This paper is the study of vehicle license plate recognition. We use the scan window, 2-D Haar Discrete Wavelet Transform and Artificial Neural Networks. This paper consists of three main parts. First part is to locate and extract license plate. Second part is the Artificial Neural Network training of the license plate. Third part is a real time scans recognition. We use two times Haar Discrete Wavelet Transform image, and extract low frequency part LL of vehicles license plate image, makes image coefficient became original of one-sixteen, thus reducing image coefficients, so can to fast recognition.This method is to use scan the license plate recognition, but we do not license plate recognition for each character. Experiment result can be a high recognition rate.