智能监控中的运动目标检测与跟踪算法研究
摘要
在智能视频监控系统中,目标检测和跟踪的实质是通过对视频序列处理来提取运动目标,并得到目标在每帧视频中的位置、大小和运动速度等信息,起着承上启下的作用。其性能对智能视频监控系统的有效性会产生直接的影响,是智能视频监控系统极为重要的组成部分。
本文首先介绍了Meanshift和Camshift算法的基本原理,给出了Camshift算法的搜索过程以及Meanshift和Camshift在视频目标跟踪中的应用。然后针对运动目标速度过快的情况,研究了Kalman滤波的相关知识。介绍了Kalman滤波器的基本原理与工作流程,并将Kalman滤波器应用到目标跟踪领域。在此基础上提出了与Camshift算法相结合的改进算法。对于背景颜色和目标相近或是背景较为复杂的场景,本文采取了前景提取的方法进行处理。这里利用Codebook背景建模的方法提取出运动目标前景,将此二值化图像作为mask模板来去除背景信息的各种干扰,获得所需要的前景信息。在此基础上为了完善跟踪效果,适应更多的应用场合,提出了利用改进的LBP局部二值化模式算法进行纹理信息的提取,与色调分量一起作为描述运动目标的特征,提高了目标跟踪的鲁棒性及适用范围。最后综合以上算法,提出了本文的最终算法,即融入了前景提取与运动估计的改进目标跟踪算法。实验证明,本文的算法在原有算法的基础上大大提高了跟踪性能。
本文的主要创新点为结合Kalman滤波器解决运动目标速度过快问题,融入Codebook解决背景与前景区分不大的问题以及为提高跟踪精度加入纹理特征。
In the intelligent video surveillance system, target detection and tracking based on video sequence is processed to extract the moving target, and obtain the target in each frame of the video in the location, size and velocity information. It plays an important role. The performance of target detection and tracking will have a direct impact on the validity of intelligent video surveillance system. It is an extremely important part of intelligent video surveillance system.
This paper first describes the basic principle of the Meanshift and Camshift algorithm. The search process of Camshift algorithm is given as well as the application of the Meanshift and Camshift in video target tracking. Then Kalman filter is researched according to the condition of target moving so fast. The basic principle and work flow of the Kalman filter is introduced. Also Kalman filter is applied to target tracking. On this foundation this paper puts forward an improved algorithm combined with Camshift. Considering that the color of background is similar to the target or background is complex, this paper adopts the foreground extraction. Here Codebook background modeling method is used to extract the moving object. This binary image is used as a mask to remove the interference of background information and to gain the various prospects information needed. On this basis in order to improve the tracking performance and adapt to more applications, texture information is extracted by using improved local binary pattern algorithm. The tone component and texture information are both used as descriptions of the characteristics of moving target so as to improve the target track robustness and scope of application. Finally the final algorithm is presented in this paper based on the above algorithms, namely an improved target tracking algorithm combined with the foreground extraction and motion estimation. The experiments prove that the algorithm in this paper greatly improves the performance of tracking targets.
The main innovations of this paper are combining Camshift with the Kalman filter to solve the problem of the target moving so fast, combining Codebook to distinguish background with foreground and adding texture feature to improve the precision of tracking.
本文首先介绍了Meanshift和Camshift算法的基本原理,给出了Camshift算法的搜索过程以及Meanshift和Camshift在视频目标跟踪中的应用。然后针对运动目标速度过快的情况,研究了Kalman滤波的相关知识。介绍了Kalman滤波器的基本原理与工作流程,并将Kalman滤波器应用到目标跟踪领域。在此基础上提出了与Camshift算法相结合的改进算法。对于背景颜色和目标相近或是背景较为复杂的场景,本文采取了前景提取的方法进行处理。这里利用Codebook背景建模的方法提取出运动目标前景,将此二值化图像作为mask模板来去除背景信息的各种干扰,获得所需要的前景信息。在此基础上为了完善跟踪效果,适应更多的应用场合,提出了利用改进的LBP局部二值化模式算法进行纹理信息的提取,与色调分量一起作为描述运动目标的特征,提高了目标跟踪的鲁棒性及适用范围。最后综合以上算法,提出了本文的最终算法,即融入了前景提取与运动估计的改进目标跟踪算法。实验证明,本文的算法在原有算法的基础上大大提高了跟踪性能。
本文的主要创新点为结合Kalman滤波器解决运动目标速度过快问题,融入Codebook解决背景与前景区分不大的问题以及为提高跟踪精度加入纹理特征。
In the intelligent video surveillance system, target detection and tracking based on video sequence is processed to extract the moving target, and obtain the target in each frame of the video in the location, size and velocity information. It plays an important role. The performance of target detection and tracking will have a direct impact on the validity of intelligent video surveillance system. It is an extremely important part of intelligent video surveillance system.
