Annular character rapid recognition method and device

Abstract

The invention provides an annular character rapid identification method and a device. The method comprises the following steps: carrying out image gain processing on an annular character image; intercepting a circular area where an annular character band is located in the annular character image; rearranging the annular character band in the intercepted circular area into a rectangular character band; positioning a character area in the rectangular character belt; segmenting and correcting the positioned character area to obtain each character image; and performing character recognition of a neural network on each segmented character image to obtain a character recognition result. According to the method, an image processing technology is used to carry out gain processing on an image so asto improve imaging quality and reduce requirements for shooting equipment and a shooting environment; a rectangular rearrangement strategy of an annular belt is innovatively adopted, so that the workload of subsequent steps is reduced, and the problem becomes simple; the character areas are positioned, segmented and corrected to obtain each character image, so that the character division is clear, the recognition is more accurate, and the recognition rate is improved.

Description

technical field [0001] The invention relates to the fields of machine learning and machine vision, in particular to a method and device for fast recognition of circular characters. Background technique [0002] Character recognition technology is a technology that locates the position of the characters, processes the image of the area and judges it through the method of character recognition. The main types of text recognition technology are OCR (Optical Character Recognition) and ICR (Intelligent Character Recognition). [0003] In text recognition, we must first locate the area where the text is located. Traditional methods include gradient-based algorithms, methods based on character texture changes, and methods based on grayscale histograms. New methods include area-based convolutional neural networks, One- shot learning and Zero shot learning. [0004] Region-based convolutional neural networks belong to the field of deep learning. Representative algorithms include RC...

AI Technical Summary

Problems solved by technology

[0007] Although the character recognition technology has been developed for nearly a hundred years, the current character recognition adaptation scene is still relatively single, almost only able to recognize characters arranged in rows in the traditional sense, and basically has no adaptability to characters arranged in a circle

Images