Abstract: According to application requirement of video-based traffic statistics in video surveillance,the paper proposed improved background detection and tracking count method．Each pixel of a frame is to be updated in the traditional Gaussian background modeling and the number of Gaussian distribution is fixed which makes resources consumption increase．This article proposed that the update area is need to be fund firstly and then update the area．The region is updated by using dynamic adjusting Gaussian distribution method．At the same time,considering the characteristics of the mean and variance,and seting respectively their update rate．Connected domain analysis is applied to create the human node and get the node's centroid．Finally according to the nodes’ centroid which has been created in the list,first search forward is used．All target pixels are searched in the next frame which are used to determine the new position of the object in the video．Experiments show that the algorithm is simple and feasible and implements accurate tracking of multiple targets．The system achieves statistical data with high accuracy．

Key words: Video surveillance,Traffic statistics,Background model,Mixture Gaussian model,Tracking of multiple targets,Connected domain analysis