运动目标检测与跟踪技术研究
|
摘要
随着信息技术的高速发展以及安全形势的迫切需要,人们对安防设备的智能性要求日益提高,智能视频监控技术已成为当今较为前沿的研究课题。其中,运动目标检测和跟踪技术作为智能视频监控的核心技术日益受到业界的关注。
本文在参阅国内外大量相关文献基础上,针对视频监控系统中运动目标检测与跟踪功能难以兼顾可靠性与稳定性的缺点,着重研究视频监控系统中运动目标检测和跟踪的关键算法。在运动目标检测技术方面,针对帧间差分法和背景差分法两种算法的不足,提出了一种帧间差分法与高斯分布背景建模相结合的目标检测算法。在采用帧间差分法中,引入了“分块赋权值”的思想和通过设置上下限两个阈值判断是否出现运动目标,从而提高了目标检测的准确性和鲁棒性,实现了对运动目标的准确检测。在运动目标跟踪技术方面,针对已有粒子滤波算法中存在的退化问题和样本贫化问题,提出了一种针对重采样过程改进的粒子滤波算法,通过合理加权线性组合的方法对粒子权值进行重新赋值,改善了样本集的多样性,提高了粒子滤波算法的估计与跟踪能力。最后,进行了仿真实验,证明了本文提出的目标检测算法和跟踪算法的可行性和有效性。
Along with the fast development of information technology as well as the urgent need of security situation, the requirement for higher-powered security equipment is ever-increasing. Now, the techniques of video intelligent surveillance have been become a essential study. Moving target detection technology and tracking technology, which is the core of intelligent video surveillance technology, are becoming increasingly subject to the industry’s attention.
On the basis of a large number of reference literature at home and abroad, as to the shortcomings that the moving target detection and tracking capabilities difficult to balance reliability and stability in the intelligent monitoring and control system, we focused on studied the key algorithm of moving target detection and target tracking in the video surveillance system. On the research of moving target detection, as to faults of frame difference method and background difference method, the thesis puts forward a new method that uses the combination of frame difference method with Gaussian distribution background model. In the use of frame difference method, introduced the concept of“Empowerment block value”thought, and by setting the upper and lower threshold to determine whether appears moving targets. This method improves the accuracy and robustness of detection algorithm, implement moving target detection. On the research of moving target tracking, Specific to the particle degradation problem and sample dilution phenomenon of particle filter method, this paper proposes a new particle filter algorithm that as to resample process, by the way of reasonable weighted linear combination to particle weight re-assignment, the algorithm improve the diversity of the samples, and improved the estimation and tracking abilities. Finally, the thesis designs simulation experiment, and the results show the moving detection method and tracking method that this paper proposed are effectiveness.
本文在参阅国内外大量相关文献基础上,针对视频监控系统中运动目标检测与跟踪功能难以兼顾可靠性与稳定性的缺点,着重研究视频监控系统中运动目标检测和跟踪的关键算法。在运动目标检测技术方面,针对帧间差分法和背景差分法两种算法的不足,提出了一种帧间差分法与高斯分布背景建模相结合的目标检测算法。在采用帧间差分法中,引入了“分块赋权值”的思想和通过设置上下限两个阈值判断是否出现运动目标,从而提高了目标检测的准确性和鲁棒性,实现了对运动目标的准确检测。在运动目标跟踪技术方面,针对已有粒子滤波算法中存在的退化问题和样本贫化问题,提出了一种针对重采样过程改进的粒子滤波算法,通过合理加权线性组合的方法对粒子权值进行重新赋值,改善了样本集的多样性,提高了粒子滤波算法的估计与跟踪能力。最后,进行了仿真实验,证明了本文提出的目标检测算法和跟踪算法的可行性和有效性。
Along with the fast development of information technology as well as the urgent need of security situation, the requirement for higher-powered security equipment is ever-increasing. Now, the techniques of video intelligent surveillance have been become a essential study. Moving target detection technology and tracking technology, which is the core of intelligent video surveillance technology, are becoming increasingly subject to the industry’s attention.
