基于视频的车辆检测与跟踪技术研究
|
摘要
随着现代社会经济的飞速发展,有效的城市道路交通管理在人们的经济、社会活动中的重要性日益显著。因此,深入研究解决城市交通问题就有着极为重要的现实意义。其中,智能交通系统(ITS)的研究被提到了重要位置,运动车辆的有效检测和准确跟踪是现代智能交通研究的核心部分,而利用计算机视觉技术进行交通状况检测与信息采集作为智能交通系统领域中的一个重要课题,其中最基础的部分是运动车辆的检测与跟踪,这成为许多国家的研究热点。
本文针对基于视频的车辆检测与跟踪系统中的相关问题进行了较为具体的研究,在总结和分析现有的车辆检测和跟踪技术的基础上,针对车辆的检测、车辆识别与跟踪、阴影去除等方面进行了分析研究与部分改进。实验结果表明给出的算法是有效的。
本文主要包括的研究工作如下:
1、基于视频序列的车辆检测。在这一部分,本文提出了一种改进的基于视频的车辆检测流程。通过现有检测方法的分析比较,对基于视频的车辆目标提取的常用方法进行了综述,其中包括帧间差分法、背景差分法和边缘检测法,图像分割部分,重点介绍了基于区域的分割算法,并通过实验验证了帧差法改进流程后的检测结果。
2、基于视频序列的车辆跟踪。在这一部分,本文主要提出了一种新的基于区域的车辆跟踪算法。在基于区域的车辆跟踪中,为了提高匹配精度,经常采用多个特征进行匹配。基于简单有效的考虑,本文所提出的区域跟踪算法选择运动车辆的质心位置及目标大小作为匹配特征。实验证明,算法具有可行性,可运用于实时环境,检测与跟踪结果也比较令人满意。
3、车辆检测过程中的阴影去除。在这一部分,本文主要针对视频车辆检测系统中的关键步骤——视频检测中的阴影去除进行深入研究,在分析了阴影产生的原因和阴影的特点之后,综合利用灰度图像及其差分后的二值化图像,提出了一种基于背景差分的检测与去除阴影的新算法。实验证明,该算法能够较好地去除运动车辆的阴影,同时保留比较完整的车辆目标信息,这为准确提取运动车辆目标奠定了坚实的基础。
With the rapid development of modern social-economic, the importance of effective urban road traffic management has become more and more notable in our daily life. Therefore, the deep research on urban transportation problems has very important practical significance. The research of ITS, intelligent transportation system, comes to an issue of consequence, and the effective detection and accurate tracking of moving vehicles are the core of this research. The traffic condition detection and information collection by computer vision system has become a hot topic in the field of ITS, in which the most basic issue is the detection and tracking of moving vehicles.
This paper makes specific studies on video-based vehicle detection and tracking system, based on summary and analyzes the existing technology on vehicle detection and tracking, some improvements are given for vehicle recognition and tracking, and shadow elimination during vehicle recognition is studied. Experimental results indicate that our algorithm is effective.
The main contents of this paper are organized as follows:
1、Vehicle detection based on video sequence. In this section, we present an improved video-based vehicle detection process. Through analysis and comparison of existing detection methods, we reviewed vehicle-based commonly used video object extraction methods, including frame difference method, background difference method and edge detection method, focusing on segmentation algorithm based on region. The experiments show that the improved frame difference is effective.
2、Vehicle tracking based on video sequence. In this section, this paper proposes an innovative region-based vehicle tracking algorithm. In region-based vehicle tracking, multiple features match is usually used to improve the matching accuracy. For simple and effective consideration, the proposed tracking algorithm is this paper selects the regional center location of mass and target size of the moving vehicle as the matching feature. Experimental results indicate that the algorithm is feasible and can apply to real-time environment. And the experimental results of tracking are satisfactory.
