运动车辆检测水平集提取的研究
|
摘要
运动目标的检测、提取及流量统计是计算机图像领域研究的重点，有着非常重要的应用价值。作为该领域最值得研究的内容之一，图像处理技术是实现运动目标检测及流量统计的一种有效方法，引起了各国研究人员的广泛关注。
本文对常见运动目标检测的方法进行了分析，针对传统帧间差分法在光照、天气变化明显时的不适应性，提出了一种基于C-V模型的运动目标水平集提取新方法：首先，对运动区域进行初始检测，在对帧间差分法进行改进的基础上通过相邻视频帧的相减，选用自适应阈值判断出当前视频帧中的运动目标像素；其次，将检测出的运动目标区域经形态学处理后通过定义最小化能量函数构建运动目标轮廓提取的水平集C-V模型，实现运动目标轮廓的提取。实验结果表明，该方法的边缘准确率和检出率更高，能够更有效地提取运动目标。
在利用上述方法检测到运动车辆的基础上，对多车道虚拟检测线车流量统计的方法进行了改进。它使用单条虚拟检测线技术实现了多车道情况下的车流量统计，能有效改善跨道行驶的车流量统计问题。作为实验验证，开发了一个车流量统计系统，该系统分为4个模块：视频载入模块、虚拟检测线设置模块、视频控制模块、车流量统计模块。实验结果表明，本文所提方法具有很好的实时性，识别率较高。
Motion object detection, extraction and traffic flow statistics is a hotresearch in image domain of computer, and has very important applicationvalues. As the most worthy of study content, image processing technology isthe realization of motion detection and flow statistics is an effective method,it has aroused the extensive attention of researchers.
In this paper, the common methods of target detection are analyzed, amotion object extraction algorithm based on level set C-V model is proposedin view of the traditional frame difference method in light, weather changesapparent inelasticity: firstly, on the moving area in initial detection, framedifference algorithm was improved and adjacent moving object pixels incurrent image were detected through adaptive threshold and adjacent videoframe subtracting technique. The proposed approach can improve the changeof light and weather effective; secondly, Construct and realize the level setC-V model of extracting the moving object contour though being processedby morphology after motion object detection. Experimental results show thatthe proposed approach for extracting moving object, which edge anddetecting accuracy rate is more powerful and can effectively extract themoving object.
Based on the above method to detect moving vehicles, multi-lane testline traffic flow statistics method was improved. It uses a single virtualdetection line technology to realize traffic flow statistics of multi lane, andimprove the cross road traffic flow statistics effectively. As an experiment,develop a traffic flow statistics system, this system divided into four modules,which are video load module, virtual detecting line set module, video controlmodule, traffic flow statistics module. Experimental results show that theproposed method has the following characteristics: the real-time property andhigh recognize rate.
本文对常见运动目标检测的方法进行了分析，针对传统帧间差分法在光照、天气变化明显时的不适应性，提出了一种基于C-V模型的运动目标水平集提取新方法：首先，对运动区域进行初始检测，在对帧间差分法进行改进的基础上通过相邻视频帧的相减，选用自适应阈值判断出当前视频帧中的运动目标像素；其次，将检测出的运动目标区域经形态学处理后通过定义最小化能量函数构建运动目标轮廓提取的水平集C-V模型，实现运动目标轮廓的提取。实验结果表明，该方法的边缘准确率和检出率更高，能够更有效地提取运动目标。
在利用上述方法检测到运动车辆的基础上，对多车道虚拟检测线车流量统计的方法进行了改进。它使用单条虚拟检测线技术实现了多车道情况下的车流量统计，能有效改善跨道行驶的车流量统计问题。作为实验验证，开发了一个车流量统计系统，该系统分为4个模块：视频载入模块、虚拟检测线设置模块、视频控制模块、车流量统计模块。实验结果表明，本文所提方法具有很好的实时性，识别率较高。
Motion object detection, extraction and traffic flow statistics is a hotresearch in image domain of computer, and has very important applicationvalues. As the most worthy of study content, image processing technology isthe realization of motion detection and flow statistics is an effective method,it has aroused the extensive attention of researchers.
