基于轨迹分析的交通目标异常行为识别

针对交通监控中运动目标的异常行为识别问题,提出一种基于轨迹分析的异常行为识别方法.首先,引入目标的空间位置、运动速度、运动方向和大小尺寸等特征参数对轨迹进行描述和聚类,以提高对目标轨迹的区分和识别能力;然后,提出一种行为识别数据库的建立和调用方法,并以实际交通场景为例,详细说明了数据库的建立和调用过程;最后,采用基于Bayes最优化的方法对轨迹进行联合匹配和边缘匹配,并根据匹配情况调用行为识别数据库对目标行为进行识别.实验结果表明,该方法切实有效,具有一定的实际应用价值.

Traffic Object's Abnormal Behavior Recognition Based on Trajectory Analysis

In view of the problems of recognizing moving object’s abnormal behavior on traffic surveillance,an algorithm on recognizing abnormal behaviors based on trajectory analysis is proposed in this paper.Firstly,four parameters of the object such as space coordinates,velocity,direction and size,are introduced to describe and cluster the trajectories in order to improve the capability of distinguishing and recognizing different objects’ trajectories.Secondly,a method of establishing and calling the database on behavior recognizing is presented in this paper,and an actual traffic scene is taken as an example which explains the establishment and calling process of the database in detail.Finally,a method based on Bayes optimization is adopted to match the trajectory,including union match and edge match,and the database is called according to the match situation to identify the object’s behavior.The experimental results show that this method is effective,and has some practical value.