3、Shadow elimination during vehicle detection based on video sequence. In this section, we focus on shadow elimination, which is a key step, in video-based vehicle detection system. On analyzing the causes and characteristics of the shadow, with the comprehensive utilization of gray image and binary image after difference, we present a novel shadow detection and elimination algorithm based on the background. Experiments show that our algorithm can remove the shadow of moving vehicle effectively and still reserve majority of target information, which lay a stable ground for extracting moving vehicle target.
本文针对基于视频的车辆检测与跟踪系统中的相关问题进行了较为具体的研究,在总结和分析现有的车辆检测和跟踪技术的基础上,针对车辆的检测、车辆识别与跟踪、阴影去除等方面进行了分析研究与部分改进。实验结果表明给出的算法是有效的。
本文主要包括的研究工作如下:
1、基于视频序列的车辆检测。在这一部分,本文提出了一种改进的基于视频的车辆检测流程。通过现有检测方法的分析比较,对基于视频的车辆目标提取的常用方法进行了综述,其中包括帧间差分法、背景差分法和边缘检测法,图像分割部分,重点介绍了基于区域的分割算法,并通过实验验证了帧差法改进流程后的检测结果。
2、基于视频序列的车辆跟踪。在这一部分,本文主要提出了一种新的基于区域的车辆跟踪算法。在基于区域的车辆跟踪中,为了提高匹配精度,经常采用多个特征进行匹配。基于简单有效的考虑,本文所提出的区域跟踪算法选择运动车辆的质心位置及目标大小作为匹配特征。实验证明,算法具有可行性,可运用于实时环境,检测与跟踪结果也比较令人满意。
3、车辆检测过程中的阴影去除。在这一部分,本文主要针对视频车辆检测系统中的关键步骤——视频检测中的阴影去除进行深入研究,在分析了阴影产生的原因和阴影的特点之后,综合利用灰度图像及其差分后的二值化图像,提出了一种基于背景差分的检测与去除阴影的新算法。实验证明,该算法能够较好地去除运动车辆的阴影,同时保留比较完整的车辆目标信息,这为准确提取运动车辆目标奠定了坚实的基础。
With the rapid development of modern social-economic, the importance of effective urban road traffic management has become more and more notable in our daily life. Therefore, the deep research on urban transportation problems has very important practical significance. The research of ITS, intelligent transportation system, comes to an issue of consequence, and the effective detection and accurate tracking of moving vehicles are the core of this research. The traffic condition detection and information collection by computer vision system has become a hot topic in the field of ITS, in which the most basic issue is the detection and tracking of moving vehicles.
This paper makes specific studies on video-based vehicle detection and tracking system, based on summary and analyzes the existing technology on vehicle detection and tracking, some improvements are given for vehicle recognition and tracking, and shadow elimination during vehicle recognition is studied. Experimental results indicate that our algorithm is effective.
The main contents of this paper are organized as follows:
1、Vehicle detection based on video sequence. In this section, we present an improved video-based vehicle detection process. Through analysis and comparison of existing detection methods, we reviewed vehicle-based commonly used video object extraction methods, including frame difference method, background difference method and edge detection method, focusing on segmentation algorithm based on region. The experiments show that the improved frame difference is effective.
2、Vehicle tracking based on video sequence. In this section, this paper proposes an innovative region-based vehicle tracking algorithm. In region-based vehicle tracking, multiple features match is usually used to improve the matching accuracy. For simple and effective consideration, the proposed tracking algorithm is this paper selects the regional center location of mass and target size of the moving vehicle as the matching feature. Experimental results indicate that the algorithm is feasible and can apply to real-time environment. And the experimental results of tracking are satisfactory.
3、Shadow elimination during vehicle detection based on video sequence. In this section, we focus on shadow elimination, which is a key step, in video-based vehicle detection system. On analyzing the causes and characteristics of the shadow, with the comprehensive utilization of gray image and binary image after difference, we present a novel shadow detection and elimination algorithm based on the background. Experiments show that our algorithm can remove the shadow of moving vehicle effectively and still reserve majority of target information, which lay a stable ground for extracting moving vehicle target.
引文
[1]陆化普,殷亚峰,史其信.新一代道路交通系统.中国公路学报,1997,10(6):70~76.
