Video detection is widely used in detecting traffic volume in public traffic these days.There are mainly three methods of vehicle tracking in traffic volume detecting: background subtraction methodology,inter-frame differencing and optical flow.And background subtraction methodology is the most widely used one.And the most popular algorithms for background extraction in the current work are the Mixture of Gaussians Model and the Codebook model.Because of the busy traffic of the given video,affects from the sunshine,unstable lighting,shadows of the vehicles and congestion,neither the background subtraction methodology nor the inter-frame differencing methodology can solve the above problems well.So,this paper tries to combine these two technics to track the vehicles correctly.This algorithm is implement in TI high performance processor TMS320DM6446 hardware system.The experiment shows that this method improves the accuracy very well. |