Real-time detection of traffic flow combining virtual detection-line and contour feature

Abstract

This paper proposes a novel method for real-time detection of traffic flow, which can count the number of vehicles for various occlusions accurately. In our method, by utilizing a virtual detection-line, the procedure of tracking which is always employed by traditional approaches is neglected. In this paper, the spatio-temporal contour model is firstly presented to reconstruct contour which contains spatio-temporal information when vehicles get across the detection-line. Then, spatio-temporal contour description model and reasoning model for counting are utilized to describe the features of spatio-temporal contour and calculate the count of vehicles, respectively. The proposed method has been tested on several video sequences with different degrees of occlusion, and achieved more than 97% of detection rate. The proposed method is validated to be simplicity and real-time, and can be used in traffic surveillance system.