车牌字符识别是车牌识别系统中的关键环节。采用图像处理和神经网络相结合的方法设计新的车牌字符识别算法，先对分割出的车牌字符进行归一化处理，然后进行SOBEL边缘检测和角点特征提取，最后输入BP神经网络进行训练、识别，其中BP神经网络模型属于改进型神经网络。通过一系列神经网络训练和仿真实验，车牌识别速度和正确率得到了明显的提高。

The license plate character recognition is a key link in the license plate recognition system. A new license plate character recognition algorithm is designed combining image processing and neural network method. First we split the normalized plate character, and then proceed to the SOBEL edge detection and the corner feature extraction. Finally enter the BP neural network training recognition. BP neural network model is a modified neural network. Through a series of neural network training and simulation experiment, the license plate recognition speed and accuracy have been improved significantly.