License Plate Image Recognition Based on Deep Learning under Complex Conditions

Abstract:
License plate is the "ID card" of a vehicle, which contains the information of the owner and the vehicle. As the identification mark of the vehicle, license plate recognition is a key part of the whole intelligent transportation system. This paper uses the method based on deep learning to detect and recognize the license plate. Due to the multi-scale detection ability of SSD target detection network, this paper selects SSD target detection algorithm in the process of license plate detection, and improves the non maximum suppression algorithm. In the license plate non segmentation character recognition, the convolution neural network is used to extract the license plate features. According to the size of the license plate text, the window of the last two pooling layers in the network is improved to better retain the license plate text information, enhance the license plate character recognition ability, and improve the robustness of license plate recognition.