[2] S.Gupte, O.Masoud, RFK Martin,“Detection and classification of vehicles”, IEEE Transactions on Intelligent Transportation Systems,2002,3 1:37～47.
[3]张伟.基于视觉的运动车辆检测与跟踪[D].上海交通大学工学博士学位论文,2007.
[4] Texas Instruments. SPRS200F, TMS320DM642 Video/Imaging Fixed– Point Digital Signal Processor Data Manual[S].2004.
[5] Qiang Ji, Xiaojie Yang. Real-time eye, gaze, and face pose tracking for monitoring driver vigilance. Real-Time Imaging, 2002, 8:357~377.
[6] Jose Melo, Andrew Naftel, Alexandre Bernardin, et al. Detection and Classification of Highway Lanes Using Vehicle Motion Trajectories [J]. IEEE Transactions on Intelligent Transportation Systems, 2006,7(2):188~200.
[7]朱志刚,徐光佑,扬波.自动交通监测系统的二维时空图像方法.中国图像图形学报,1996,1(2):101~10.
[8]吴裙文,张学工.应用小波变换和PAC进行车辆的静态图像检测[J].清华大学学报,2002,42(11):1560~1564.
[9]周志宇,汪亚明,黄文清.基于遗传算法的指数熵在车辆运动检测中的应用[J].浙江工程学院学报,2003,20(9):202~205.
[10]肖旺新,张雪,黄卫.交通视频图像的多尺度自适应阈值边缘检测[J].土木工程学报,2003,37(6):89~94.
[11]周欣.基于方向分形维的高速公路车辆边缘检测研究[J].信号处理,2004,20(3):258~262.
[12] ANZAIY, KATOT, HIGASHIKUBOM, et al. Development of an integrated video imaging vehicle detector [J].SEI Technical Re2view,2005,59 (6).
[13] Liu G Y, Chen R, Deng Y et al. Video based 3D human body motion capture[J]. Journal of Computer-Aided Design and Computer Graphics, 2006,18(l):82~88.
[14]车德欣,李小平.基于小波分析和局部变量的车型识别仁[J].军民两用技术与产品.2006,3:29~33.
[15]郁梅,蒋刚毅.智能交通系统中的计算机视觉技术应用[J].计算机工程与应用,2001,(10):101~104.
[16]伍友龙.基于图像分析的高速公路交通事件检测算法研究[J].长沙理工大学硕士学位论文,2005,3.
[17]张玲,陈丽敏,何伟等.基于视频的改进帧差法在车流量检测中的应用[J].重庆大学学报(自然科学版),2004,27(s):3~34.
[18]明英,边馥苓,蒋晶珏,王钢.对光照、阴影和反光具有鲁棒性的变化检测算法及实现[J].计算机工程与应用,2004(24):23~26.
[19]李庆忠,陈显华,王立红.视频监控中运动目标检测和识别方法[J].计算机工程,2004;30(16):143~145.
[20]黄文清,汪亚明.基于自适应混合模型和区域特征的多目标跟踪与检测[J].中国图形图像学报,2000:5(12):1024~1029.
[21] A.J.SCHOFIELD,P.A. MEHTA and T.J.STONHAM.A system for counting people in video images using neural networks to identify the background scene. Pattern Recognition, 1996, 29(8):1421~1428.
[22]谷口庆治,数字图像处理------应用篇,科学出版社,2002,291~302.
[23]严柏军,郑链,王克勇一种改进的基于光流的运动目标的检测算法[J].红外技术,2001,28(3):351~354.
[24] KmaedaY. Minoh Miehihiko a Human Motion Estimation Method Using 3-suceessive Video Frames. Proceedings of International Conference on Virtual System sand Multimedia,1996:135~140.
[25]王春波,张卫东.智能交通系统运动车辆的视觉检测[J].红外与毫米波学报,2001,20(2):81~86.
[26] Kim J B, Park H S.A Real-time Region-based Motion Segmentation Using Adaptive Thresholding and K-means Clustering. M Brooks, D Corbtett, M StumPtner. Springer, 2001, 30(6):213~224.