This paper first describes the basic principle of the Meanshift and Camshift algorithm. The search process of Camshift algorithm is given as well as the application of the Meanshift and Camshift in video target tracking. Then Kalman filter is researched according to the condition of target moving so fast. The basic principle and work flow of the Kalman filter is introduced. Also Kalman filter is applied to target tracking. On this foundation this paper puts forward an improved algorithm combined with Camshift. Considering that the color of background is similar to the target or background is complex, this paper adopts the foreground extraction. Here Codebook background modeling method is used to extract the moving object. This binary image is used as a mask to remove the interference of background information and to gain the various prospects information needed. On this basis in order to improve the tracking performance and adapt to more applications, texture information is extracted by using improved local binary pattern algorithm. The tone component and texture information are both used as descriptions of the characteristics of moving target so as to improve the target track robustness and scope of application. Finally the final algorithm is presented in this paper based on the above algorithms, namely an improved target tracking algorithm combined with the foreground extraction and motion estimation. The experiments prove that the algorithm in this paper greatly improves the performance of tracking targets.
The main innovations of this paper are combining Camshift with the Kalman filter to solve the problem of the target moving so fast, combining Codebook to distinguish background with foreground and adding texture feature to improve the precision of tracking.
引文
[1]王亮,胡卫明,谭铁牛人运动的视觉分析综述[J]计算机学报2002 25(3)225-237
[2]Frederick Tung John S. Zelek David Clausi Efficient Target Recovery Using STAGE for Mean-shift Tracking[J] Canadian Conference on Computer and Robot Vision 2009 16-22
[3]罗的国,韩萍Meanshif目标跟踪的两级更新自适应窗宽算法第二十五届中国(天津)2011 IT、网络、信息技术、电子、仪器仪表创新学术会议2011 5-9
[4]李琳嶙,费树岷,金立左基于Mean shift的改进目标跟踪算法中国智能自动化会议论文集2009 1383-1388
[5]胡波基于Mean Shift算法和卡尔曼滤波的视频目标跟踪技术研究[学位论文]宁波宁波大学2010 37-38
[6]田洪宁基于运动目标检测跟踪的智能监控系统设计实现[学位论文]北京北京工业大学2011 23-24
[7]P. Hidayatullah H. Konik CAMSHIFT Improvement on Multi-Hue and Multi-Object Tracking [J] 2011 International Conference on Electrical Engineering and Informatics Bandung Indonesia 2011
[8]Shengluan Huang Jingxin Hong. Moving Object Tracking System Based On Camshift And Kalman Filter[J] IEEE 2011 1423-1426
[9]徐萧萧,杨浩,陈宗海一种基于改进Camshift的目标跟踪算法[J]系统仿真技术及其应用2009 11(5)851-854
[10]Madhur Mehta Chandni Goyal M.C. Srivastava R.C. Jain. Real Time Object Detection and Tracking:Histogram Matching and Kalman Filter Approach[J] IEEE 2010 796-801
[11]WANG Xiangyu LI Xiujuan. The Study of Moving Target Tracking Based on Kalman-CamShift in the Video[J] IEEE 2010
[12]李阳 智能监控系统中运动目标检测与跟踪算法研究[学位论文]哈尔滨 哈尔滨工程大学硕士学位论文2010 43-46
[13]查宇飞,张育,毕笃彦一种基于粒子滤波的自适应运动目标跟踪方法[J]第十二届全国图象图形学学术会议2005 333-338
[14]Hyung-Bok Kim Kwee-Bo Sim A Particular Object Tracking in an Environment of Multiple Moving Objects[J] International Conference on Control Automation and Systems 2010 1053-1056
[15]李军,张华,单梁 一种基于Mean shift和粒子滤波的综合目标跟踪算法[J]南京理工大 学学报2009 167(33)220-225
[16]Da Tang Yu-Jin Zhang Combining Mean-shift and Particle Filter for Object Tracking[J] Sixth International Conference on Image and Graphics 2011 771-776
