Vehicle Segmentation Approach Based on the Space-Time and Self-Similarity of Background

Abstract

Vehicle segmentation is the key of the traffic video image analysis, the movement state of vehicles can be understudied to get traffic flow parameters. Based on the traffic scene characterized by gradual change, the paper present the background model reconstruction with time-space and self-similarity of background. Video frame image is differenced from background, and binary to get the vehicle template. According to the template, movement vehicle will be segmented from it, the results of the experiment confirmed the effectiveness of algorithm.