Character recognition is the key of recognition systems for license plate，and Chinese character recognition is the most difficult point. A kind of method that the gray character image is transformed into horizontal，vertical，charge-offs，Na energy feature image by Gabor filters is proposed. Non-linear transformation is designed to regulate the outputs of Gabor filters adaptively which is used to improve the performances for low quality images. Finally we rely on network method for feature extraction and advanced BP neural network as the classification to enhance the rate of the recognition of license plate Chinese character.