In this paper, the common methods of target detection are analyzed, amotion object extraction algorithm based on level set C-V model is proposedin view of the traditional frame difference method in light, weather changesapparent inelasticity: firstly, on the moving area in initial detection, framedifference algorithm was improved and adjacent moving object pixels incurrent image were detected through adaptive threshold and adjacent videoframe subtracting technique. The proposed approach can improve the changeof light and weather effective; secondly, Construct and realize the level setC-V model of extracting the moving object contour though being processedby morphology after motion object detection. Experimental results show thatthe proposed approach for extracting moving object, which edge anddetecting accuracy rate is more powerful and can effectively extract themoving object.
Based on the above method to detect moving vehicles, multi-lane testline traffic flow statistics method was improved. It uses a single virtualdetection line technology to realize traffic flow statistics of multi lane, andimprove the cross road traffic flow statistics effectively. As an experiment,develop a traffic flow statistics system, this system divided into four modules,which are video load module, virtual detecting line set module, video controlmodule, traffic flow statistics module. Experimental results show that theproposed method has the following characteristics: the real-time property andhigh recognize rate.
引文
[1]郝旭.智能交通系统的社会经济效益分析[J].科技与经济，2004,4(16):36-37.
[2] Stauffer C, Grimson E. Learning patterns of activity using real-time tracking[J].IEEE Transactions on Pattern Recognition and Machine Intelligence,2000,22(8):747-757.
[3] Wren C, Azarbayejani A, Darrell T, et al. Real-time tracking of the human body [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1997,19(7):780-785.
[4] Stauffer C, Grimson W E L. Adaptive background mixture models for Real-timetracking[C]//IEEE Computer Society Conference on Computer Vision and PatternRecognition, Fort. Collins, USA,1999:246-252.
[5]代科学,李国辉,涂丹,等.监控视频运动目标检测减背景技术的研究现状和展望[J].中国图象图形学报,2006,11(7):919-927.
[6] Fejes S, Davis L S. What can projections of flow fields tell us about the visualmotion[C]//International Conference on Computer Vision,Bombay,India,1998:979-986.
[7] Lipton A, Fujiyoshi H, Patil R. Moving target classification and tracking fromreal-time video[C]//IEEE Workshop on Applications of Computer Vision,Princeton,NJ,USA,1998:8-l4.
[8]朱付平,田捷,林瑶,葛行飞.基于Level Set方法的医学图像分割[J].软件学报,2002,13(9):1866-1872.
[9]钱芸,张英杰.水平集的图像分割方法综述[J].中国图象图形学报.2008,13(1).7-13.
[10] Li C, Xu C, Gui C, et al. Level set evolution without re-initialization: a new variationalformulation[C]. IEEE Computer Society Conference on Computer Vision and PatternRecognition. San Diego, CA,USA:IEEE Press,2005,1：430-436.
[11] Stanley Osher and Ronald Fedkiw. Level Set Methods and Dynamic Implicit Surfaces[M]. Springer-Verlag,2002.
[12] Osher, S., and Sethian, J.A., Fronts Propagating with Curvature-Dependent Speed:Algorithms Based on Hamilton–Jacobi Formulations[J]. Journal of ComputationalPhysics.1988,79(1):12–49.
[13] D. Adalsteinsson, J. Sethian. A Fast Level Set Method for Propagating Interface[J].Journal of Computational Physics.1995,118:269-277.
[14] Vese L, Chan T. A Multiphase Level Set Framework for Image Segmentation Using theMumford and Shah Model[J]. International Journal of Computer Vision,2002,50(3):271-293.
[15]王怡,周明全,耿国华.基于简化Mumford-Shah模型的水平集图像分割算法[J].计算机应用,2006,26(8):1848-1850．
[16] Chan T, Vese L. Active Contours Without Edges[J]. IEEE Trans. on Imag. Proc.,2001,10(2):266-277.
[17] Caselles V, Sapiro G, Kimmel R. Geodesic active contours[J]. International Journal ofComputer Vision.1997,22(1)：61-79.
[18] Chan T, Vese L. Active contours without edges[J]. IEEE Transactions on ImageProcessing,2001,10(2):266-277.
[19]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.计算机应用研究,2005,9:9-14.
