A character recognition method and system

Abstract

The present invention provides a text recognition method and system, the method comprising: acquiring a text image of a natural scene to be recognized; inputting the text image of a natural scene into a text recognition model to obtain text content; wherein, the text recognition model is formed by a visual recognition network , a visual information optimization network and a semantic information optimization network; the visual information optimization network and the semantic information optimization network are used to generate corresponding text recognition supplementary information according to the character characteristics output by the text recognition network for the text recognition network based on text Identify supplementary information and optimize text recognition results. The present invention uses the visual information optimization network and the semantic information optimization network to obtain global information at the two levels of visual features and language results, and obtain visual and semantic clues as supplementary information for the iterative execution of the recognition process by the visual recognition network, thereby Gradually obtain more robust recognition results and improve the accuracy of text recognition.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a character recognition method and system. Background technique [0002] Text recognition is one of the popular research topics in computer vision, and it is also an important research direction in the application of vision-based pattern recognition. Text recognition technology needs to transform the located text area into a symbol that can be recognized and operated by the computer. Since the text in natural scenes often presents a variety of different forms, the text feature representation obtained through recognition requires strong expressive ability and discrimination. force. [0003] Deep learning is a popular research direction in the field of machine learning in recent years. It has achieved great success in the fields of computer vision and natural language processing, especially deep convolutional neural networks. The three structural features of downsampling ...

AI Technical Summary

Problems solved by technology

[0004] However, the existing text recognition technology based on deep learning mainly uses the recursive neural network to decode text. When recognizing characters, it can only use the text information that has been recognized, and cannot obtain global perception, resulting in the accuracy of text recognition. lower

Images