基于全局运动特征的地铁乘客异常行为检测
|
摘要
城市轨道交通中,车站站台是乘客候车长时间滞留的公共场所。突发的异常事件往往会引起乘客的恐慌,造成盲目奔跑甚至带来更严重的踩踏事件。以城市轨道交通乘客打架斗殴异常事件为研究对象,以城市轨道交通现行CCTV视频监控系统为数据源,基于金字塔改进型L-K光流的计算方法,针对视频帧图像中有效点像素运动矢量分析运动强度和方向两大类指标,提取图像全局运动特征,并获取打架斗殴行为阈值测试范围,从而有效、正确地实现该种行为的异常判断,为达到智能分析实时检测打架斗殴异常行为并报警提供具有可行性的理论基础和技术依据,从而减轻工作人员的劳动强度、强化工作效率,提高城市轨道交通运行组织水平。
In urban rail transit,station platform is a public place for passengers to wait for a long time.Unexpected events often cause panic among passengers,causing a blind run and even more serious stampede.The study on the abnormality of fights in urban rail transit was taking as the object of study and the current CCTV video surveillance system of urban rail transit was taking as the data source.Based on the improved L-K optical flow calculation method in Pyramid,we extracted the global motion characteristics of the images and got the threshold range of the fighting behavior based on the two index of effective point pixel motion vector in the video frame image,and analyzed the global motion characteristics of the image,so as to effectively and correctly realized the abnormal behavior of this kind of behavior and provide a feasible theoretical basis and technical basis to achieve intelligent analysis,real-time detection of abnormal behavior and alarm.This study can reduce the labor intensity of the staff,strengthen the working efficiency and improve the organization level of urban rail transit operation.
In urban rail transit,station platform is a public place for passengers to wait for a long time.Unexpected events often cause panic among passengers,causing a blind run and even more serious stampede.The study on the abnormality of fights in urban rail transit was taking as the object of study and the current CCTV video surveillance system of urban rail transit was taking as the data source.Based on the improved L-K optical flow calculation method in Pyramid,we extracted the global motion characteristics of the images and got the threshold range of the fighting behavior based on the two index of effective point pixel motion vector in the video frame image,and analyzed the global motion characteristics of the image,so as to effectively and correctly realized the abnormal behavior of this kind of behavior and provide a feasible theoretical basis and technical basis to achieve intelligent analysis,real-time detection of abnormal behavior and alarm.This study can reduce the labor intensity of the staff,strengthen the working efficiency and improve the organization level of urban rail transit operation.
引文
[1]Palwasha A,Paulo C,Henrique S.AutomaticVisual Detection of Human Behavior:A Review from2000 to 2014[J].Expert Systems withApplications,2015:6935-6956.
[2]WANG Xiao-fei,GAO Ming-liang,HE Xiao-hai,et al.An Abnormal Crowd Behavior Detection Algorithm based on Fluid Mechanics[J].Journal of Computer,2014,9(5):1144-1149.
[3]FU Bo,LI Wen-hui,Chen B,et al.AbnormalBehavior Detection based on Weighted Energy of Optical Flow[J].Journal of Jilin University,2013,43(6):1644-1649.
[4]朱海龙,刘鹏,刘家锋,等.人群异常状态检测的图分析方法[J].自动化学报,2012,38(5):742-750.
[5]XIONG Guo-gang,WU Xin-yu,CHENYe-lun,et al.AbnormalCrowd BehaviorDetection based on the Energy Model[C].Information and Automation(ICIA),2011IEEE International Conference on Date of Conference,2011:495-500.
[6]CUI Xin-yi,LIU Qing-shan,GAO Ming-chen,et al.Abnormal Detection Using Interaction Energy Potentials[C].The 24th IEEE Conference on Computer Vision and Pattern Recognition,2011:3161-3167.
[7]Simon B,Daniel S,Lewis J P,et al.A Database and Evaluation Methodology for Optical Flow[J].International Journal of Computer Vision,2011,92(1):1-31.
[8]刘赏,董林芳.人群运动方向异常检测算法[J].计算机科学,2013,40(11):337-340.
[9]Horn B K P,Schunck BG.Determining Optical Flow[J].Artificial Intelligence,1980,17(81):185-203.
[10]Bruhn A,Weickert J,Schn9rrC.Lucas/Kanade Meets Horn/Schunck:Combining Local and Global Optic Flow Methods[J].International Journal of Computer Vision,2005,61(3):211-231.
[11]Eero P,Simoncelli,Edward H.Probability D,Stributions of Optical Flow[C].IEEE Conference on Computer Vision and Pattern Recognition,1991:310-315.
[12]Baker S.Lucas-kanade 20 Years on:A Unifiying Framework[J].International Journal of Computer Vision,2004,56(3):221-255.
[13]Bouguet J Y.Pyramidal Implementation of the Lucas Kanade Feature Tracker Description of the Algorithm[J].Acta Pathologica Japonicat,2000,22(2):363-381.
[2]WANG Xiao-fei,GAO Ming-liang,HE Xiao-hai,et al.An Abnormal Crowd Behavior Detection Algorithm based on Fluid Mechanics[J].Journal of Computer,2014,9(5):1144-1149.
[3]FU Bo,LI Wen-hui,Chen B,et al.AbnormalBehavior Detection based on Weighted Energy of Optical Flow[J].Journal of Jilin University,2013,43(6):1644-1649.
[4]朱海龙,刘鹏,刘家锋,等.人群异常状态检测的图分析方法[J].自动化学报,2012,38(5):742-750.
[5]XIONG Guo-gang,WU Xin-yu,CHENYe-lun,et al.AbnormalCrowd BehaviorDetection based on the Energy Model[C].Information and Automation(ICIA),2011IEEE International Conference on Date of Conference,2011:495-500.
[6]CUI Xin-yi,LIU Qing-shan,GAO Ming-chen,et al.Abnormal Detection Using Interaction Energy Potentials[C].The 24th IEEE Conference on Computer Vision and Pattern Recognition,2011:3161-3167.
[7]Simon B,Daniel S,Lewis J P,et al.A Database and Evaluation Methodology for Optical Flow[J].International Journal of Computer Vision,2011,92(1):1-31.
[8]刘赏,董林芳.人群运动方向异常检测算法[J].计算机科学,2013,40(11):337-340.
[9]Horn B K P,Schunck BG.Determining Optical Flow[J].Artificial Intelligence,1980,17(81):185-203.
[10]Bruhn A,Weickert J,Schn9rrC.Lucas/Kanade Meets Horn/Schunck:Combining Local and Global Optic Flow Methods[J].International Journal of Computer Vision,2005,61(3):211-231.
[11]Eero P,Simoncelli,Edward H.Probability D,Stributions of Optical Flow[C].IEEE Conference on Computer Vision and Pattern Recognition,1991:310-315.
[12]Baker S.Lucas-kanade 20 Years on:A Unifiying Framework[J].International Journal of Computer Vision,2004,56(3):221-255.
[13]Bouguet J Y.Pyramidal Implementation of the Lucas Kanade Feature Tracker Description of the Algorithm[J].Acta Pathologica Japonicat,2000,22(2):363-381.