[17]熊亮,刘伟铭基于背景Codebook模型的前景检测算法[J1科学技术与工程2010 9(10)2118-2121
[18]Atif Ilyas Mihaela Scuturici Serge Miguet Real Time Foreground-Background Segmentation Using a Modified Codebook Model[J] 2009 Advanced Video and Signal Based Surveillance 2009 454-459
[19]Kyungnam Kim Thanarat H. Chalidabhongse David Harwood Larry Davis Real-time foreground-background segmentation using codebook model [J] SCIENCE DIRECT 2005
[20]Amit Pal Gerald Schaefer M. Emre Celebi ROBUST CODEBOOK-BASED VIDEO BACKGROUND SUBTRACTION[J] International Conference on Acoustics Speech and Signal Processing 2010 1146-1149
[21]Qiu Tu Yiping Xu Manli Zhou Box-based Codebook Model for Real-time Objects Detection[J] Proceedings of the 7th World Congress on Intelligent Control and Automation 2008 7621-7625
[22]Jiu XU Ning JIANG Satoshi GOTO Block-based Codebook Model with Oriented-Gradient Feature for Real-time Foreground Detection[J] MMSP 2011
[23]I-Ting Sun Shih-Chung Hsu and Chung-Lin Huang A HYBRID CODEBOOK BACKGROUND MODEL FOR BACKGROUND SUBTRACTION[J] Structural Insulated Panel Association 2011 96-101
[24]JIFENG NING LEI ZHANG DAVID ZHANG CHENGKE WU ROBUST OBJECT TRACKING USING JOINT COLOR-TEXTURE HISTOGRAM[J] International Journal of Pattern Recognition and Artificial Intelligence 2009 7(23)1245-1263
[25]吴良健基于Cam shift和Kalman滤波的多目标跟踪算法研究[学位论文]武汉中南民族大学2010
[26]陈禹基于改进的CAMSHIFT算法的视频目标跟踪系统的研究与设计[学位论文]太原太原理工大学2012
[27]T. Kim S. Lee J. Paik Combined shape and feature-based video analysis and its application to non-rigid object tracking[J] The Institution of Engineering and Technology 2011 5(1)87-100
[28]Yong Chu Hong Zhu Dong Wang A Codebook-based Video Moving Objects Detecting Method[J] International Conference on Information Acquisition 2006
[29]Zulfiqar Hasan Khan Irene Yu-Hua Gu Andrew Backhouse JOINT PARTICLE FILTERS AND MULTI-MODE ANISOTROPIC MEAN SHIFT FOR ROBUST TRACKING OF VIDEO OBJECTS WITH PARTITIONED AREAS[J] ICIP 2009
[30]Mahnaz Janipoor Deilamani Rahebe Niaraki Asli Moving Object Tracking Based On Mean Shift Algorithm and Features Fusion[J] IEEE 2009
[31]Alper Yilmaz Object Tracking by Asymmetric Kernel Mean Shift with Automatic Scale and Orientation Selection[J] IEEE 2007
[32]Katharina Quast Andre Kaup SHAPE ADAPTIVE MEAN SHIFT OBJECT TRACKING USING GAUSSIAN MIXTURE MODELS[J]
[33]Jung Uk Chol Seung Hun Jin Xuan Dai Phaml Jae Wook Jeon Object Tracking Circuit using Particle Filter with Multiple Features[J] SICE-ICASEInternationalJointConference 2006
[34]Zulfiqar Hasan Khan Irene Yu-Hua Gu Andrew G Backhouse TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY[J] IEEE 2011 21(1)
[2]Frederick Tung John S. Zelek David Clausi Efficient Target Recovery Using STAGE for Mean-shift Tracking[J] Canadian Conference on Computer and Robot Vision 2009 16-22
[3]罗的国,韩萍Meanshif目标跟踪的两级更新自适应窗宽算法第二十五届中国(天津)2011 IT、网络、信息技术、电子、仪器仪表创新学术会议2011 5-9
[4]李琳嶙,费树岷,金立左基于Mean shift的改进目标跟踪算法中国智能自动化会议论文集2009 1383-1388
[5]胡波基于Mean Shift算法和卡尔曼滤波的视频目标跟踪技术研究[学位论文]宁波宁波大学2010 37-38
[6]田洪宁基于运动目标检测跟踪的智能监控系统设计实现[学位论文]北京北京工业大学2011 23-24
[7]P. Hidayatullah H. Konik CAMSHIFT Improvement on Multi-Hue and Multi-Object Tracking [J] 2011 International Conference on Electrical Engineering and Informatics Bandung Indonesia 2011
[8]Shengluan Huang Jingxin Hong. Moving Object Tracking System Based On Camshift And Kalman Filter[J] IEEE 2011 1423-1426
[9]徐萧萧,杨浩,陈宗海一种基于改进Camshift的目标跟踪算法[J]系统仿真技术及其应用2009 11(5)851-854
