Abstract: This paper proposed a vehicle retrograde detection method on urban single lane based on video .First , background subtraction was used for moving vehicle detection .Second ,Harris corner features were combined to comple-ment the color features of Mean-shift to improve tracking accuracy ,from which motion trajectory of a vehicle can be com-puted .Finally ,the computed motion trajectory was analyzed for vehicle retrograde determination .Comparative experi-ments were conducted to evaluate retrograde detection accuracy and computational complexity .Experiment results show that the proposed detection method works well for vehicle retrograde detection .The average recognition rate is about 87 . 55% ,which is similar to that of the comparative method .However ,the proposed method has 19 .56 ms less computation time than the compared method ,therefore ,with fast processing speed .