License plate recognition system based on car license plate character target of a computer vision system. Due to obtain the vehicle images affected by outside factors such as illumination and noise interference, usually with complexity and uncertainty, lead to complex scenarios license plate character recognition more difficult. In order to improve the recognition rate, this paper use BP network to the fuzzy processing is used to identify the training of the positioning plate, using Matlab to experiment. With convolution neural network for license plate recognition experiments, compared with BP algorithm, the simulation and experiment for the proposed algorithm, compare two methods found the convolutional neural network algorithm for license plate character recognition has high recognition rate, is widely used gradually.