In this paper a system for the recognition of Chinese vehicle license plates is present. Images acquired under real conditions are often anamorphic. A method is designed to adjust the anamorphic vehicle license plates. Images are also influenced by weather and rays, which often make gray scale inhomogeneous. A preprocessing operation is used to solve this problem. Using pattern matching for character recognition, we can avoid isolating the characters and enhance the correctness of character extracting. To the question that few characters are easy to be confused, we construct BP neural networks to complete effective character recognition.