A method of extraction prospect, which uses the background image LBP (local binary pattern) texture and current frame image LBP texture similarity analysis, was put forward. This method overcomes the sensitivity to illumination methods in vehicle detection, such as frame difference method and background difference method. The Camshift tracking algorithm combines the H,S and V components, the improved LBP texture of the joint histogram with the centroid tracking by pyramid L-K optical flow. This method can effectively solve two problems: one that the similar background color may lead to the tracking of the target area bigger, and the other that the search window position accuracy is low when dealing with large displacement between frames of video. The experimental results prove that the method has good detection and tracking effect.