交通视频中的Kalman滤波的多车辆跟踪算法

Multiple vehicle tracking algorithm based on Kalman filter in traffic video

提出了一种交通视频中的Kalman滤波的多车辆跟踪算法.该算法利用Kalman滤波器反馈控制系统估计运动状态进行预测和修正,并为运动目标建立模型;利用当前车辆的信息对下一帧目标的位置进行预测,以便缩小目标的搜索范围和搜索时间,从而快速跟踪车辆.利用车辆的外接矩形框大小、质心等特征对车辆进行特征匹配,为交通视频中的车辆建立对应关系,利用新的系统参数更新模型,获得车辆的轨迹,如此反复,从而实现对车辆的跟踪.实验结果表明,此算法运算速度很快,对于车辆这样的快速运动目标,也具有较好的跟踪效果.

A novel method of multiple vehicles tracking algorithm based on Kalman filter in traffic video is proposed in this paper.The Kalman filter is applied to the feedback control system to predict and modify the motion state,as well as to build models for moving objects.The current information of the vehicle is used to forecast its position in the next frame so as to reduce the searching scope and searching time for this object,thus the vehicle can be tracked quickly.The features such as the size of out-connected rectangle frame and centroid of the vehicle are used for feature matching,which can establish the correspondences between vehicles in successive frames.Thereafter,using new system parameters,object models will be updated,and the tracks of vehicles are obtained.So repeatedly,vehicle tracking is achieved.To evaluate the performance of the proposed method,experiments have been conducted on traffic video sequences.Experimental results show this proposed method can operate at a high speed and successfully track rapid moving vehicles.