基于视频流的车辆检测和跟踪算法的研究
|
摘要
智能交通系统是一种智能化的系统,通过对道路交通流的实时检测,它能够全方位、实时准确和高效的对路面上的车辆进行检测和跟踪,然后根据道路运行状况以及交通流的动态变化迅速做出诱导控制,在一定程度上减轻了道路拥塞程度,缓解了道路交通压力,降低了事故发生率。
本文对基于视频的运动车辆检测识别、跟踪和违章检测等方面做了分析研究并进行了大量的实验,使用Visual C++ 6.0作为开发环境,并结合Intel公司的OpenCV图像处理库进行开发。通过使用OpenCV技术,实时的捕捉视频流数据,对图像进行灰度化和平滑处理,用背景差和帧间差相结合的方法得到感兴趣区域后,进行二值化处理,此时得到的图像黑色为背景,白色为前景运动目标。
在车辆检测识别阶段,本文采用了轮廓提取法提取二值图中的轮廓,然后计算轮廓的外接矩形,对满足一定条件的矩形认为是车辆,用矩形框圈起来这就完成了对视频中的车辆的检测和识别。在车辆跟踪阶段,提出了颜色和质心距离相结合的多特征匹配算法来跟踪已经检测识别出的车辆。分类研究了交通违章事件等行为,分别对违章停车、车辆逆向行驶、违章变道压线等车辆违章行为做了实验,检测识别出视频中的车辆违规事件并对违章信息进行管理。
At present, motor vehicles are increasing. This causes traffic problems worse and often generates a number of traffic accidents. Man-made illegal events include running red lights, speeding, illegal lane change. Intelligent Transportation System is an intelligent system, through the real-time detection of road traffic flow, it can detect and track vehicles accurately-According to the traffic flow on the road, the system can response quickly. It can reduce road congestion, ease the traffic pressure and reduce the accident rate through taking certain measures. Currently, the intelligent traffic system has been used in vehicle detection and tracking, calculating speed.
This paper makes a lot of experiments on motion detection and identification of vehicles. Using Visual C++ 6.0 as the development environment, combined with OpenCV image processing library for development. We can play the video stream and read a frame of the video image by OpenCV. In the process of studying algorithm, the image processing functions is provided by OpenCV library.
After capturing video frame by OpenCV library functions, the color space is converted and the video image is processed. The background subtraction and frame difference have advantages and disadvantages. If the combination of these two methods, the performance was complementary, it can not only detect the movement of vehicles and also detect stationary vehicles. At the same time. the result of frame difference gets improved effectively. So this paper extracted the region of interest through background subtraction and the frame difference method. After eroding and dilating the binary image, the image with black background and a full internal structure of vehicles is gotten. Then the contour of binary image is extracted and the bounding rectangle of the contours is calculated. If the bounding rectangle meets certain threshold conditions, it is considered to be vehicle and would be saved to the container which has been initialized. If the bounding rectangle does not satisfy the threshold conditions, it is not vehicle and may be pedestrians, road signs and other interferences and so on. The vehicle detection and identification are completed.
In the stage of vehicle tracking, mass center of vehicle reflects the position of the vehicle, and it is the key to extract the mass center correctly. It can not be too large in the adjacent two frames. Color is a feature of vehicles. After establishing the color probability model of vehicle and doing normalized histogram, two cars can be distinguished through comparing the color histogram distance. This article proposed a multi-feature matching method to track vehicles. Rationally extract mass center and color of the vehicle, and then compare the contour with vehicles in containers. If the value is matched, the contour is considered to the same vehicle and then set the same serial number. If the value is not matched, the contour is considered to be a new car and set a new serial number for the car. Multi-feature matching method is proved to be simple and effective.
There are different characteristics in illegal behavior of different vehicles. It is the most critical issues to detect and identify the illegal vehicles accurately and timely. In this paper, illegal parking, vehicle reverse driving, illegal behavior of vehicles are studied. The steps are as follows:the system detects whether there are the appearance of moving objects in the region. If so. further confirm whether it is the vehicles moving target of traffic. And then use the appropriate algorithm to determine what kind of illegal behavior in the monitor area that it belongs to and record the appropriate information for subsequent processing.
In summary, this paper did follow researches on detection and identification vehicles, tracking and violation detection. In the detection and identification of the vehicle stage, the article proposes to detect vehicles with extracting contours method in the binary image. In the stage of vehicle tracking, the article proposes multi-feature matching method to track the vehicle movements in the video, At last, this paper analyzes and studies illegal parking, vehicle reverse driving, illegal lane changed.
