This paper proposes a new model for vehicle detection. The main principles of this paper are to describe the distribution of vehicle appearance by using compact vehicle appearance prototypes to match the individual appearance prototype, and applying a simple distance standard, to collect and classify all the obtained regional information into feature library, to obtain the probability of similar appearance depending on the number of matched vehicle appearances and the density of coverage in target detection, and eventually to distinguish and recognize different parts of the vehicle appearances by setting the threshold. Experimental results show that the use of this detection model can facilitate and effective monitoring vehicle in varieties of weather conditions.