多车辆跟踪时分割粘连车辆的方法

Method for Touching Vehicles Segmentation in Multi-vehicles Tracking

针对复杂交通系统中的多个车辆粘连问题,提出了一种分割粘连车辆的方法。当两辆车粘连时,分别计算它们的平均灰度值之后取差值,如果该差值大于设定的阈值,则判断它们是水平粘连还是垂直粘连,并确定投影分割的范围,在该范围内采用灰度均值投影的方法确定粘连边界并将它们分割,当灰度相似或者3个及以上的车辆粘连时,根据Kalman预测信息为每个粘连车辆确定分割窗口将它们分割。还提出了一种快速有效的相邻帧中车辆匹配方法。实验表明,文中方法能快速有效地将粘连车辆分割,并且计算复杂度低。

Based on the problem of multi-vehicles touching in complex traffic system, an algorithm is proposed to segment touching vehicles. When two vehicles are touching, if their difference of average gray is bigger than the threshold which is provided, the two vehicles can be judged to be horizontal touching or vertical touching and the line of touching can be identified by means of projection of average gray in the projection area. When two vehicles which have little difference of gray or more than three vehicles are touching, a segmentation window is framed based on the information of Kalman prediction to segment the touch- ing vehicles. Then, a method of matching vehicles in successive frames is proposed which is fast and effective. Experimental results show that this algorithm can segment touching vehicles fast and effectively and has low computational complexity.