Convolutional neural network for recogniton of licence plate symbols

This paper discusses an approach to the license plate character recognition based on convolutional neural networks (CNN, convolutional neural network). Using a convolutional neural network for image recognition is caused by two main reasons: 1) reduction of the complexity of the neural network training and output calculation in comparison with a classical multilayer perceptron, which is important in the field of image processing and analysis, and 2) increased robustness of CNN to different distortion applied to symbols recognition in contrast to classical neural networks and other methods of image classification. The paper describes the design process of the convolution neural network for recognition of license plate characters. At the end conducted comparing the quality of the developed CNN recognition method and traditional template matching method.