基于视频的城市交通拥挤检测系统的研究
|
摘要
急剧增长的机动车保有量与十分有限的路网容量之间的严重不平衡，造成了如今全世界范围内各大城市普遍可见的城市交通拥挤现象。许多实践证明，单独从交通需求或供给上试图解决交通拥挤都是很难奏效的，而且我国很多城市也面临着无地可用来新建道路设施的尴尬。此时，交通管理就显得尤为重要，完善的管理能最大限度的发挥出路网的容量，反之则会降低路网容量。近些年来，一种智能且高效的交通管理手段正应运而生----智能交通系统。IntelligentTransportation System，简称ITS。智能交通系统是将先进的信息技术、通讯技术、传感技术、控制技术以及计算机技术等有效地集成运用于整个交通运输管理体系，而建立起的一种在大范围内、全方位发挥作用的，实时、准确、高效的综合运输和管理系统。许多发达国家和地区早就已经开始了对智能交通系统的研究和使用，我国近些年来也逐渐加大了对这一领域的研究，其广泛的应用前景和经济社会价值正使得对智能交通系统的研究成为新兴热门的研究课题。
智能交通系统是一个庞大而复杂的系统，本研究中提出的基于视频的城市交通拥挤检测系统就是其中一个最基础的子系统，主要实现对城市主要道路进行实时交通状态的检测和判别，并将准确的检测结果提供给交通管理部门，为其提供必要的决策依据。各国研究人员都展开了对这一系统的研究，但是目前国内成型的交通拥挤检测系统却不多。本文根据当前国内外先进的硬件和软件技术设计了一套基于视频的智能交通拥挤检测系统，并详细介绍了系统各部分及其实现过程。
本文中提出的核心算法包括两个方面：拥挤判别主要交通参数的检测算法及交通拥挤检测算法。拥挤判别主要交通参数的检测基于先进的视频检测技术，首先通过简单的背景差法获得车辆目标，但是由于背景差法的局限性使得提取的车辆目标中包含车辆阴影，因此通过归一化互相关算法将车辆阴影去除以获得完整的车辆目标；再运用虚拟检测线和虚拟检测面相结合的改进虚拟线圈检测方法检测本研究中所需的三个交通参数----流量、速度和占有率。交通拥挤检测算法首先分析各国交通拥挤状态的定义并将交通拥挤状态按4个等级划分----畅通、轻度拥挤、拥挤和严重拥挤，并通过借鉴其他领域提出的蚁群算法和模糊C均值聚类算法获得交通拥挤状态的4个等级聚类中心，再通过计算当前交通参数对各等级标准的隶属度来确定当前交通参数所反映的交通拥挤状态。
本文中系统实现主要包括两个部分，硬件部分主要介绍了本研究采用的处理平台----TI公司研发的高性能数字信号处理器(TMS320DM642)，以及其视频采集和回放的硬件设备配置；软件部分主要介绍了系统的软件开发环境，并重点讲述了系统各个模块具体软件实现过程，并结合一个具体实例，对本研究中提出的算法进行可靠性和可行性验证。
最后对本文系统做了全面总结，并提出了本系统存在的不足及本人对于系统中有待改进部分的构想和思路。
With the rapid growth of the motor vehicle and the very limited road networkcapacity unbalanced, the major cities in the worldwide all have a big problem ----urban traffic congestion. Practice has proved that it is very difficult to solve the trafficcongestion only with the traffic demand or traffic supply. In china many cities all facewith the embarrassment of no available land to create the new road facilities. At thistime, traffic management becomes particularly important, and the better trafficmanagement can maximize the network capacity, otherwise it will reduce the capacityof the road network. In recent years, an intelligent and efficient means of trafficmanagement have emerged ---- Intelligent Transportation Systems, abbr. ITS.Intelligent Transportation Systems applies the effectively integrate advancedinformation technology, the communication technology, the sensor technology, thecontrol technology and the computer technology to the entire system, and builds aall-round role, large scale , real-time, accurate and efficient integrated transportationmanagement system. Many developed countries and regions have already begun toresearch and use of intelligent transportation system, and China has graduallyincreased the research in this area in recent years, its broad application prospects,economic value and social value make the study of the intelligent transportationsystem become an emerging hot research topic.