[27] Du Yuren, Zhou Aijun. Moving Object Detection for Video Monitoring Systems. The eighth international conference on electronic measurement and instruments 2007 XI'AN.
[28] Pankaj Kumar, Surendra Ranganath, Huang Weimin, et al. Frame work for real-time behavior interpretation from traffic video [J]. IEEE Transactions on Intelligent Transportation Systems, 2005, 6(1): 43~53.
[29]胡小锋,赵辉.visual c++/MATLAB图像处理与识别应用案例精选[M],北京:人民邮电出版社,2003.
[30]张玲,叶海炳,何伟一种基于边缘信息的改进车辆检测方法[J].重庆大学学报(自然科学版),2001,27(11):56~58.
[31]周爱军,杜宇人,基于视频图像Harris角点检测的车型识别[J].扬州大学学报(自然科学版), 2008,11(1):67~70.
[32]胡会珍.基于视觉的运动车辆识别与跟踪方法研究.中北大学硕士学位论文,2008.4:4~5.
[33]章毓晋.图象分割[D].北京:科学出版社,2001.
[34]韩思奇,王蕾.图像分割的阈值法综述.系统工程与电子技术,2002, (24):91~95.
[35]贾允,丁艳,刘泽平.改进图像阈值分割算法的研究[J].光学技术,2005,31(1):155~157.
[36] Francois G, Meyer, Patrick B, Region based tracking using a fine motion models. in long image sequences[J].Computer Vision, Graphics and Image Processing, 1994,60(2):119-140.
[37] Jorge B, Bober M, Pla F, Motion and intensity based segmentation and its application to traffic monitoring[C]. Proceedings of International Conference on Image Analysis and Processings, 1997:502~509.
[38] Grinias I, Tziritas G. A semi-automatic seeded region-growing algorithm for video object localization and tracking [J].Signal Processing: Image Communication, 2001(16):977~986.
[39] Hall, D. Nascimento, J. Ribeiro, P. et al. Comparison of target detection algorithms using adaptive background models [C]. Visual Surveillance and Performance Evaluation of Tracking and Surveillance, 2005. 2nd Joint IEEE International Workshop on, 2005:113~120.
[40] R. Cucchiara, C. Grana, M. Piccardi, et al. Improving Shadow Suppression in Moving Object Detectionwith HSV Color Information. IEEE Intelligent Transportation System Conference Proceedings, 2001:334~339.
[41]王圣男,郁梅,蒋刚毅.智能交通系统中基于视频图像处理的车辆检测与跟踪方法综述[J].计算机应用研究,2005,9:9~14.
[42] Jen-Chao Taj, Shung-Tsang Tseng. Real-time Image Tracking for Automatic Traffic Monitoring and Enforcement Applications Visual track in[J].Image and Vision Computing 2004,22(6): 640~649.
[43]林海涵,唐慧明.基于视频的车辆检测和分析算法[J].江南大学学报:自然科学版,2007,6(3):323~326.
[44] Nadimi S, Bhanu B. Physical models for moving shadow and object detection in video[J].IEEE Trans Patt Anal and Mach Intell, 2004,26:1079~1087.
[45]袁基炜,史忠科.一种运动车辆的阴影消除新方法[J].西安交通大学学报,2005,39(6):598~602.
[46] Wang H, Suter D, Schindler K. Adaptive object tracking based on an effective appearance filter. IEEE Transactions on Patten Analysis and Machine intelligence, 2007, 29(9):1661~1667.
[47]丁德志.基于视频序列的车辆检测与跟踪算法研究与应用.山东师范大学工学硕士学位论文,2009.
[48]刘思峰,谢乃明.灰色系统理论及其应用(第四版).北京:科学出版社,2008:105~110.
[49]高岚,董慧颖,兰利宝.自适应背景下运动目标阴影检测算法研究[J].现代电子技术,2008,31(6),59-61.
[50]罗军辉,冯平.Matlab7.0在图像处理中的应用[M].北京:机械工业出版社,2005.
