In order to obtain the character information of the license plate from the license plate image, it is necessary to study the character recognition system. The license plate recognition process is divided into four parts: image preprocessing, license plate location, license plate correction and segmentation, and license plate character recognition. For the first three links, image median filtering, Sobel operator edge detection, mathematical morphology processing, Otsu binarization and Hough transform correction were completed, and a single standard character image was obtained. In the license plate character recognition link, firstly, according to the position of Chinese characters, the template matching algorithm is used to identify Chinese characters. For digital and alphabetic characters, the character images are segmented, and the number of connected regions of each part of the image is obtained to obtain character shape features. And training BP neural network to recognize characters. Finally, a system that can complete the recognition of the license plate characters in the image is obtained.