Intelligent transportation system is a large and complex system, the video-basedurban traffic congestion detection system proposed in this study is one of the mostbasic subsystem. The main fuction of this system is to detecte and discriminate thereal-time traffic status of the main roads in the city, then provide the accurate results tothe public security traffic management department and help them make the rightdecision. Many researchers have carried out many studies of this system, but the wholetraffic congestion detection system is seldom in our country. In this paper, based on theadvanced hardware and software technology design a set of intelligent video-basedtraffic congestion detection system, and introduce all the design details of the system and the implementation process.
In this paper, the core algorithm includes two aspects: the traffic parametersdetection algorithm and congestion detection algorithm. The traffic parametersdetection algorithm uses the advanced video detection technology for trafficparameters detection. First, through a simple background difference method to obtainthe vehicle target, but with the limitations of the background subtraction methodmake the extraction of the vehicle target with the vehicle shadow, then the normalizedcorrelation algorithm is used to remove the vehicle shadow and then get the wholevehicle targets; Then combinate the virtual test line and the virtual test face to thevirtual coil detection method to detect the three traffic parameters---- traffic flow,traffic speed and the traffic sharing. In the part of the congestion detection algorithm,analyze the definition of the countries congestion state and then divide the congestionstate in to four levels ---- smooth, mild crowding, congestion and severe overcrowding,then refernce the classical algorithm in related areas ---- Ant Colony Algorithm andFuzzy C-means Clustering Algorithm to get the traffic congestion status of the fourhierarchical clustering center. Calculate the grade standard of membership of thecurrent traffic parameters to detemine the current traffic state of the traffic parameters.
The part of the system design in this study includes two aspects. Hardware sectionmainly introduces a processing platform ----TMS320DM642, which is ahigh-performance didital signal processor developed by TI, then introduces somethingabout the hardware configuration of the video capture and the playback; Softwaresection describes the system software development environment, and focus on thesoftware implmentation process of each subsystem module. Then combined with aconcrete example, verificate the reliability and the feasibility of the algorithmproposed in this study.
At last, make a comprehensive summary of the whole thesis, showing theshortcoming of this system and some ideas of some parts needed to be improved.
智能交通系统是一个庞大而复杂的系统，本研究中提出的基于视频的城市交通拥挤检测系统就是其中一个最基础的子系统，主要实现对城市主要道路进行实时交通状态的检测和判别，并将准确的检测结果提供给交通管理部门，为其提供必要的决策依据。各国研究人员都展开了对这一系统的研究，但是目前国内成型的交通拥挤检测系统却不多。本文根据当前国内外先进的硬件和软件技术设计了一套基于视频的智能交通拥挤检测系统，并详细介绍了系统各部分及其实现过程。
本文中提出的核心算法包括两个方面：拥挤判别主要交通参数的检测算法及交通拥挤检测算法。拥挤判别主要交通参数的检测基于先进的视频检测技术，首先通过简单的背景差法获得车辆目标，但是由于背景差法的局限性使得提取的车辆目标中包含车辆阴影，因此通过归一化互相关算法将车辆阴影去除以获得完整的车辆目标；再运用虚拟检测线和虚拟检测面相结合的改进虚拟线圈检测方法检测本研究中所需的三个交通参数----流量、速度和占有率。交通拥挤检测算法首先分析各国交通拥挤状态的定义并将交通拥挤状态按4个等级划分----畅通、轻度拥挤、拥挤和严重拥挤，并通过借鉴其他领域提出的蚁群算法和模糊C均值聚类算法获得交通拥挤状态的4个等级聚类中心，再通过计算当前交通参数对各等级标准的隶属度来确定当前交通参数所反映的交通拥挤状态。
本文中系统实现主要包括两个部分，硬件部分主要介绍了本研究采用的处理平台----TI公司研发的高性能数字信号处理器(TMS320DM642)，以及其视频采集和回放的硬件设备配置；软件部分主要介绍了系统的软件开发环境，并重点讲述了系统各个模块具体软件实现过程，并结合一个具体实例，对本研究中提出的算法进行可靠性和可行性验证。
最后对本文系统做了全面总结，并提出了本系统存在的不足及本人对于系统中有待改进部分的构想和思路。
