基于凹点匹配的黏连车辆图像分割方法

Image segmentation method for adhesion vehicles based on concave points matching

在利用智能交通监控系统对过往车辆进行检测时,车辆之间相互重叠、黏连的图像会给车辆计数和特征提取带来很大困难。本文针对车辆黏连图像提出一种简单快速的凹点匹配分割算法。该算法主要通过计算黏连车辆的凸闭包,得出黏连区域的凹区;再根据点到特定对角线的最小或最大距离,确定原图像的凹点集合;最后选择最优分割线对黏连车辆进行分割。仿真结果证实该算法精度高,并且稳定性好。

For the application of the intelligent traffic monitoring system to detect the passing vehicles,it is known that the overlapping and adhesion of the image will bring great difficulties to the vehicles count and feature extraction.To solve the problem of vehicles adhesion,a simple and fast algorithm for image segmentation was proposed in this paper.This algorithm began with the calculation of the convex closure of the vehicles adhesion and getting the concave area of adhesion area,followed by the determination of a set of concave points of the original image based on the minimum or maximum distance of point to diagonal distance,ended with the choice of the optimal split line to segment adhesion vehicles.Simulation results showed that the algorithm has high accuracy and good stability.