Research on optical character recognition based on MLP

Character recognition in images has become an important research direction in machine vision. With the development of deep learning technology, the accuracy of optical character recognition is constantly improving. In this paper, the principle and processing method of character recognition technology are deeply studied, and the key technologies in the process of character recognition such as image preprocessing, character segmentation and character recognition are discussed. Based on Otsu threshold method, a weighted threshold method is proposed, which is very useful for character segmentation of the overall darkening image. The accuracy of character segmentation is further improved by combining connected region detection and morphology. A classifier model based on multi-layer perceptron (MLP) is adopted to realize the effective recognition of RMB prefix numbers. The experimental results show that the character recognition method in this paper has high accuracy and robustness, and can adapt to the needs of real-time systems.