On the basis of a large number of reference literature at home and abroad, as to the shortcomings that the moving target detection and tracking capabilities difficult to balance reliability and stability in the intelligent monitoring and control system, we focused on studied the key algorithm of moving target detection and target tracking in the video surveillance system. On the research of moving target detection, as to faults of frame difference method and background difference method, the thesis puts forward a new method that uses the combination of frame difference method with Gaussian distribution background model. In the use of frame difference method, introduced the concept of“Empowerment block value”thought, and by setting the upper and lower threshold to determine whether appears moving targets. This method improves the accuracy and robustness of detection algorithm, implement moving target detection. On the research of moving target tracking, Specific to the particle degradation problem and sample dilution phenomenon of particle filter method, this paper proposes a new particle filter algorithm that as to resample process, by the way of reasonable weighted linear combination to particle weight re-assignment, the algorithm improve the diversity of the samples, and improved the estimation and tracking abilities. Finally, the thesis designs simulation experiment, and the results show the moving detection method and tracking method that this paper proposed are effectiveness.
引文
[1] A.HamPapur, L .Brown. Application. Technologies and application. Information, Communications and Signal Processing. Proceedings of the 2003 Joint Conference of the Fourth Internation Conference. 2003, Vol.2. 1133-1138.
[2] MeKennaT. Video surveillance and human activity recognition for anti-terrors and force protection. IEEE Conference on Advanced Video and signal Based Surveillance. 2003. 22-23.
[3] E.Montseny, J.Frau. specialized processors for real-time image analysis.Workshop Proceedings, Barcelona, Spain, 1991.9.
[4]赵荣椿.数字图像处理导论.西安:西北工业大学出版社, 2000.
[5] RobertT.Collins, Alan J.Lipton. A System for Video Surveillance and Monitoring. Carnegie Mellon University. 2000.
[6]中国科学院自动化研究所模式识别国家重点实验室视觉监控研究组,智能视觉监控,http;//www.sinosurveillance.com/.
[7] Haritaoglu I, Harwood.D, etal. Real-time Surveillance of people and Their Activities. IEEE Transactions on Pattern Analysis and Machine Intelligence.2000, 22(8). 809-830.
[8] Lipto and H.Fujiyoshi, etc. Moving Target Classification and Tracking from Real-time Video. IEEE workshop Application of Computer Vision.1998, 10. 8-14.
[9] Foresti, GL. Object recognition and tracking for remote video surveillance. IEEE Transaction of Circuits and Systems for Video Technology.1999, 9(7). 1045-1062.
[10] A.Murat Tekalp. Digital video processing. Tsinghua University Press. 1998. 380.
[11] Horn B, Schunch B.Detemining optical flow. Artificial Intelligence.1981.185-203.
[12] Haritaoglu and Harwood, Davis. Real-time system for detecting and tracking People. Proceedings of Third IEEE International Conference of Automatic Face and Gesture Recoginition.1998. 222-227.
[13] Stringa, E.and Regazzoni. Real-time video-shot detection for scene surveillance applications. IEEE Transactions of Image Processing. 2000, 1, 9(1). 19-69.
[14] Robert T.Collins, Alan J.Lipto. A System for Video Surveillance and Monitoring. Carnegie Mellon University. 2000.
[15] Haritaoglu I, Harwood.D. Real-time Surveillance of People and Their Activities. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(8).809-830.
[16]魏志强,纪筱鹏,冯业伟.基于自适应背景图像更新的运动目标检测方法.电子学报.2005, 33(12). 2261-2264.
[17] Gutchess D, Trajkovics M, Cohen-Solal E. A background model initialization algorithm for video surveillance .Computer Vision.2001.ICCV 2001. Proceedings Eighth IEEE International Conference. 2001, 6, Vol.1. 733-740.
[18] Jwu-Sheng Hu, Tzung-Min Su, Shr-Chi Jeng. Robust Background Subtraction with Shadow and Highlight. Removal for Indoor Surveillance Intelligent Robots and Systems, 2006 IEEE/RSJ International Conference. 2006, 10. 4545-4550.
[19] Cezar Silveira, Jacques J, Rosito Jung C. A Background Subtraction Model Adapted to Illumination Changes. Image Processing, 2006 IEEE International Conference. 2006, 10. 1817-1820.
[20]郑江滨.视频监视方法研究.西北工业大学博士学位论文.2002. 21-22.
[21]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.
[22] C.Wren, A. Azarbayejani, T.Darrel. Real-time Tracking of the Human Body. IEEE Trans. 1997, 6, 19(7). 780-785.
