Traffic States Detection based on Block Matching

A traffic states detection method is proposed to represent the road traffic conditions accurately and reduce labor intensive monitoring work in this paper. The edge of image, which is named as edge characteristic, is applied to describe the number of vehicles running on the road, and the average density of vehicle edge pixels is used to substitute the road occupation ratio. Meanwhile block matching method is used for estimating the speed of traffic flow. According the extracted feature parameters, a classifier is designed to classify the traffic states in three categories (congested, slow, and smooth). The experimental results showed the proposed algorithm is robust. Keywords-Traffic state detection; Traffic flow speed; Block matching; Vehicle edge density; Background subtract