Aiming at the problem that wrong extraction of vehicles shadow will cause low reliability and will not satisfy the real-time requirement in intelligent transportation system, an improved model based on HSV color space and difference of background is proposed. According to the analysis of the model, value is pushed out. The different values of shadow, vehicles and background can be used to extract shadows from target vehicles, and background subtraction is also used. Experiments on different scenes suggest that high effective detection rate of the proposed method can be reached, and this method is also effective to the reliability and real-time requirement in ITS.