[23]郑江滨、张艳宁、冯大淦.视频监视中运动目标的检测与跟踪算法.系统工程与电子技术.2002, 24(10). 34-37.
[24]张文杰,戚飞虎,江卓军.实时视频监控系统中运动目标检测和跟踪的一种实用方法.上海交通大学学报.2002,36(12).1837-1840.
[25] C.Stauffer, W.E.L.Gramson. Adaptive background mixture models for real-time tracking. IEEE Confence on Computer Vision and Pattern Recognition.1999. 23-25.
[26] Goujou, E.and Miteran,J. Human detection with video surveillance system. Proceedings of the IEEE IECON 21 st International Conference.1995, vol.2. 747-757.
[27]张孝令等,贝叶斯动态模型及其预测,山东科技出版社,1992.1.
[28] A.Doucet, S.Godsill, and C.Andrieu. On sequential Monte Carlo sampling methods for Bayesian filtering[J]. Statisit.Computer, 2000, vol.10. 197-208.
[29] A.Doucet, J.de Freitas, N.Gordon. Sequential Monte Carlo Methods in Practice[M]. New-York: Springer-Verlag. 2001.
[30]盛骤等,概率论与数理统计[M].北京:高等教育出版社,1989.
[31]方再根.计算机模拟和蒙特卡罗方法论.北京:北京工业学院出版社,1988. 88-96.
[32]徐钟济.蒙特卡罗方法.上海:上海科学技术出版社,1985. 123-143.
[33]李庆扬,关治,白峰杉.数值计算原理.[M].北京:清华大学出版社,2000.
[34] A.Doucet, S.Godsill, Andrieu. On sequential Monte Carlo sampling methods for Bayesian filtering[J]. Statist. Computer, 2000, 10. 197-208.
[35]徐进.视监控系统中的活动目标检测与跟踪算法的研究.山东大学.2004.05.
[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]谌剑,严平,张静远.权值优化组合粒子滤波算法研究.计算机工程与应用. 2009,45(24). 33-35.
[38]何立,孙函,黄永等.MODIS 1B数据的重采样方法研究.遥感应用.2007, (3). 37-43.
[39]吴刚,唐振民,耿峰等.引入前帧加权采样的粒子滤波目标跟踪算法.计算机科学.2009, 36(8). 215-216,246.
[40]常文革,宋千,梁甸农.重采样差值技术实现与应用.系统工程与电子技术.2000,22(5).87-89.
[41]裴福俊,孙新蕊,刘红云等.均匀重采样粒子滤波器在SINS初始对准中的应用.北京工业大学学报.2008, 34(6). 567-571.
[42]高显忠,赵伟,侯中喜.粒子滤波改进算法研究.弹箭与制导学报. 2009, 29(3). 241-244.
[43] J.S.Liu, R.Chen. Sequential Monte Carlo Methods of Dynamic System. Journal of American Statistician.1998. 1032-1044.
[44] Y.Chen, Y.Rui and T.S.Huang. JPDAF Based HMM for Real-time Contour Tracking. In: Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recongnition. Kauai, Hawaii. 2001,1.543-550.
[45] N.J.Gordon, D.J.Salmond, and A.F.M.Smith. Novel Approach to Nonlinear/Non-Gaussian Bayesian State Estimation. In: IEEE Proceedings on Radar and Signal Processing.1993, 140(2). 107-111.
[46] M.S.Arulampalam, S.Maskell and N.Gordon, etal. A Tutorial on Particle Filters for Online Nonlinear/Non-Gaussian Bayesian Tracking. IEEE Transactions Signal Processing. 2002, 50(2). 174-188.
[47] W.Grimson, C.Stauffer, R.Romano. Using adaptive tracking to classify and Monitor activities in a site. IEEE Confence on Computer Vision and Pattern Recognition. 1998. 22-29.
[48] D.Murray, A.Basu.Motion tracking with an active camera. IEEE Trans. 1994, 16(5). 449-459.
[49] S.Ribaric, G.Adrinek, S.Segvic. Real-time active visual tracking system. IEEE Melecon, 2004, 12(15).231-234.
[50] I.Cohen, G.Medioni. Detecting and tracking moving objects for video surveillance. IEEE Conference on Computer Vision and Pattern Recognition.1999.23-25.
[51] Jianbo Shi, Carlo Tomasi. Good features to track. IEEE Conference on Computer Vision and Pattern Recongition.1994. 593-600.
