Real-time vehicle queue estimation of large-scale traffic scene

Abstract

This paper describes a method for the real-time vehicle queue estimation of large-scale traffic scene. The method is based on improving the vehicle presence detection and movement analysis method to combine the temporal and spatial information of neighbor frame images, and adding a mechanism for eliminating false detection. The results of experiments are presented by estimating queue length from the highly cluttered video of a high-speed toll station entrance. The accuracy of the system is evaluated by comparing the queue estimation results with the manual annotations. Experimental results indicate that the method is simple and works stably, which makes it potentially be used in broad application areas.