With the rapid growth of the motor vehicle and the very limited road networkcapacity unbalanced, the major cities in the worldwide all have a big problem ----urban traffic congestion. Practice has proved that it is very difficult to solve the trafficcongestion only with the traffic demand or traffic supply. In china many cities all facewith the embarrassment of no available land to create the new road facilities. At thistime, traffic management becomes particularly important, and the better trafficmanagement can maximize the network capacity, otherwise it will reduce the capacityof the road network. In recent years, an intelligent and efficient means of trafficmanagement have emerged ---- Intelligent Transportation Systems, abbr. ITS.Intelligent Transportation Systems applies the effectively integrate advancedinformation technology, the communication technology, the sensor technology, thecontrol technology and the computer technology to the entire system, and builds aall-round role, large scale , real-time, accurate and efficient integrated transportationmanagement system. Many developed countries and regions have already begun toresearch and use of intelligent transportation system, and China has graduallyincreased the research in this area in recent years, its broad application prospects,economic value and social value make the study of the intelligent transportationsystem become an emerging hot research topic.
Intelligent transportation system is a large and complex system, the video-basedurban traffic congestion detection system proposed in this study is one of the mostbasic subsystem. The main fuction of this system is to detecte and discriminate thereal-time traffic status of the main roads in the city, then provide the accurate results tothe public security traffic management department and help them make the rightdecision. Many researchers have carried out many studies of this system, but the wholetraffic congestion detection system is seldom in our country. In this paper, based on theadvanced hardware and software technology design a set of intelligent video-basedtraffic congestion detection system, and introduce all the design details of the system and the implementation process.
In this paper, the core algorithm includes two aspects: the traffic parametersdetection algorithm and congestion detection algorithm. The traffic parametersdetection algorithm uses the advanced video detection technology for trafficparameters detection. First, through a simple background difference method to obtainthe vehicle target, but with the limitations of the background subtraction methodmake the extraction of the vehicle target with the vehicle shadow, then the normalizedcorrelation algorithm is used to remove the vehicle shadow and then get the wholevehicle targets; Then combinate the virtual test line and the virtual test face to thevirtual coil detection method to detect the three traffic parameters---- traffic flow,traffic speed and the traffic sharing. In the part of the congestion detection algorithm,analyze the definition of the countries congestion state and then divide the congestionstate in to four levels ---- smooth, mild crowding, congestion and severe overcrowding,then refernce the classical algorithm in related areas ---- Ant Colony Algorithm andFuzzy C-means Clustering Algorithm to get the traffic congestion status of the fourhierarchical clustering center. Calculate the grade standard of membership of thecurrent traffic parameters to detemine the current traffic state of the traffic parameters.
The part of the system design in this study includes two aspects. Hardware sectionmainly introduces a processing platform ----TMS320DM642, which is ahigh-performance didital signal processor developed by TI, then introduces somethingabout the hardware configuration of the video capture and the playback; Softwaresection describes the system software development environment, and focus on thesoftware implmentation process of each subsystem module. Then combined with aconcrete example, verificate the reliability and the feasibility of the algorithmproposed in this study.
At last, make a comprehensive summary of the whole thesis, showing theshortcoming of this system and some ideas of some parts needed to be improved.
引文
[1]公安部：全国机动车保有量达2.17亿辆[OL].http://society.people.com.cn/GB/223276/15203743.html
[2]智能交通系统引发期待[OL].http://news.sina.com.cn/c/sd/2010-11-22/154921513841_4.shtml
[3]郭蓓.基于视频图像处理的高速公路隧道超速检测报警系统的研究与开发[D] .2006年
[4]罗刚.日本智能交通技术的发展状况及值得借鉴的经验[J].道路交通与安全.2004年02期
[5]张秀媛.公路自动事件检测技术[J].系统工程理论于实践, 2001, 6(6):118. 24.
