一种新的车辆遮挡检测与分割方法

New method for detecting and dividing vehicles blocking

遮挡是图像处理和运动物体跟踪过程中比较难解决的问题。针对交通视频中车辆遮挡问题,提出了一种基于统计模型的车辆遮挡分割算法。该算法在提取出运动车辆的基础上,使用横向-纵向扫描方法对运动车辆前景模板进行去空洞处理,得到完整的运动车辆区域;利用统计模型判断是否发生遮挡,如果判断车辆发生遮挡,在加入纠错机制的前提下得到正确的遮挡点,同时确定遮挡区域;利用边缘提取方法分割出遮挡区域,得到完全分隔的车辆。实验结果表明,该方法能够很好地解决车辆部分遮挡问题。

Blocking is a difficult problem to solve in the process of image processing and moving object tracking. Aiming at the problem for vehicles blocking in traffic videos, this paper proposes a vehicles blocking segmentation algorithm based on a statistical model. To get the whole area of moving vehicles, the algorithm removes empty space in the foreground template, using the horizontal-vertical scanning method, based on the extraction of moving vehicles. Then the occurrence of blocking is determined by the statistical model. If the vehicles are blocked, the method will get the right blocking points and identify the blocking regions under the premise of adding error correc- tion mechanism. Finally, it has divided out the blocking area by edge detection, so as to obtain the completely sepa- rated vehicles. Besides, the experimental results show that this method is able to well solve the problem of partial blocking of vehicles.