Research on segmentation of overlapped vehicles based on feature points on contour

Abstract

Occlusion in the monitoring video is a problem often encountered in the moving vehicles detection, tracking and identification. In practice, the moving vehicles that are needed to be tracked are often overlapped in the image. As a result, the mistakes in the targets segment and traffic parameters calculation are the problems much more difficult to solve. Generally, the moving target occlusion in video can be divided into two categories, one is the occlusion between the targets, the other one is the occlusion between the target and background. Considering the occlusion between the targets, a vehicles segmentation algorithm, which is based on feature points on contour, is presented in this paper. First of all, it is needed to detect and judge that if the occlusion between the targets was exist in the moving region according to the calculation of the duty cycle and the ratio of length and width. Then, the connective vehicles region that involving vehicle occlusion will be segmented into independent targets according segmentation algorithm mentioned above. MATLAB is used as the simulation platform to implement the functions of the vehicles segmentation. Experiment results show that the algorithm presented in this paper is effective and feasible, and also have good application prospects.