[6]林东.交通事件视频检测系统嵌入式软件研究与实现[D] .长安大学.2009
[7] Panos G. Michalopoulus, Vehicle Detection Video Through Image Processing: TheAutoscope System, IEEE TRANSACTION ON VEHICULAR TECHNOLOGY,VOL.40, No. 1, FEBRUARY 1991, P21-29
[8]乔光军,杨兆选.基于虚拟线的交通信息视频检测技术及应用[J].电子技术应.2005年09期
[9]汪勤,黄山,张洪斌等.基于视频图像处理的交通事件检测系统[J].计算机应用研究.2008第7期
[10]刘卫宁,曾恒,孙棣华,赵敏.基于视频检测技术的交通拥挤判别模型[J].计算机应用研究.2010第8期
[11]王俊,曾致,付祥胜.一种视频交通检测器的设计与实现[J].电视技术.2005第10期
[12]陈永康,劳云腾,李志慧,王殿海.复杂环境下背景更新和阴影检测算法研究[C].2005年海峡两岸智能交通运输系统学术研讨会暨第二届同舟交通论坛论文集.2005年
[13]吴睿.基于视频的城市交通事件检测系统的研究[D].南京邮电大学.2007年
[14] Payne H J,R D Hdfcnl eh,H C Knobel.Development and testing of incidentdetection algorithms[J].research methodology and demiled results,DeportNo-FHWA-RD-76-20 Federal Highway Administration,1976,2:121-126
[15] Utah Department of Transportation,University of Utah.INCIDENT DETECTIONALGORITHM EVALUATION.March 2001
[16] Peter T, Martin, Joseph Perrin, Blake Hansen. Incident Detection AlgorithmEvaluation[M]. 2001
[17] Justin Black,Indu Sreedevi.Automatic Incident Detection Algorithms [M] .2001.
[18] Levin M,Kranae G M.Incident detection:A Bayesian Approach[J].TransportationResearch Record,1978:52-58.
[19] CHEN C C, YIH Y, WU Y C. Auto bias selection for learning based schedulingsystems[J].International Journal of Production Research,1999,37(9):1987-2002.
[20] Haruo ISHIDA, Hideki FURUYA, Shin-ichiro Kai, Naohisa OKAMOTO.TRAVELSPEED AND TRAFFIC CONGESTIONS RECOGNITION ON EXPRESSWAYS.Journal of the Eastern Asia Society for Transportation Studies, Vol.5, October,2003
[21]姜桂艳,王江锋,冮龙晖,张晓东.基于ANN技术的城市快速路交通拥挤自动识别算法研究[C].第八届国际交通新技术应用大会论文集.2004年第8期
[22]房丽侠,魏连雨,闫伟阳.基于决策树的交通拥挤状态预测[J].河北工业大学学报; 2010年2期
[23]窦慧丽,王国华.基于模糊聚类和判别分析的交通状态提取算法[J].交通信息与安全.2010年2期
[24]邹亮,徐建闽,朱玲湘.基于融合技术的道路交通状态判别模型[J].清华大学学报（自然科学版）.2007年2期
[25]杨祖元,黄席樾,杜长海等.基于FFCM聚类的城市交通拥堵判别研究[J].计算机应用研究,2008,25(9):2768-2770.
[26]周伟,罗石贵.基于模糊综合识别的事件检测算法[J].西安公路交通大学学报.2001年2期
[27]周雪铭.基于小波分析和神经网络的交通事件自动检测[J.公路与汽运.2002年4期
[28]姜桂艳.道路交通状态判别技术与应用[M].人民交通出版社.2004,9
[29]杜豫川,周小鹏,郝媛,孙立军.基于实时检测数据的交通拥挤检测算法研究[C].2006年第二届中国智能交通年会集.2006
[30]普运伟,金炜东,朱明等.核模糊c-均值算法的聚类有效性研究[J].计算机科学.2007,34(2):207—210.
[31]黄红星.基于蚁群算法的动态模糊聚类分析[OL].中国科技论文在线,2005.
