English character segmentation method and device based on object detection network

Abstract

The embodiment of the invention discloses an English character segmentation method and device based on an object detection network. Through the English character segmentation method and device, segmentation accuracy of English characters in a picture can be improved. The method comprises the steps that S1, a to-be-segmented picture is acquired, and an English text line region is extracted from theto-be-segmented picture; and S2, the English text line region is input into the object detection network obtained through pre-training, and a segmentation result of the English characters in the to-be-segmented picture is obtained, wherein training samples used during training of the object detection network include indentation adhesion samples, extension adhesion samples, viewing angle adhesionsamples and special effect adhesion samples.

Description

technical field [0001] Embodiments of the present invention relate to the field of character recognition, and in particular to a method and device for segmenting English characters based on an object detection network. Background technique [0002] In recent years, fields related to text recognition have benefited from convolutional neural networks to receive more attention. Generally speaking, the text recognition process is expanded from high-level text lines to low-level characters. The positioning of the text line and the recognition process of a single character have achieved remarkable results. Character segmentation is indispensable for obtaining a single character area from the text line area. However, the work related to text segmentation is still developing slowly. It has become a major factor that limits the effect of text recognition at present. Improving the accuracy of character segmentation can solve the shortcomings in the current text recognition process, ...

AI Technical Summary

Problems solved by technology

The characters that stick to each other bring great difficulties to the segmentation work, and the traditional method is not satisfactory in the effect of character segmentation

In addition to the sticking phenomenon, the large range of character size changes in the text line, the inclination and distortion of the scene picture, the complex font and artistic format of the text also increase the difficulty of character segmentation, and it is convenient to find a new character segmentation method. It is particularly urgent and full of challenges

Images