运动目标空间模式辨识与异常交通行为检测  

Spatial pattern recognition and abnormal traffic behavior detection of moving object

作　　者：胡宏宇[1,2] 王庆年[2] 曲昭伟[1] 李志慧[1] 

机构地区：[1]吉林大学交通学院,长春130022 [2]吉林大学汽车仿真与控制国家重点实验室,长春130022

出　　处：《吉林大学学报（工学版）》2011年第6期1598-1602,共5页Journal of Jilin University:Engineering and Technology Edition

基　　金：'863'国家高技术研究发展计划项目(2009AA11Z210);国家自然科学基金项目(50808092);吉林省科技发展计划项目(20080432);博士后科学基金面上项目(20110491307);吉林大学科学前沿与交叉学科创新项目(201103146)

摘　　要：提出了运动目标的空间运动模式辨识与异常交通行为检测方法。利用改进Hausdor-ff距离的轨迹空间度量方法构建了轨迹集合的空间距离相似度矩阵。根据谱聚类算法学习轨迹的空间分布,提取了运动目标的典型运动模式。在此基础上,提出了基于Bayes分类器的轨迹空间运动模式匹配方法,进而检测异常交通行为。以实际交通场景中的车辆换道行为检测为例,验证了方法的有效性。该方法可以为运动目标交通行为特性以及交通管理控制等方面的研究提供技术支持。The spatial motion pattern recognition method and the abnormal traffic hehavior detection method of the moving object were proposed.The spatial distance similarity matrix of the trajectory dataset was constructed using the mearurement method in the trajectory space with the improved Hausdorff distance.The typical motion pattern of the moving object was extracted by the spatial distributions of the trajectory dataset learnt by the spectral clustering algorithm.A motion pattern matching algorithm in the trajectory space was presented for the abnormal traffic behavior detection based on Bayes classifier.The vehicle lane-changing behavior in the realistic scenario was recognized as an example,and the result was encouraging.The proposed method provides a technical support for the studies of the traffic behavior of moving object and the intelligent traffic management and control.

关 键 词：交通运输工程 视频处理 运动模式辨识 BAYES分类器 

分 类 号：U491[交通运输工程—交通运输规划与管理]