[52]袁天鑫,最佳估计原理.北京:国防工业出版社,1980.
[53]陈建良.自回归模型的参数估计.山东工业大学学报.1996, 26(3). 226-230.
[54]范春宇.视频跟踪技术研究.西安:西安电子科技大学硕士论文,2007.
[55] M.Sanjeev Arulampalam, Simon Maskell, Neil Gordon. A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking [J]. IEEE Transactions on Signal Processing, 2002, 50(2). 174-188.
[2] MeKennaT. Video surveillance and human activity recognition for anti-terrors and force protection. IEEE Conference on Advanced Video and signal Based Surveillance. 2003. 22-23.
[3] E.Montseny, J.Frau. specialized processors for real-time image analysis.Workshop Proceedings, Barcelona, Spain, 1991.9.
[4]赵荣椿.数字图像处理导论.西安:西北工业大学出版社, 2000.
[5] RobertT.Collins, Alan J.Lipton. A System for Video Surveillance and Monitoring. Carnegie Mellon University. 2000.
[6]中国科学院自动化研究所模式识别国家重点实验室视觉监控研究组,智能视觉监控,http;//www.sinosurveillance.com/.
[7] Haritaoglu I, Harwood.D, etal. Real-time Surveillance of people and Their Activities. IEEE Transactions on Pattern Analysis and Machine Intelligence.2000, 22(8). 809-830.
[8] Lipto and H.Fujiyoshi, etc. Moving Target Classification and Tracking from Real-time Video. IEEE workshop Application of Computer Vision.1998, 10. 8-14.
[9] Foresti, GL. Object recognition and tracking for remote video surveillance. IEEE Transaction of Circuits and Systems for Video Technology.1999, 9(7). 1045-1062.
[10] A.Murat Tekalp. Digital video processing. Tsinghua University Press. 1998. 380.
[11] Horn B, Schunch B.Detemining optical flow. Artificial Intelligence.1981.185-203.
[12] Haritaoglu and Harwood, Davis. Real-time system for detecting and tracking People. Proceedings of Third IEEE International Conference of Automatic Face and Gesture Recoginition.1998. 222-227.
[13] Stringa, E.and Regazzoni. Real-time video-shot detection for scene surveillance applications. IEEE Transactions of Image Processing. 2000, 1, 9(1). 19-69.
[14] Robert T.Collins, Alan J.Lipto. A System for Video Surveillance and Monitoring. Carnegie Mellon University. 2000.
[15] Haritaoglu I, Harwood.D. Real-time Surveillance of People and Their Activities. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(8).809-830.
[16]魏志强,纪筱鹏,冯业伟.基于自适应背景图像更新的运动目标检测方法.电子学报.2005, 33(12). 2261-2264.
[17] Gutchess D, Trajkovics M, Cohen-Solal E. A background model initialization algorithm for video surveillance .Computer Vision.2001.ICCV 2001. Proceedings Eighth IEEE International Conference. 2001, 6, Vol.1. 733-740.
[18] Jwu-Sheng Hu, Tzung-Min Su, Shr-Chi Jeng. Robust Background Subtraction with Shadow and Highlight. Removal for Indoor Surveillance Intelligent Robots and Systems, 2006 IEEE/RSJ International Conference. 2006, 10. 4545-4550.
[19] Cezar Silveira, Jacques J, Rosito Jung C. A Background Subtraction Model Adapted to Illumination Changes. Image Processing, 2006 IEEE International Conference. 2006, 10. 1817-1820.
[20]郑江滨.视频监视方法研究.西北工业大学博士学位论文.2002. 21-22.
[21]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.
[22] C.Wren, A. Azarbayejani, T.Darrel. Real-time Tracking of the Human Body. IEEE Trans. 1997, 6, 19(7). 780-785.
[23]郑江滨、张艳宁、冯大淦.视频监视中运动目标的检测与跟踪算法.系统工程与电子技术.2002, 24(10). 34-37.
[24]张文杰,戚飞虎,江卓军.实时视频监控系统中运动目标检测和跟踪的一种实用方法.上海交通大学学报.2002,36(12).1837-1840.
[25] C.Stauffer, W.E.L.Gramson. Adaptive background mixture models for real-time tracking. IEEE Confence on Computer Vision and Pattern Recognition.1999. 23-25.
