Video/image detection is modern and effective for detection of traffic flow, since it can be installed simply and integrates surveillance and detection. Mixed traffic flow exists in so many developing countries that it emphasizes the importance of relevant information collection methods. Through the analysis of the characteristics of mixed traffic flow and available methods on video/image detection, a full-scene image detection framework based on background subtraction is proposed. In this paper the research on background model and shadow detection are illustrated. The extended running average algorithm for obtaining and updating representation of the background scene is proposed and compared with background model based on kernel density estimation and three-parameter background model in efficiency. A shadow model is proposed, considering the significant impact of the moving objects shadow to detection efficiency. After analyzing samples which were captured from real traffic scene, in several color spaces such as RGB and HSV, the utility of the proposed method is demonstrated through experiments on several scenes. (a) For the covering entry of this conference, please see ITRD abstract no. 0612AR242E.
Abstract