基于轮廓特征点凹性分析的遮挡车辆分割算法

在智能交通系统的拍摄场景中,由于车辆间距过近和摄像角度原因,引起车辆遮挡的现象,增加了目标车辆检测和跟踪的难度。根据轮廓特征点,结合轮廓凹凸性,提出一种凹陷区域检测与分割算法。首先采用背景差分法提取车辆区域,根据车辆区域外接矩形的长宽比和占空比判断是否是多车遮挡,同时通过凸包分析算法提取遮挡凹陷区域;然后通过Freeman链码确定凹陷区域的轮廓特征点,对特征点进行凹性分析;最后匹配分割点,采用Bresenham直线生成法分割遮挡车辆。实验结果表明,该算法有效解决遮挡车辆分割不准确问题,与其他算法相比,具有较好的场景适应性。

Segmentation of occluded vehicles based on concavity analysis of contour feature points

In the shooting scenes of intelligent transportation system, the phenomenon of vehicles occlusion is caused by the distance between vehicles and the angle of camera shooting, which increases the difficulty of target vehicles detection and tracking. A concave region detection and segmentation algorithm based on contour feature points and contour concavity and convexity is proposed. Firstly, we use background difference method to extract the vehicles region. Meanwhile, according to the length-width ratio and the duty ratio of the outer rectangle of the vehicles area, the multi vehicles occlusion in the area is judged, and through convex analysis algorithm to extract concave area. Then, the edge feature points of concave area are determined by Freeman chain codes, and the feature points are concave analyzed. Finally, the Bresenham segmentation method is used to segment the occluded vehicles by matching the segmentation points. The experimental results show that the algorithm can effectively solve the problem of occluded vehicles segmentation, and has better scene adaptability compared with other algorithms.