Research on Character Detection and Recognition Algorithm of Digital Instrument

At present, the traditional image processing and machine learning methods are mostly used for the recognition of character of digital instruments, the application scenarios of which is fixed and the restrictive conditions are many. In the complex and dynamic real environment, the recognition accuracy of character is rapidly reduced, which is difficult to meet the actual application requirements. Regarding the issue above, combining traditional image processing with deep learning, a method for detection and recognition of character based on deep convolutional neural networks is proposed. The method mainly includes three steps: screen area detection, character extraction, and recognition. First, the improved SSD deep convolutional neural network is applied to locate and segment the screen area of the instrument in the input image; Then, the visual saliency map of the screen area is calculated, and the characters are obtained by binarization the saliency map based on adaptive threshold; Finally, an end-to-end fully convolutional neural network is designed to recognize characters. The experimental results show that the proposed method can effectively improve the accuracy and robustness of character recognition. On the digital instrument dataset collected without restrictions, the overall recognition accuracy reaches 86.3 % and the detection speed is 10fps, which can meet the needs of practical application.