[2] S.Gupte, O.Masoud, RFK Martin,“Detection and classification of vehicles”, IEEE Transactions on Intelligent Transportation Systems,2002,3 1:37～47.
[3]张伟.基于视觉的运动车辆检测与跟踪[D].上海交通大学工学博士学位论文,2007.
[4] Texas Instruments. SPRS200F, TMS320DM642 Video/Imaging Fixed– Point Digital Signal Processor Data Manual[S].2004.
[5] Qiang Ji, Xiaojie Yang. Real-time eye, gaze, and face pose tracking for monitoring driver vigilance. Real-Time Imaging, 2002, 8:357~377.
[6] Jose Melo, Andrew Naftel, Alexandre Bernardin, et al. Detection and Classification of Highway Lanes Using Vehicle Motion Trajectories [J]. IEEE Transactions on Intelligent Transportation Systems, 2006,7(2):188~200.
[7]朱志刚,徐光佑,扬波.自动交通监测系统的二维时空图像方法.中国图像图形学报,1996,1(2):101~10.
[8]吴裙文,张学工.应用小波变换和PAC进行车辆的静态图像检测[J].清华大学学报,2002,42(11):1560~1564.
[9]周志宇,汪亚明,黄文清.基于遗传算法的指数熵在车辆运动检测中的应用[J].浙江工程学院学报,2003,20(9):202~205.
[10]肖旺新,张雪,黄卫.交通视频图像的多尺度自适应阈值边缘检测[J].土木工程学报,2003,37(6):89~94.
[11]周欣.基于方向分形维的高速公路车辆边缘检测研究[J].信号处理,2004,20(3):258~262.
[12] ANZAIY, KATOT, HIGASHIKUBOM, et al. Development of an integrated video imaging vehicle detector [J].SEI Technical Re2view,2005,59 (6).
[13] Liu G Y, Chen R, Deng Y et al. Video based 3D human body motion capture[J]. Journal of Computer-Aided Design and Computer Graphics, 2006,18(l):82~88.
[14]车德欣,李小平.基于小波分析和局部变量的车型识别仁[J].军民两用技术与产品.2006,3:29~33.
[15]郁梅,蒋刚毅.智能交通系统中的计算机视觉技术应用[J].计算机工程与应用,2001,(10):101~104.
[16]伍友龙.基于图像分析的高速公路交通事件检测算法研究[J].长沙理工大学硕士学位论文,2005,3.
[17]张玲,陈丽敏,何伟等.基于视频的改进帧差法在车流量检测中的应用[J].重庆大学学报(自然科学版),2004,27(s):3~34.
[18]明英,边馥苓,蒋晶珏,王钢.对光照、阴影和反光具有鲁棒性的变化检测算法及实现[J].计算机工程与应用,2004(24):23~26.
[19]李庆忠,陈显华,王立红.视频监控中运动目标检测和识别方法[J].计算机工程,2004;30(16):143~145.
[20]黄文清,汪亚明.基于自适应混合模型和区域特征的多目标跟踪与检测[J].中国图形图像学报,2000:5(12):1024~1029.
[21] A.J.SCHOFIELD,P.A. MEHTA and T.J.STONHAM.A system for counting people in video images using neural networks to identify the background scene. Pattern Recognition, 1996, 29(8):1421~1428.
[22]谷口庆治,数字图像处理------应用篇,科学出版社,2002,291~302.
[23]严柏军,郑链,王克勇一种改进的基于光流的运动目标的检测算法[J].红外技术,2001,28(3):351~354.
[24] KmaedaY. Minoh Miehihiko a Human Motion Estimation Method Using 3-suceessive Video Frames. Proceedings of International Conference on Virtual System sand Multimedia,1996:135~140.
[25]王春波,张卫东.智能交通系统运动车辆的视觉检测[J].红外与毫米波学报,2001,20(2):81~86.
[26] Kim J B, Park H S.A Real-time Region-based Motion Segmentation Using Adaptive Thresholding and K-means Clustering. M Brooks, D Corbtett, M StumPtner. Springer, 2001, 30(6):213~224.
