8) :11-16.
[101] D. Comaniciu, V. Ramesh, and P. Meer, "Real-Time Tracking of Non-Rigid Objects Using Mean Shift," [J] IEEE Computer Vision and Pattern Recognition, 2000,2: 142-149.
[102] D Comaniciu, P Meer, Mean shift: a robust approach toward feature space analysis, [J] IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, 24(5):603-619.
[103] Collins RT. Mean-Shift blob tracking through scale space[C]. Proceeding of the conference on Computer Vision and Pattern Recognition, 2003:234~240.
[104]左军毅,梁彦,赵春晖,潘泉,张洪才等. Mean Shift跟踪算法中尺度自适应策略的研究[J].中国图象图形学报.2008,13(9): 1750-1757.
[105]高建坡,王煜坚,杨浩,吴镇扬.基于均值移动和椭圆拟合的人脸跟踪算法[J].东南大学学报(自然科学版),2006,36(6):897-902.