“0†and “O†Recognition Based on Deep Learning

Abstract

The traditional algorithm has achieved good recognition for the recognition of most characters, but the recognition rate of the number “0†and the letter “O†is only 80-90%, which is difficult to meet the actual needs of the industry. For the recognition of these two characters, a recognition method based on deep learning is proposed. First, image preprocessing is performed on the characters, and then the sample data is manually labeled, and 40,000 training samples and 10,000 test sample images are obtained by the data samples enhancement. The results show that the CNN network can achieve more than 99% recognition rate, the training sample time is about 5 minutes, and 1000 images can be recognized in 1 second. Both the recognition speed and the recognition effect can meet the actual needs of the industry.