本文针对车牌字符识别系统在工程应用中存在识别准确率不高、效率低的问题，从工程实践的角度描述了一种新的基于BP神经网络的识别系统在车牌字符识别中的应用。详细介绍了车牌字符识别中的样本集和测试集的组织、图像二值化、归一化、图像去噪、神经网络构建和训练。实践结果表明,本系统适用于自然场景中的车牌自动识别问题，并且具有较强的适应性。

In view of the shortcomings of the automobile license plate identification systems, such as the low identification accuracy and efficiency under practical conditions, a new identification system based on the BP network is designed. In terms of engineering application,the character identification of automobile license plates is addressed in detail, including building the training sets of samples, image binarization, normalization, removal of noise,and neural network construction. The experimental results show that the system has good performance even when the  images have low quality and the license plates are located in a complicated natural scene.