本文对基于视频的运动车辆检测识别、跟踪和违章检测等方面做了分析研究并进行了大量的实验,使用Visual C++ 6.0作为开发环境,并结合Intel公司的OpenCV图像处理库进行开发。通过使用OpenCV技术,实时的捕捉视频流数据,对图像进行灰度化和平滑处理,用背景差和帧间差相结合的方法得到感兴趣区域后,进行二值化处理,此时得到的图像黑色为背景,白色为前景运动目标。
在车辆检测识别阶段,本文采用了轮廓提取法提取二值图中的轮廓,然后计算轮廓的外接矩形,对满足一定条件的矩形认为是车辆,用矩形框圈起来这就完成了对视频中的车辆的检测和识别。在车辆跟踪阶段,提出了颜色和质心距离相结合的多特征匹配算法来跟踪已经检测识别出的车辆。分类研究了交通违章事件等行为,分别对违章停车、车辆逆向行驶、违章变道压线等车辆违章行为做了实验,检测识别出视频中的车辆违规事件并对违章信息进行管理。
At present, motor vehicles are increasing. This causes traffic problems worse and often generates a number of traffic accidents. Man-made illegal events include running red lights, speeding, illegal lane change. Intelligent Transportation System is an intelligent system, through the real-time detection of road traffic flow, it can detect and track vehicles accurately-According to the traffic flow on the road, the system can response quickly. It can reduce road congestion, ease the traffic pressure and reduce the accident rate through taking certain measures. Currently, the intelligent traffic system has been used in vehicle detection and tracking, calculating speed.
This paper makes a lot of experiments on motion detection and identification of vehicles. Using Visual C++ 6.0 as the development environment, combined with OpenCV image processing library for development. We can play the video stream and read a frame of the video image by OpenCV. In the process of studying algorithm, the image processing functions is provided by OpenCV library.
After capturing video frame by OpenCV library functions, the color space is converted and the video image is processed. The background subtraction and frame difference have advantages and disadvantages. If the combination of these two methods, the performance was complementary, it can not only detect the movement of vehicles and also detect stationary vehicles. At the same time. the result of frame difference gets improved effectively. So this paper extracted the region of interest through background subtraction and the frame difference method. After eroding and dilating the binary image, the image with black background and a full internal structure of vehicles is gotten. Then the contour of binary image is extracted and the bounding rectangle of the contours is calculated. If the bounding rectangle meets certain threshold conditions, it is considered to be vehicle and would be saved to the container which has been initialized. If the bounding rectangle does not satisfy the threshold conditions, it is not vehicle and may be pedestrians, road signs and other interferences and so on. The vehicle detection and identification are completed.
In the stage of vehicle tracking, mass center of vehicle reflects the position of the vehicle, and it is the key to extract the mass center correctly. It can not be too large in the adjacent two frames. Color is a feature of vehicles. After establishing the color probability model of vehicle and doing normalized histogram, two cars can be distinguished through comparing the color histogram distance. This article proposed a multi-feature matching method to track vehicles. Rationally extract mass center and color of the vehicle, and then compare the contour with vehicles in containers. If the value is matched, the contour is considered to the same vehicle and then set the same serial number. If the value is not matched, the contour is considered to be a new car and set a new serial number for the car. Multi-feature matching method is proved to be simple and effective.
There are different characteristics in illegal behavior of different vehicles. It is the most critical issues to detect and identify the illegal vehicles accurately and timely. In this paper, illegal parking, vehicle reverse driving, illegal behavior of vehicles are studied. The steps are as follows:the system detects whether there are the appearance of moving objects in the region. If so. further confirm whether it is the vehicles moving target of traffic. And then use the appropriate algorithm to determine what kind of illegal behavior in the monitor area that it belongs to and record the appropriate information for subsequent processing.
In summary, this paper did follow researches on detection and identification vehicles, tracking and violation detection. In the detection and identification of the vehicle stage, the article proposes to detect vehicles with extracting contours method in the binary image. In the stage of vehicle tracking, the article proposes multi-feature matching method to track the vehicle movements in the video, At last, this paper analyzes and studies illegal parking, vehicle reverse driving, illegal lane changed.
引文
[1]葛广英.智能交通系统中的视频监控技术[J].电视技术,2006,286(4)：89-92.
[2]潘秦华.车辆目标检测与交通流量检测系统的研究[D].西安电子科技大学.2005.
[3]T.Abramczuk.A Microcomputer Based TV Detector for Road Traffic[J]. In Symposium on Road Research Program,T.1984.3(2):145-147.
[4]马晓宇.基于视频分割与跟踪算法的车流量统计[D].浙江大学.2008.
[5]魏武,张起森,王明俊,黄中祥.基于计算机视觉和图像处理的交通参数检测[J].信息与控制,2001.30(3)：257-261.
[6]张丽.车辆视频检测与跟踪系统的算法研究[D].浙江大学.2003.
