摘要
针对交通道路中车辆的异常行为辨识问题,提出了一种基于轨迹分析的车辆异常行为辨识方法。利用改进的Hausdorff距离计算轨迹的相似度矩阵,根据谱聚类算法学习轨迹的空间分布模式,利用最小平均距离提取运动模式的中心轨迹并根据轨迹的位移向量学习方向模式。在此基础上对新轨迹进行空间与方向模式混合匹配,通过匹配结果检测方向异常的车辆。与此同时,对每类运动模式进行速度特征提取,将95%的行驶的速度作为正常速度区间,5%的行驶的速度作为异常速度区间。通过数据集验证了该方法可以准确地识别出速度异常的车辆,具有一定的实际应用价值。
Aiming at the problem of vehicle abnormal behavior identification in traffic road,a vehicle abnormal behavior identification method based on trajectory analysis is proposed.Firstly,the improved Hausdorff distance is used to calculate the spatial distance similarity matrix of the trajectory.Then,the spatial distribution pattern of the trajectory is learned according to the spectral clustering algorithm.Finally,the center trajectory of the motion pattern is extracted by using the minimum average distance,and the direction pattern is learned according to the displacement vector of the trajectory.On this basis,the new trajectory is mixed with spatial and directional pattern matching,and the vehicles with abnormal direction are detected through the matching results.Meanwhile,the speed feature of the motion mode is extracted.The speed of 95%of the drivers is regarded as the normal speed range,and the speed of 5%of the drivers is regarded as the abnormal speed range.Through the data set verification,the method can accurately identify the abnormal speed of the vehicle,and has a certain practical application value.
作者
丁华
杨文杰
姜超
DING Hua;YANG Wenjie;JIANG Chao(School of Automotive and Traffic Engineering,Jiangsu University,Zhengjiang 212013,China)
出处
《重庆理工大学学报（自然科学）》
CAS
北大核心
2022年第7期62-69,共8页
Journal of Chongqing University of Technology：Natural Science
基金
国家重点研发计划项目(2019YFB1600500)。
关键词
交通工程
轨迹分析
异常行为检测
混合模式匹配
traffic engineering
trajectory analysis
abnormal behavior detection
mixed pattern match
