Research on character correction method based on machine learning

Traditional industrial production is dominated by labor force, and the quality inspection personnel in production line have a large workload and long working time, which is easy to reduce the detection efficiency and increase the error rate due to visual fatigue. Therefore, it is urgent to use machine vision technology to identify characters instead of manual. However, when the image is obtained from the camera, the character may be inverted, which will lead to the phenomenon that "W" is recognized as "m" in the process of character recognition, which seriously affects the recognition accuracy. In order to solve this problem, this paper proposes a character correction method based on deep learning. Firstly, Otsu method and global threshold algorithm are used to segment the image. Then, the method of introducing directional gradient histogram (HOG) and support vector machine (SVM) is proposed to classify and correct the character images extracted before. The experimental results show that the algorithm proposed in this paper has sufficient practicability and robustness, realizes the automatic correction of inverted characters, and can effectively improve the accuracy of character recognition.