[7]李毅,孙正兴,远博,张岩.一种改进的帧差和背景减相结合的运动检测方法[J].中国图象图形学报,2009,14(6)：1162-1168.
[8]Elgammal A. Harwood D, Davis L S. Non-parametric model for background subtraction[C]. Proceedings of the 6th European Conference on Computer Vision.2000: 751-767.
[9]Wren C. Azarbayejani A. Darrell T, Pentland A. Pfinder:Real-Time tracking of the human body [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence.1997. 19(7):780-785.
[10]Stauffer C. Grimson W. Adaptive background mixture models for real-time tracking[C]. Proceedings of IEEE International Conference on Computer Vision and Pattern Recognition. Fort Collins:IEEE Press.1999.2:246-252.
[11]Stauffer C, Grimson W. Learning patterns of activity using real-time tracking[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence.2000.22(8):747-757.
[12]李江春.运动目标检测与跟踪算法的研究与实现[D].吉林大学.2009.
[13]Benjamin Coifnan et al.a real-time computer vision system for vehicle tracking and traffic surveillance[J].Transportation Research PartC:Emerging Technologies. 1998.6 (4):271-288.
[14]Grinias I. Tziritas G, A semi-automatic seeded region growing algorithm for video object localization and tracking[J]. Signal Processing:Image Communication 2001.16(10): 977-986.
[15]WangYue. Teoh Earn Khwang. Shen Dinggang. Lane detection and tracking using B-snake[J]. Image and Vision Computing.2004.22(4):269-280.
[16]KimWon,Lee Choon-Young.Lee Ju-Jang.Tracking moving object using Snake's jump based on image flow[J].Mechatronics.2001.11(2):199-226.
[17]刘瑞祯.于仕琪. OpenCV教程-基础篇[M].北京：北京航空航天大学出版社.2007，6.
[18]Takaba S. A Traffic Flow Measuring System Using a Solid State Sensor[C]. Proc IEEE Conf on Road Traffic Data Collection, London, UK,1984:45-56.
[19]Paragios N, Deriche R.Geodesic active contours and level sets for the detection and tracking of moving objects[J].IEEE Transactions on Pattern Analysis and Machine Interface.2000.22 (3):266-280.
[20]Fejes S, Davis L S. What can projections of flow fields tell us about the visual motion In: Proceeding of International Conference on Computer Vision[C].Bombay,India.1998: 979-986.
[21]MiIIan Sonka, Vaclav Hlavac, Roger Boyle(艾海舟,武勃等译).图象处理分析与机器视觉[M].第二版.北京：人民邮电出版社,2003：469-491.
[22]Barron J L.Fleet D J.Beauehemin S S.Systems and Experiment Performance of Optical Flow Techniques [J]. International Journal of Computer Vision.1994.12(1):43-77.
[23]王洪建.李志敏,基于视频图像的车辆流量实时检测系统[J].光学精密工程,2005,13：222-225.
[24]Hoose N, IMPACT:An Image Analysis Tool for Motor Analysis and Surveillance[J]. Traffic Engineering Control.1992,23(4):140-147.
[25]Michalopoulos P G. Vehicle Detection Video Through Image Processing:The Autoscope System[J]. IEEE Trans on Vehicular Technology,1991.40(1):21-29.
[26]赵发科,施毅.车辆跟踪中的背景初始化与更新方法研究[J].交通信息与安全,2009,,27(4)：16-21.
[27]刘玉秋.基于视频流的车辆检测算法[D].吉林大学.2008.
[28]张便利,常胜江,李江卫,王凯.基于彩色直方图分析的智能视频监控系统[J].物理学报,2006,55(12)：6399-6403.
[29]刘雪,常发亮,王华杰.基于改进Camshift算法的视频对象跟踪方法[J].图像处理,2007,7(3)：156-159.
[30]严捷丰,陈功,刘学亮,周荷琴.一直视频检测车辆位置违章的几何方法[J].小型微型计算机系统,2009,3(3)：498-502.
[31]王宁.基于视频的车辆跟踪与交通事件检测[D].四川大学.2005.
[32]刘晓薇,胡振民,余鹤龄.违章车辆视频检测算法的研究[J].华东交通大学学报,2007,,24(5)：64-66.
[33]刘肃亮.交通车辆违章智能视频监控系统的设计与实现[D].西北大学.2004.
[34]肖习雨,张昌凡,龙永红.基于视频的车辆违章监测方法[J].湖南工业大学学报,2009,23(6)：20-23.
[2]潘秦华.车辆目标检测与交通流量检测系统的研究[D].西安电子科技大学.2005.
[3]T.Abramczuk.A Microcomputer Based TV Detector for Road Traffic[J]. In Symposium on Road Research Program,T.1984.3(2):145-147.
[4]马晓宇.基于视频分割与跟踪算法的车流量统计[D].浙江大学.2008.
