Character line detection method and device

Abstract

The embodiment of the invention provides a character line detection method and device. In the embodiment of the invention, a preset YOLO model is used for simultaneously detecting a position of a character line in a to-be-detected image, an inclination angle of the character line, a forward / backward direction of characters included in the character line and a language type of the characters included in the character line. According to the embodiment of the invention, no self-adaptive binarization method needs to be used for extracting the characters in the image, thus cases of reducing detection accuracy of the character line due to illumination or shadows can be avoided, no bag-type feature classifier needs to be used for determining the character direction and the character language typein the character line, and cases of reducing the detection accuracy of the character line by lower generalization ability thereof is avoided. Generalization ability of the YOLO model of the embodiment of the invention is higher than that of bag-type feature classifiers, and thus the embodiment of the invention can improve the detection accuracy of the character line as compared with the prior art.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a method and device for detecting character lines. Background technique [0002] At present, there is a need to detect text in images in many occasions, such as collecting images containing ID cards, driving licenses, driver's licenses or business cards, and then detecting text information such as names, numbers or positions in the images. Wherein, each text message contains multiple texts arranged in a row, for example, the name "Zhang San" contains two Chinese characters, the ID number contains eighteen numbers, and the position contains more than two Chinese characters, etc. . [0003] Among them, each text information is a text line composed of multiple texts. When it is necessary to recognize text information in an image, it is usually necessary to first determine the text line in the image, and then use OCR (Optical Character Recognition, optical text recognition) techno...

AI Technical Summary

Problems solved by technology

[0006] First of all, when using the adaptive binarization method to extract the text in the image, it is often affected by light or shadows, resulting in missing characters, or causing the extracted text to contain non-text noise, which may lead to the detection of text. The line does not exactly match the actual text line in the image, which reduces the detection accuracy of the text line

[0007] Secondly, the generalization ability of the bag-type feature classifier is limited by the size of the dictionary and its corresponding feature vectors. The ability to transform is low, which in turn reduces the detection accuracy of text lines

Images