According to the actual application requirements, the requirements of recognition of license plate recognition system with high accuracy, proposes an improved neural network license plate recognition algorithm based on neural network recognition algorithm on the standard was improved in the standard neural network algorithm in the method of increasing the inertia impulse batch processing is improved and tested by training a large number of samples, the experimental results show that the neural network recognition algorithm improved the recognition rate and processing speed is greatly improved compared with the standard neural network character recognition has not improved algorithm has been applied in residential parking lot, achieves the application requirements, and proves that compared the improved neural network license plate recognition algorithm and standard the neural network algorithm in the realtime recognition accuracy has been greatly improved.