交通监控视频中的车辆异常行为检测

Abnormal Vehicle Behavior Detection Using Traffic Surveillance Video

构建了利用交通监控视频对车辆异常行为进行检测的系统框架。使用改进Surendra背景差分与三帧差分相结合的算法进行车辆目标检测,结合Cam Shift算法与Kalman滤波器进行车辆目标跟踪,提取车辆质心绘制运动轨迹,针对车辆运动方向判别、违章变道、调头等行为提出了检测方法。实验结果表明,提出的交通监控视频中的车辆异常行为检测系统具有较高的实时性与准确性,部署简易快速,维护成本低廉,可以满足当今智能交通系统日益增长的需求。

A system framework for abnormal vehicle behavior detection using traffic surveillance video is constructed. Improved Surendra background difference algorithm with the combination of three-frame difference algorithm is introduced for vehicle detection. The combination of CamShift algorithm and Kalman filter are used for vehicle targets tracking. Vehicle centroids are extracted to draw the vehicle trajectory, then detection methods are proposed for movement direction judgment, illegal lane change, turning around, etc. Experimental results show that the proposed system has high real-time and accuracy effects, and can be easily deployed, as well as costs low maintenance, which can meet the growing demand in today＇s intelligent transportation system.