[26] Goujou, E.and Miteran,J. Human detection with video surveillance system. Proceedings of the IEEE IECON 21 st International Conference.1995, vol.2. 747-757.
[27]张孝令等,贝叶斯动态模型及其预测,山东科技出版社,1992.1.
[28] A.Doucet, S.Godsill, and C.Andrieu. On sequential Monte Carlo sampling methods for Bayesian filtering[J]. Statisit.Computer, 2000, vol.10. 197-208.
[29] A.Doucet, J.de Freitas, N.Gordon. Sequential Monte Carlo Methods in Practice[M]. New-York: Springer-Verlag. 2001.
[30]盛骤等,概率论与数理统计[M].北京:高等教育出版社,1989.
[31]方再根.计算机模拟和蒙特卡罗方法论.北京:北京工业学院出版社,1988. 88-96.
[32]徐钟济.蒙特卡罗方法.上海:上海科学技术出版社,1985. 123-143.
[33]李庆扬,关治,白峰杉.数值计算原理.[M].北京:清华大学出版社,2000.
[34] A.Doucet, S.Godsill, Andrieu. On sequential Monte Carlo sampling methods for Bayesian filtering[J]. Statist. Computer, 2000, 10. 197-208.
[35]徐进.视监控系统中的活动目标检测与跟踪算法的研究.山东大学.2004.05.
[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]谌剑,严平,张静远.权值优化组合粒子滤波算法研究.计算机工程与应用. 2009,45(24). 33-35.
[38]何立,孙函,黄永等.MODIS 1B数据的重采样方法研究.遥感应用.2007, (3). 37-43.
[39]吴刚,唐振民,耿峰等.引入前帧加权采样的粒子滤波目标跟踪算法.计算机科学.2009, 36(8). 215-216,246.
[40]常文革,宋千,梁甸农.重采样差值技术实现与应用.系统工程与电子技术.2000,22(5).87-89.
[41]裴福俊,孙新蕊,刘红云等.均匀重采样粒子滤波器在SINS初始对准中的应用.北京工业大学学报.2008, 34(6). 567-571.
[42]高显忠,赵伟,侯中喜.粒子滤波改进算法研究.弹箭与制导学报. 2009, 29(3). 241-244.
[43] J.S.Liu, R.Chen. Sequential Monte Carlo Methods of Dynamic System. Journal of American Statistician.1998. 1032-1044.
[44] Y.Chen, Y.Rui and T.S.Huang. JPDAF Based HMM for Real-time Contour Tracking. In: Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recongnition. Kauai, Hawaii. 2001,1.543-550.
[45] N.J.Gordon, D.J.Salmond, and A.F.M.Smith. Novel Approach to Nonlinear/Non-Gaussian Bayesian State Estimation. In: IEEE Proceedings on Radar and Signal Processing.1993, 140(2). 107-111.
[46] M.S.Arulampalam, S.Maskell and N.Gordon, etal. A Tutorial on Particle Filters for Online Nonlinear/Non-Gaussian Bayesian Tracking. IEEE Transactions Signal Processing. 2002, 50(2). 174-188.
[47] W.Grimson, C.Stauffer, R.Romano. Using adaptive tracking to classify and Monitor activities in a site. IEEE Confence on Computer Vision and Pattern Recognition. 1998. 22-29.
[48] D.Murray, A.Basu.Motion tracking with an active camera. IEEE Trans. 1994, 16(5). 449-459.
[49] S.Ribaric, G.Adrinek, S.Segvic. Real-time active visual tracking system. IEEE Melecon, 2004, 12(15).231-234.
[50] I.Cohen, G.Medioni. Detecting and tracking moving objects for video surveillance. IEEE Conference on Computer Vision and Pattern Recognition.1999.23-25.
[51] Jianbo Shi, Carlo Tomasi. Good features to track. IEEE Conference on Computer Vision and Pattern Recongition.1994. 593-600.
[52]袁天鑫,最佳估计原理.北京:国防工业出版社,1980.
[53]陈建良.自回归模型的参数估计.山东工业大学学报.1996, 26(3). 226-230.
[54]范春宇.视频跟踪技术研究.西安:西安电子科技大学硕士论文,2007.
[55] M.Sanjeev Arulampalam, Simon Maskell, Neil Gordon. A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking [J]. IEEE Transactions on Signal Processing, 2002, 50(2). 174-188.