[32]康琦,汪镭,安静等.群体智能计算[C] .2004年生命系统建模仿真国际会议暨第一届全国生命系统建模仿真学术会议论文集.2004年10月
[33]鲁一杰,赵攀,彭强,江永全.基于DSP的视频车辆实时检测系统的实现[J].计算机应用研究.2009年2期
[34]王玮,张卫宁,张大志,史英.DM642在网络视频传输系统中的应用[J];自动化仪表,2007年2期
[35]叶磊.智能视频监控嵌入式实现及传输保护研究[D].上海交通大学.2009年
[36]王一.基于TMS320DM642的H.264编解码器的移植与优化[D].天津大学.2008年
[37] TMS320DM642[OL]. http://baike.baidu.com/view/1013760.htm
[38] SEED-VPM642用户指南(Rev. B).2005年
[39] Texas Instruments Incorporated, HPA Digital Audio Video SLES087, September2003
[40] Texas Instrumnets Incorprated著,牛金海等编译.Code_Composer_Studio(CCS)集成开发环境(IDE)入门指导书V10
[41] Texas Instruments Incorprated .DSP集成开发环境CCS开发指南
[42] opencv2.0_使用手册[OL]. http://wenku.baidu.com/view.2009年5月
[43] opencv中文手册[OL].http://wenku.baidu.com/view
[44]沈华.基于插值和主元素分析的人脸超分辨率算法研究[D].湖南大学.2010年
[45] Texas Instruments.TMS320C6000 DSP Enhanced Direct Memory Access (EDMA)Controller Reference Guide, Application Report SPRU 234b; March 2005
[46]任丽香.TM S320C6000系列DSPs的原理与应用[M] .电子L业出版社.2001年6月
[47]王念旭等.DSP基础与应用系统设计[M] .北航出版社.2002年8月
[48]张雄伟等.DSP芯片的原理与开发[M] .电子工业出版社.2001年1月
[2]智能交通系统引发期待[OL].http://news.sina.com.cn/c/sd/2010-11-22/154921513841_4.shtml
[3]郭蓓.基于视频图像处理的高速公路隧道超速检测报警系统的研究与开发[D] .2006年
[4]罗刚.日本智能交通技术的发展状况及值得借鉴的经验[J].道路交通与安全.2004年02期
[5]张秀媛.公路自动事件检测技术[J].系统工程理论于实践, 2001, 6(6):118. 24.
[6]林东.交通事件视频检测系统嵌入式软件研究与实现[D] .长安大学.2009
[7] Panos G. Michalopoulus, Vehicle Detection Video Through Image Processing: TheAutoscope System, IEEE TRANSACTION ON VEHICULAR TECHNOLOGY,VOL.40, No. 1, FEBRUARY 1991, P21-29
[8]乔光军,杨兆选.基于虚拟线的交通信息视频检测技术及应用[J].电子技术应.2005年09期
[9]汪勤,黄山,张洪斌等.基于视频图像处理的交通事件检测系统[J].计算机应用研究.2008第7期
[10]刘卫宁,曾恒,孙棣华,赵敏.基于视频检测技术的交通拥挤判别模型[J].计算机应用研究.2010第8期
[11]王俊,曾致,付祥胜.一种视频交通检测器的设计与实现[J].电视技术.2005第10期
[12]陈永康,劳云腾,李志慧,王殿海.复杂环境下背景更新和阴影检测算法研究[C].2005年海峡两岸智能交通运输系统学术研讨会暨第二届同舟交通论坛论文集.2005年
[13]吴睿.基于视频的城市交通事件检测系统的研究[D].南京邮电大学.2007年
[14] Payne H J,R D Hdfcnl eh,H C Knobel.Development and testing of incidentdetection algorithms[J].research methodology and demiled results,DeportNo-FHWA-RD-76-20 Federal Highway Administration,1976,2:121-126
[15] Utah Department of Transportation,University of Utah.INCIDENT DETECTIONALGORITHM EVALUATION.March 2001
[16] Peter T, Martin, Joseph Perrin, Blake Hansen. Incident Detection AlgorithmEvaluation[M]. 2001
[17] Justin Black,Indu Sreedevi.Automatic Incident Detection Algorithms [M] .2001.