[27] Du Yuren, Zhou Aijun. Moving Object Detection for Video Monitoring Systems. The eighth international conference on electronic measurement and instruments 2007 XI'AN.
[28] Pankaj Kumar, Surendra Ranganath, Huang Weimin, et al. Frame work for real-time behavior interpretation from traffic video [J]. IEEE Transactions on Intelligent Transportation Systems, 2005, 6(1): 43~53.
[29]胡小锋,赵辉.visual c++/MATLAB图像处理与识别应用案例精选[M],北京:人民邮电出版社,2003.
[30]张玲,叶海炳,何伟一种基于边缘信息的改进车辆检测方法[J].重庆大学学报(自然科学版),2001,27(11):56~58.
[31]周爱军,杜宇人,基于视频图像Harris角点检测的车型识别[J].扬州大学学报(自然科学版), 2008,11(1):67~70.
[32]胡会珍.基于视觉的运动车辆识别与跟踪方法研究.中北大学硕士学位论文,2008.4:4~5.
[33]章毓晋.图象分割[D].北京:科学出版社,2001.
[34]韩思奇,王蕾.图像分割的阈值法综述.系统工程与电子技术,2002, (24):91~95.
[35]贾允,丁艳,刘泽平.改进图像阈值分割算法的研究[J].光学技术,2005,31(1):155~157.
[36] Francois G, Meyer, Patrick B, Region based tracking using a fine motion models. in long image sequences[J].Computer Vision, Graphics and Image Processing, 1994,60(2):119-140.
[37] Jorge B, Bober M, Pla F, Motion and intensity based segmentation and its application to traffic monitoring[C]. Proceedings of International Conference on Image Analysis and Processings, 1997:502~509.
[38] Grinias I, Tziritas G. A semi-automatic seeded region-growing algorithm for video object localization and tracking [J].Signal Processing: Image Communication, 2001(16):977~986.
[39] Hall, D. Nascimento, J. Ribeiro, P. et al. Comparison of target detection algorithms using adaptive background models [C]. Visual Surveillance and Performance Evaluation of Tracking and Surveillance, 2005. 2nd Joint IEEE International Workshop on, 2005:113~120.
[40] R. Cucchiara, C. Grana, M. Piccardi, et al. Improving Shadow Suppression in Moving Object Detectionwith HSV Color Information. IEEE Intelligent Transportation System Conference Proceedings, 2001:334~339.
[41]王圣男,郁梅,蒋刚毅.智能交通系统中基于视频图像处理的车辆检测与跟踪方法综述[J].计算机应用研究,2005,9:9~14.
[42] Jen-Chao Taj, Shung-Tsang Tseng. Real-time Image Tracking for Automatic Traffic Monitoring and Enforcement Applications Visual track in[J].Image and Vision Computing 2004,22(6): 640~649.
[43]林海涵,唐慧明.基于视频的车辆检测和分析算法[J].江南大学学报:自然科学版,2007,6(3):323~326.
[44] Nadimi S, Bhanu B. Physical models for moving shadow and object detection in video[J].IEEE Trans Patt Anal and Mach Intell, 2004,26:1079~1087.
[45]袁基炜,史忠科.一种运动车辆的阴影消除新方法[J].西安交通大学学报,2005,39(6):598~602.
[46] Wang H, Suter D, Schindler K. Adaptive object tracking based on an effective appearance filter. IEEE Transactions on Patten Analysis and Machine intelligence, 2007, 29(9):1661~1667.
[47]丁德志.基于视频序列的车辆检测与跟踪算法研究与应用.山东师范大学工学硕士学位论文,2009.
[48]刘思峰,谢乃明.灰色系统理论及其应用(第四版).北京:科学出版社,2008:105~110.
[49]高岚,董慧颖,兰利宝.自适应背景下运动目标阴影检测算法研究[J].现代电子技术,2008,31(6),59-61.
[50]罗军辉,冯平.Matlab7.0在图像处理中的应用[M].北京:机械工业出版社,2005.