基于改进分水岭算法的粘连车辆图像分割

Separation of Merged Vehicles Based on Improved Watershed Algorithm

为解决基于光流法的交通监控视频车流量检测过程中出现的车辆前景粘连情况,本文提出一种基于欧氏距离的分水岭粘连车辆分割算法.算法通过计算车辆前景中像素点距离背景的最小欧式距离,对阈值筛选后的区域进行质心的选取;以质心为起点构建分水岭算法的集水盆地,生成分水岭轮廓;将分水岭轮廓与原粘连图像结合完成粘连车辆的分割.本文进行各种类型车辆粘连图像分割实验,结果表明,本文提出的算法对于车辆粘连情况分割成功率高,获得了良好的车辆识别效果.

In order to solve the problem of vehicle foreground adhesion in traffic monitoring videos for traffic flow detection based on optical flow method,this paper proposes a watershed adhesion vehicle segmentation algorithm based on Euclidean distance.The algorithm calculates the centroid of the thresholded area by calculating the minimum Euclidean distance between the pixel and the background in the foreground of the vehicle.The centroid is used as the starting point to construct the watershed basin of the watershed algorithm to generate the watershed contour;the watershed contour is combined with the original adhesion image to complete the segmentation of adhered vehicles.In this paper,various types of vehicle adhesion image segmentation experiments were carried out.The results show that the proposed algorithm has a higher success rate for vehicle adhesion segmentation and achieves good vehicle recognition.