[5]魏武,张起森,王明俊,黄中祥.基于计算机视觉和图像处理的交通参数检测[J].信息与控制,2001.30(3)：257-261.
[6]张丽.车辆视频检测与跟踪系统的算法研究[D].浙江大学.2003.
[7]李毅,孙正兴,远博,张岩.一种改进的帧差和背景减相结合的运动检测方法[J].中国图象图形学报,2009,14(6)：1162-1168.
[8]Elgammal A. Harwood D, Davis L S. Non-parametric model for background subtraction[C]. Proceedings of the 6th European Conference on Computer Vision.2000: 751-767.
[9]Wren C. Azarbayejani A. Darrell T, Pentland A. Pfinder:Real-Time tracking of the human body [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence.1997. 19(7):780-785.
[10]Stauffer C. Grimson W. Adaptive background mixture models for real-time tracking[C]. Proceedings of IEEE International Conference on Computer Vision and Pattern Recognition. Fort Collins:IEEE Press.1999.2:246-252.
[11]Stauffer C, Grimson W. Learning patterns of activity using real-time tracking[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence.2000.22(8):747-757.
[12]李江春.运动目标检测与跟踪算法的研究与实现[D].吉林大学.2009.
[13]Benjamin Coifnan et al.a real-time computer vision system for vehicle tracking and traffic surveillance[J].Transportation Research PartC:Emerging Technologies. 1998.6 (4):271-288.
[14]Grinias I. Tziritas G, A semi-automatic seeded region growing algorithm for video object localization and tracking[J]. Signal Processing:Image Communication 2001.16(10): 977-986.
[15]WangYue. Teoh Earn Khwang. Shen Dinggang. Lane detection and tracking using B-snake[J]. Image and Vision Computing.2004.22(4):269-280.
[16]KimWon,Lee Choon-Young.Lee Ju-Jang.Tracking moving object using Snake's jump based on image flow[J].Mechatronics.2001.11(2):199-226.
[17]刘瑞祯.于仕琪. OpenCV教程-基础篇[M].北京：北京航空航天大学出版社.2007，6.
[18]Takaba S. A Traffic Flow Measuring System Using a Solid State Sensor[C]. Proc IEEE Conf on Road Traffic Data Collection, London, UK,1984:45-56.
[19]Paragios N, Deriche R.Geodesic active contours and level sets for the detection and tracking of moving objects[J].IEEE Transactions on Pattern Analysis and Machine Interface.2000.22 (3):266-280.
[20]Fejes S, Davis L S. What can projections of flow fields tell us about the visual motion In: Proceeding of International Conference on Computer Vision[C].Bombay,India.1998: 979-986.
[21]MiIIan Sonka, Vaclav Hlavac, Roger Boyle(艾海舟,武勃等译).图象处理分析与机器视觉[M].第二版.北京：人民邮电出版社,2003：469-491.
[22]Barron J L.Fleet D J.Beauehemin S S.Systems and Experiment Performance of Optical Flow Techniques [J]. International Journal of Computer Vision.1994.12(1):43-77.
[23]王洪建.李志敏,基于视频图像的车辆流量实时检测系统[J].光学精密工程,2005,13：222-225.
[24]Hoose N, IMPACT:An Image Analysis Tool for Motor Analysis and Surveillance[J]. Traffic Engineering Control.1992,23(4):140-147.
[25]Michalopoulos P G. Vehicle Detection Video Through Image Processing:The Autoscope System[J]. IEEE Trans on Vehicular Technology,1991.40(1):21-29.
[26]赵发科,施毅.车辆跟踪中的背景初始化与更新方法研究[J].交通信息与安全,2009,,27(4)：16-21.
[27]刘玉秋.基于视频流的车辆检测算法[D].吉林大学.2008.
[28]张便利,常胜江,李江卫,王凯.基于彩色直方图分析的智能视频监控系统[J].物理学报,2006,55(12)：6399-6403.
[29]刘雪,常发亮,王华杰.基于改进Camshift算法的视频对象跟踪方法[J].图像处理,2007,7(3)：156-159.
[30]严捷丰,陈功,刘学亮,周荷琴.一直视频检测车辆位置违章的几何方法[J].小型微型计算机系统,2009,3(3)：498-502.
[31]王宁.基于视频的车辆跟踪与交通事件检测[D].四川大学.2005.
[32]刘晓薇,胡振民,余鹤龄.违章车辆视频检测算法的研究[J].华东交通大学学报,2007,,24(5)：64-66.
[33]刘肃亮.交通车辆违章智能视频监控系统的设计与实现[D].西北大学.2004.
[34]肖习雨,张昌凡,龙永红.基于视频的车辆违章监测方法[J].湖南工业大学学报,2009,23(6)：20-23.