OCR image character recognition and character correction method and system

Abstract

The invention discloses an OCR image character recognition and character correction method. In the character recognition module, a Chinese character deformation degree Pr function is constructed and fitted by adopting a multi-stage neural network, and the network reflects the deformation degree Pr of a target Chinese character by taking image CNN data and four additional features of horizontal, vertical, left-falling and right-falling as variables and taking GAN recognition degrees of different degrees as training values of the deformation degree; in the character correction module, a second-stage similar character distinguishing network is additionally arranged and used for carrying out high-precision similar character distinguishing on the optimal recognition result determined by the training network for the first time, and due to the arrangement of a second-stage network, the complexity of the first-stage network can be reduced, and the overall generalization ability of the networkcan be improved. The method and the system are mainly used for identifying machine-made invoices, various forms and documents, are high in recognition precision, high in recognition speed and strong in adaptability, and have strong correction capability for partial information loss and recognition errors; compared with the recognition effect of the traditional OCR recognition technology, the recognition accuracy is greatly improved.

Description

technical field [0001] The invention relates to the technical field of Chinese character recognition, in particular to a method and system for OCR image character recognition and character correction. Background technique [0002] OCR (Optical Character Recognition, Optical Character Recognition) technology is to convert the text of various bills, newspapers, books, manuscripts and other printed materials into image information through optical input methods such as scanning, and then use text recognition technology to convert the image information into usable computer input technology. [0003] With the continuous development of image sensors, especially the exponential increase in the number of various mobile phones and professional (such as security) cameras, computer image data is increasing rapidly; but the image quality is relatively lower than that of traditional scanners or various professional cameras; traditional Chinese character OCR technology faces the problem t...

AI Technical Summary

Problems solved by technology

According to relevant market surveys in 2018, many traditional OCR manufacturers in the market did not perform well in the testing of various bills photographed by mobile phones. The field has achieved good results, but due to the large base of Chinese characters, the required training data set is thousands of times more than that of Western character sets (conservative estimates), so the performance of Chinese character OCR on the open AI platform is quite poor on poor images. Ideal, and end-to-end deep neural networks have natural misidentifications and are vulnerable to attacks

Images