Distinguishing moving objects from traffic video by the dynamic background skeleton based model

This paper proposed an simple and efficient method to detect moving objects in traffic video based on the dynamic background skeleton model(be called DBSB below), after listing some existing methods for moving objects detection. The defects of other models of moving object detection is given firstly, and then a rough set based idea DBSB on how to perfect the defects is presented. With the DBSB, the background model is established on the historical data constructed mainly by the skeletons, with which the bad effects produced by illumination or season changes can be filtered out, of the components. The judgment of one belongs to the background or is a moving object is done with this model and the background model is updated with this time judgment result consequently. The experimental results show that this method can shield various disturbances, the adaptability and efficiency can be improved greatly with this method.