[10]Madhur Mehta Chandni Goyal M.C. Srivastava R.C. Jain. Real Time Object Detection and Tracking:Histogram Matching and Kalman Filter Approach[J] IEEE 2010 796-801
[11]WANG Xiangyu LI Xiujuan. The Study of Moving Target Tracking Based on Kalman-CamShift in the Video[J] IEEE 2010
[12]李阳 智能监控系统中运动目标检测与跟踪算法研究[学位论文]哈尔滨 哈尔滨工程大学硕士学位论文2010 43-46
[13]查宇飞,张育,毕笃彦一种基于粒子滤波的自适应运动目标跟踪方法[J]第十二届全国图象图形学学术会议2005 333-338
[14]Hyung-Bok Kim Kwee-Bo Sim A Particular Object Tracking in an Environment of Multiple Moving Objects[J] International Conference on Control Automation and Systems 2010 1053-1056
[15]李军,张华,单梁 一种基于Mean shift和粒子滤波的综合目标跟踪算法[J]南京理工大 学学报2009 167(33)220-225
[16]Da Tang Yu-Jin Zhang Combining Mean-shift and Particle Filter for Object Tracking[J] Sixth International Conference on Image and Graphics 2011 771-776
[17]熊亮,刘伟铭基于背景Codebook模型的前景检测算法[J1科学技术与工程2010 9(10)2118-2121
[18]Atif Ilyas Mihaela Scuturici Serge Miguet Real Time Foreground-Background Segmentation Using a Modified Codebook Model[J] 2009 Advanced Video and Signal Based Surveillance 2009 454-459
[19]Kyungnam Kim Thanarat H. Chalidabhongse David Harwood Larry Davis Real-time foreground-background segmentation using codebook model [J] SCIENCE DIRECT 2005
[20]Amit Pal Gerald Schaefer M. Emre Celebi ROBUST CODEBOOK-BASED VIDEO BACKGROUND SUBTRACTION[J] International Conference on Acoustics Speech and Signal Processing 2010 1146-1149
[21]Qiu Tu Yiping Xu Manli Zhou Box-based Codebook Model for Real-time Objects Detection[J] Proceedings of the 7th World Congress on Intelligent Control and Automation 2008 7621-7625
[22]Jiu XU Ning JIANG Satoshi GOTO Block-based Codebook Model with Oriented-Gradient Feature for Real-time Foreground Detection[J] MMSP 2011
[23]I-Ting Sun Shih-Chung Hsu and Chung-Lin Huang A HYBRID CODEBOOK BACKGROUND MODEL FOR BACKGROUND SUBTRACTION[J] Structural Insulated Panel Association 2011 96-101
[24]JIFENG NING LEI ZHANG DAVID ZHANG CHENGKE WU ROBUST OBJECT TRACKING USING JOINT COLOR-TEXTURE HISTOGRAM[J] International Journal of Pattern Recognition and Artificial Intelligence 2009 7(23)1245-1263
[25]吴良健基于Cam shift和Kalman滤波的多目标跟踪算法研究[学位论文]武汉中南民族大学2010
[26]陈禹基于改进的CAMSHIFT算法的视频目标跟踪系统的研究与设计[学位论文]太原太原理工大学2012
[27]T. Kim S. Lee J. Paik Combined shape and feature-based video analysis and its application to non-rigid object tracking[J] The Institution of Engineering and Technology 2011 5(1)87-100
[28]Yong Chu Hong Zhu Dong Wang A Codebook-based Video Moving Objects Detecting Method[J] International Conference on Information Acquisition 2006
[29]Zulfiqar Hasan Khan Irene Yu-Hua Gu Andrew Backhouse JOINT PARTICLE FILTERS AND MULTI-MODE ANISOTROPIC MEAN SHIFT FOR ROBUST TRACKING OF VIDEO OBJECTS WITH PARTITIONED AREAS[J] ICIP 2009
[30]Mahnaz Janipoor Deilamani Rahebe Niaraki Asli Moving Object Tracking Based On Mean Shift Algorithm and Features Fusion[J] IEEE 2009
[31]Alper Yilmaz Object Tracking by Asymmetric Kernel Mean Shift with Automatic Scale and Orientation Selection[J] IEEE 2007
[32]Katharina Quast Andre Kaup SHAPE ADAPTIVE MEAN SHIFT OBJECT TRACKING USING GAUSSIAN MIXTURE MODELS[J]
[33]Jung Uk Chol Seung Hun Jin Xuan Dai Phaml Jae Wook Jeon Object Tracking Circuit using Particle Filter with Multiple Features[J] SICE-ICASEInternationalJointConference 2006
[34]Zulfiqar Hasan Khan Irene Yu-Hua Gu Andrew G Backhouse TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY[J] IEEE 2011 21(1)