基于轮廓特征拐点的遮挡车辆分离方法研究

Research on occlusion vehicles image segmentation based on feature points on contour

遮挡是智能交通系统中车辆检测、跟踪和识别时较难解决的问题之一。本文主要针对智能交通系统中车辆之间的遮挡,以车辆常用的矩形模板为先验知识,首先对存在车辆遮挡的连通区域提取边缘轮廓,然后获取轮廓上的4类特征拐点,最后采用本文提出的基于轮廓特征拐点的车辆分离算法,将相互遮挡的车辆分离开来,便于后续对车辆的跟踪、匹配等处理。实验结果表明,本文提出的算法具有有效性、可行性和良好的应用前景。

Occlusion in the monitoring video was a problem often encountered in the moving vehicles detection,tracking and identification.In this paper,the occlusion among vehicles was mainly considered and a rectangle region was assumed as the shape template of a vehicle.The first step was to extract the edge curve of the connected region,which had the phenomenon of vehicles occlusion,and then got the feature points on contour,at last using the vehicles segmentation algorithm which was based on feature points on contour separate vehicles,in order to facilitate the follow-up work as tracking and matching vehicles.Experimental results showed the effectiveness,feasibility and good application prospects of the algorithm in this paper.