A text content-independent identification method for wrinkled Chinese handwriting

Abstract

The invention provides a text-content-irrelevant wrinkle Chinese handwriting identification method, and belongs to the field of handwriting identification. The method comprises the steps that (1) a handwriting text is scanned and then input to a computer, and the obtained handwriting images are preprocessed; (2) an image database is built according to the handwriting images; (3) scattering conversion is carried out on each image in the image database, and a scattering coefficient of each image is obtained; fitting is carried out on the scattering coefficients, and statistic models are built; (5) a statistic model corresponding to the handwriting image to be identified is searched for; (6) the KL distance between the handwriting image to be identified and a candidate image is calculated; (7) the identification accuracy is calculated; (8) comparing analysis and statistics are carried out on the experimental result. Compared with the prior art, the method can obtain a higher identification rate within the same time, and can well resist wrinkles and still obtain a higher correct identification rate for handwriting identification under a wrinkled condition.

Description

technical field [0001] The invention relates to the field of handwriting identification, in particular to a text-content-independent identification method for wrinkled Chinese handwriting. Background technique [0002] Handwriting recognition is a biological behavior recognition technology that authenticates writers based on handwriting. It has been widely used in security, finance and other fields, and has gradually become a research hotspot in the fields of computer vision and pattern recognition. Generally, according to whether the handwriting is related to the content of the text, it can be divided into two categories: text-related off-line handwriting recognition and text-independent off-line handwriting recognition. Off-line handwriting recognition that has nothing to do with text overcomes the shortcomings related to the writing content of the former, and has the characteristics of less constraints and easy data acquisition in practical applications, and has been wide...

AI Technical Summary

Problems solved by technology

[0004] The data processed by the above methods are all text images obtained under normal conditions, but in practical applications, the obtained handwriting text images may be text images obtained after scanning wrinkled paper, and the handwriting produced by paper wrinkling will appear to a certain extent. Translation and local elastic deformation cause certain visual differences in the same text image, and the handwriting produces changes such as local translation and local deformation. When the local feature method is used for handwriting identification, it is necessary to binarize the text image, and the wrinkled handwriting image Due to the large impact of deformation on the binarized image, and this impact is random and uncertain, the feature quality of the stroke feature extracted by the local feature method is reduced, so the local feature method is difficult to deal with wrinkled handwriting identification, so To identify the identity of wrinkled Chinese handwriting, its description features should have local translation invariance and local deformation stability

The features described by the above method do not have local translation invariance and local deformation stability, and it is difficult to deal with handwriting identification in the case of wrinkles. Therefore, the research on the identification method of wrinkled Chinese handwriting is very meaningful.

Images