[20]张晖,董育宁,基于视频的车辆检测算法综述[J].南京邮电大学学报:自然科学版,2007,27(3):88-94.
[21]焦波,李国辉,涂丹.一种用于运动目标检测的快速收敛混合高斯模型[J].中国图象图形学报，2008，13(11):2139—2143.
[22]刘静,王玲.混合高斯模型背景法的一种改进算法[J]．计算机工程与应用，2010,46(13):168—170.
[23]陈祖爵,陈潇君,何鸿.基于改进的混合高斯模型的运动目标检测[J].中国图象图形学报,2007,12(9):1585—1589.
[24] Spagnolo P, DOrazio T, LeoM, et al. Moving object segmentation by backgroundsubtraction and temporal analysis[J]. Image and Vision Computing,2006,24(5):411-423.
[25]李刚,邱尚斌,林凌,曾锐利.基于背景差法和帧间差法的运动目标检测方法.仪器仪表学报,2006,27(8):961-964.
[26]胡小锋,赵辉.Visual C++/MATLAB图像处理与识别实用案例精选[M].北京:人民邮电出版社,2004:93-100.
[27]冈萨雷斯.数字图像处理:MATLAB版[M].北京:电子工业出版社,2006.
[28]王晓丹,吴崇明.基于MATLAB的系统分析与设计—图像处理.西安:西安电子科技大学出版社.2000,9:204.
[29]杨莉,杨新.基于水平集的多运动目标分割[J].上海交通大学学报,2004,38(5):713-717.
[30]张磊,项学智,赵春晖,等.基于光流场与水平集的动目标检测[J].计算机应用,2009,29(4):972-978.
[31]张建国,张荣国,薛菲,刘焜.基于水平集方法的肿瘤图像识别[J].太原科技大学学报,2012,33(1):23-25.
[32]张荣国,刘小君,党伟超,刘焜.多目标轮廓Mumford-Shah水平集提取[J].智能系统学报,2011,6(4):360-366.
[33]张开华，周文罡，张振，等.一种改进的C-V主动轮廓模型[J].光电工程,2008，35(12)：1-6.
[34]于慧敏,尤育赛.基于水平集的多运动目标检测和分割[J].浙江大学学报:工学版,2007,41(3):412-417.
[35]王欢,任明武,杨静宇,等.一种区域级运动目标检测算法[J].模式识别与人工智能,2009,22(5):689-696.
[36]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,2004,27(5):31-33.
[37]罗欣,朱清新.改进的基于边缘检测技术的车流量统计系统[J].计算机工程,2006,32(9):228-232.
[38]张永丽,张太镒,毕建民.基于自适应背景初始化的车流量检测算法[J].微电子学与计算机,2007,24(5):138-140.
[2] Stauffer C, Grimson E. Learning patterns of activity using real-time tracking[J].IEEE Transactions on Pattern Recognition and Machine Intelligence,2000,22(8):747-757.
[3] Wren C, Azarbayejani A, Darrell T, et al. Real-time tracking of the human body [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1997,19(7):780-785.
[4] Stauffer C, Grimson W E L. Adaptive background mixture models for Real-timetracking[C]//IEEE Computer Society Conference on Computer Vision and PatternRecognition, Fort. Collins, USA,1999:246-252.
[5]代科学,李国辉,涂丹,等.监控视频运动目标检测减背景技术的研究现状和展望[J].中国图象图形学报,2006,11(7):919-927.
[6] Fejes S, Davis L S. What can projections of flow fields tell us about the visualmotion[C]//International Conference on Computer Vision,Bombay,India,1998:979-986.
[7] Lipton A, Fujiyoshi H, Patil R. Moving target classification and tracking fromreal-time video[C]//IEEE Workshop on Applications of Computer Vision,Princeton,NJ,USA,1998:8-l4.
[8]朱付平,田捷,林瑶,葛行飞.基于Level Set方法的医学图像分割[J].软件学报,2002,13(9):1866-1872.
[9]钱芸,张英杰.水平集的图像分割方法综述[J].中国图象图形学报.2008,13(1).7-13.
[10] Li C, Xu C, Gui C, et al. Level set evolution without re-initialization: a new variationalformulation[C]. IEEE Computer Society Conference on Computer Vision and PatternRecognition. San Diego, CA,USA:IEEE Press,2005,1：430-436.