[18] Levin M,Kranae G M.Incident detection:A Bayesian Approach[J].TransportationResearch Record,1978:52-58.
[19] CHEN C C, YIH Y, WU Y C. Auto bias selection for learning based schedulingsystems[J].International Journal of Production Research,1999,37(9):1987-2002.
[20] Haruo ISHIDA, Hideki FURUYA, Shin-ichiro Kai, Naohisa OKAMOTO.TRAVELSPEED AND TRAFFIC CONGESTIONS RECOGNITION ON EXPRESSWAYS.Journal of the Eastern Asia Society for Transportation Studies, Vol.5, October,2003
[21]姜桂艳,王江锋,冮龙晖,张晓东.基于ANN技术的城市快速路交通拥挤自动识别算法研究[C].第八届国际交通新技术应用大会论文集.2004年第8期
[22]房丽侠,魏连雨,闫伟阳.基于决策树的交通拥挤状态预测[J].河北工业大学学报; 2010年2期
[23]窦慧丽,王国华.基于模糊聚类和判别分析的交通状态提取算法[J].交通信息与安全.2010年2期
[24]邹亮,徐建闽,朱玲湘.基于融合技术的道路交通状态判别模型[J].清华大学学报（自然科学版）.2007年2期
[25]杨祖元,黄席樾,杜长海等.基于FFCM聚类的城市交通拥堵判别研究[J].计算机应用研究,2008,25(9):2768-2770.
[26]周伟,罗石贵.基于模糊综合识别的事件检测算法[J].西安公路交通大学学报.2001年2期
[27]周雪铭.基于小波分析和神经网络的交通事件自动检测[J.公路与汽运.2002年4期
[28]姜桂艳.道路交通状态判别技术与应用[M].人民交通出版社.2004,9
[29]杜豫川,周小鹏,郝媛,孙立军.基于实时检测数据的交通拥挤检测算法研究[C].2006年第二届中国智能交通年会集.2006
[30]普运伟,金炜东,朱明等.核模糊c-均值算法的聚类有效性研究[J].计算机科学.2007,34(2):207—210.
[31]黄红星.基于蚁群算法的动态模糊聚类分析[OL].中国科技论文在线,2005.
[32]康琦,汪镭,安静等.群体智能计算[C] .2004年生命系统建模仿真国际会议暨第一届全国生命系统建模仿真学术会议论文集.2004年10月
[33]鲁一杰,赵攀,彭强,江永全.基于DSP的视频车辆实时检测系统的实现[J].计算机应用研究.2009年2期
[34]王玮,张卫宁,张大志,史英.DM642在网络视频传输系统中的应用[J];自动化仪表,2007年2期
[35]叶磊.智能视频监控嵌入式实现及传输保护研究[D].上海交通大学.2009年
[36]王一.基于TMS320DM642的H.264编解码器的移植与优化[D].天津大学.2008年
[37] TMS320DM642[OL]. http://baike.baidu.com/view/1013760.htm
[38] SEED-VPM642用户指南(Rev. B).2005年
[39] Texas Instruments Incorporated, HPA Digital Audio Video SLES087, September2003
[40] Texas Instrumnets Incorprated著,牛金海等编译.Code_Composer_Studio(CCS)集成开发环境(IDE)入门指导书V10
[41] Texas Instruments Incorprated .DSP集成开发环境CCS开发指南
[42] opencv2.0_使用手册[OL]. http://wenku.baidu.com/view.2009年5月
[43] opencv中文手册[OL].http://wenku.baidu.com/view
[44]沈华.基于插值和主元素分析的人脸超分辨率算法研究[D].湖南大学.2010年
[45] Texas Instruments.TMS320C6000 DSP Enhanced Direct Memory Access (EDMA)Controller Reference Guide, Application Report SPRU 234b; March 2005
[46]任丽香.TM S320C6000系列DSPs的原理与应用[M] .电子L业出版社.2001年6月
[47]王念旭等.DSP基础与应用系统设计[M] .北航出版社.2002年8月
[48]张雄伟等.DSP芯片的原理与开发[M] .电子工业出版社.2001年1月