Real-time detection algorithm for moving vehicles in dynamic traffic environment

Abstract

As the video detection technology has become a hot issue in the Intelligent Transportation System (ITS), detecting the moving vehicles accurately in real-time is one of the challenging problems. This paper describes a real-time method for segmenting moving vehicles in dynamic scenes. An adaptive background update algorithm based on three-frame difference is proposed. By this method, the background is updated according to the dynamic change of surrounding light and external events. Meanwhile, for acquiring an accurate and whole segmentation result, an adaptive threshold algorithm which could adapt to real-time changes of light in the traffic environment is proposed. Experimental results of several traffic scenes are provided, which demonstrate the real-time and dynamic update of the background, and the effective segmentation by the presented method. The proposed method can be utilized in the complicated transportation environment, which lays the foundation of the practical application of the video detection technology.