[11] Stanley Osher and Ronald Fedkiw. Level Set Methods and Dynamic Implicit Surfaces[M]. Springer-Verlag,2002.
[12] Osher, S., and Sethian, J.A., Fronts Propagating with Curvature-Dependent Speed:Algorithms Based on Hamilton–Jacobi Formulations[J]. Journal of ComputationalPhysics.1988,79(1):12–49.
[13] D. Adalsteinsson, J. Sethian. A Fast Level Set Method for Propagating Interface[J].Journal of Computational Physics.1995,118:269-277.
[14] Vese L, Chan T. A Multiphase Level Set Framework for Image Segmentation Using theMumford and Shah Model[J]. International Journal of Computer Vision,2002,50(3):271-293.
[15]王怡,周明全,耿国华.基于简化Mumford-Shah模型的水平集图像分割算法[J].计算机应用,2006,26(8):1848-1850．
[16] Chan T, Vese L. Active Contours Without Edges[J]. IEEE Trans. on Imag. Proc.,2001,10(2):266-277.
[17] Caselles V, Sapiro G, Kimmel R. Geodesic active contours[J]. International Journal ofComputer Vision.1997,22(1)：61-79.
[18] Chan T, Vese L. Active contours without edges[J]. IEEE Transactions on ImageProcessing,2001,10(2):266-277.
[19]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.计算机应用研究,2005,9:9-14.
[20]张晖,董育宁,基于视频的车辆检测算法综述[J].南京邮电大学学报:自然科学版,2007,27(3):88-94.
[21]焦波,李国辉,涂丹.一种用于运动目标检测的快速收敛混合高斯模型[J].中国图象图形学报，2008，13(11):2139—2143.
[22]刘静,王玲.混合高斯模型背景法的一种改进算法[J]．计算机工程与应用，2010,46(13):168—170.
[23]陈祖爵,陈潇君,何鸿.基于改进的混合高斯模型的运动目标检测[J].中国图象图形学报,2007,12(9):1585—1589.
[24] Spagnolo P, DOrazio T, LeoM, et al. Moving object segmentation by backgroundsubtraction and temporal analysis[J]. Image and Vision Computing,2006,24(5):411-423.
[25]李刚,邱尚斌,林凌,曾锐利.基于背景差法和帧间差法的运动目标检测方法.仪器仪表学报,2006,27(8):961-964.
[26]胡小锋,赵辉.Visual C++/MATLAB图像处理与识别实用案例精选[M].北京:人民邮电出版社,2004:93-100.
[27]冈萨雷斯.数字图像处理:MATLAB版[M].北京:电子工业出版社,2006.
[28]王晓丹,吴崇明.基于MATLAB的系统分析与设计—图像处理.西安:西安电子科技大学出版社.2000,9:204.
[29]杨莉,杨新.基于水平集的多运动目标分割[J].上海交通大学学报,2004,38(5):713-717.
[30]张磊,项学智,赵春晖,等.基于光流场与水平集的动目标检测[J].计算机应用,2009,29(4):972-978.
[31]张建国,张荣国,薛菲,刘焜.基于水平集方法的肿瘤图像识别[J].太原科技大学学报,2012,33(1):23-25.
[32]张荣国,刘小君,党伟超,刘焜.多目标轮廓Mumford-Shah水平集提取[J].智能系统学报,2011,6(4):360-366.
[33]张开华，周文罡，张振，等.一种改进的C-V主动轮廓模型[J].光电工程,2008，35(12)：1-6.
[34]于慧敏,尤育赛.基于水平集的多运动目标检测和分割[J].浙江大学学报:工学版,2007,41(3):412-417.
[35]王欢,任明武,杨静宇,等.一种区域级运动目标检测算法[J].模式识别与人工智能,2009,22(5):689-696.
[36]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,2004,27(5):31-33.
[37]罗欣,朱清新.改进的基于边缘检测技术的车流量统计系统[J].计算机工程,2006,32(9):228-232.
[38]张永丽,张太镒,毕建民.基于自适应背景初始化的车流量检测算法[J].微电子学与计算机,2007,24(5):138-140.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |