A text image recognition method, device and computing device

Abstract

Embodiments of the present invention provide a text image recognition method, device, readable storage medium, and computing device. By using the VAE model to improve the data and training process, and then combining CenterNet and CRNN, the text image can be quickly and accurately resolved. The problem of detection and recognition of scratches, the method includes: obtaining the first feature of the text image; according to the first feature of the text image, identifying the text box in the text image; according to the coordinate information of the text box, Intercepting the second feature corresponding to the text box from the first feature of the text image; obtaining the third feature of the scratch of the text image; subtracting the third feature from the second feature to obtain A fourth feature corresponding to the text box; according to the fourth feature, identify the content of the text box.

Description

technical field [0001] The invention relates to the field of text image processing, in particular to a text image recognition method, device, readable storage medium and computing equipment. Background technique [0002] Text detection has a wide range of applications and is a pre-step for many computer vision tasks, such as image search, text recognition, identity authentication, and visual navigation. The main purpose of text detection is to locate the position of text lines or characters in the image. The precise positioning of text is both very important and challenging, because compared with general object detection, text has multiple directions, irregular shapes, and extreme length and width. Ratio, font, color, background and other characteristics, therefore, the algorithms that are often more successful in general object detection cannot be directly transferred to text detection. However, with the re-emergence of deep learning in recent years, the research on text de...

AI Technical Summary

Problems solved by technology

[0004] At present, in intelligent applications such as photographic judgment questions for primary school mathematics, due to problems such as the respondent’s writing and thinking habits, it often occurs that the answer is crossed out and then answered again. The crossed out part is called a scratch problem, which is called The crossed-out answers are sometimes necessary and sometimes unnecessary, but accurately grasping the content of the scratched answers will not only help to improve the results of the judgment, but also help to master the answering habits and thinking methods of the respondents, so as to provide better answers. It lays the foundation for product iteration and better help for respondents, but the current conventional detection and recognition models, especially the recognition model, cannot solve the problem of scratches well

Images