Research on Recognition and Classification of Moving Objects in Mixed Traffic Based on Video Detection

At present, mixed traffic is the major property of the traffic in some developing countries, mixed traffic parameters can be obtained by video detection which is utilized more and more for the traffic information collection. In this paper, recognition and classification of moving objects is developed. In order to obtain moving objects from the video sequence efficiently, this paper presents a background initialization algorithm based on clustering classifier, all stable non-overlapping intervals in the training sequence are located for each pixel as possible backgrounds by slip window at first; then the background interval is obtained from the classified data set by unsupervised clustering. According to mixed traffic, a simple effective feature representation algorithm is proposed, the distance between the point of object’s silhouette and the geometric center is defined as “centro-distance”, all of centro-distances along the silhouette make up of a vector, and moving objects could be recognized with the vector of centro-distance. Minimum-distance method is used to classify moving objects in mixed traffic into three categories: vehicles, bikes and persons. Temporal consistency is combined with to further improve the accuracy of the classification. Experimental results show that the proposed method can be applied well in real-world.