Research on License Plate Character Recognition Technology Based on Image Processing and Deep Learning

Abstract

Character recognition methods are applied in many fields, greatly improving work efficiency in daily life[1], such as license plate retrieval, invoice printing recognition, lottery betting codes, tax reports, etc. Digital recognition has been widely used in the field of computer vision and image recognition, and deep learning algorithms are currently popular image recognition algorithms. Deep learning has been widely studied and applied in target recognition and speech content recognition. With the rapid increase in production requirements and computer data processing speed, the application of character recognition in actual production and life is becoming more and more common[2]. It is also extremely important for automatic retrieval and real-time, fast and accurate character input. However, traditional pattern recognition and feature extraction algorithms cannot well meet the requirements of real-time and correctness in production. At the same time, due to the vigorous development of deep learning, character recognition technology based on deep learning has advantages that traditional recognition algorithms cannot match. This paper proposes a barcode recognition algorithm based on a deep neural network combined with a global optimization method. It uses a convolutional recurrent network to extract the characteristics of each character in the barcode and classify it. Compared with the traditional method, it has stronger adaptability and generalization. Chemical energy.