OCR identification method based on template matching

Abstract

The invention relates to the field of image identification and processing, in particular to an OCR identification method based on template matching. Images and words to be identified are divided into sub-images only containing single characters, wherein figures, letters, punctuation marks and character sub-images therein are respectively marked out; according to corresponding font making characteristic images are selected according to image characters to be identified; normalization processing is conducted on the sub-images to be identified and corresponding types of characteristic images. On the basis, characteristic comparison is performed, contrastive calculation is conducted on the figure, letter and punctuation mark sub-images to be identified and alphabetic character sub-images on corresponding types of characteristic templates respectively, and an XOR algorithm of the same pixel position is adopted during contrastive calculation, and error frequency statistics is performed; marks corresponding to characteristic images having the lowest error frequency serve as identification results to be output. The method achieves image identification, the calculation process is simple, human and material resources are saved, and the identification efficiency is higher.

Description

technical field [0001] The invention relates to the field of image recognition, in particular to an OCR recognition method based on template matching. Background technique [0002] With the development of society and the advancement of science and technology, the amount of knowledge created by humans is increasing exponentially. Before the emergence of electronic books, most of the knowledge was passed down in the form of books. China has produced a large number of excellent Books, these books have been damaged to varying degrees in the long river of history, so it is imminent to digitally store these books; Due to the large number of books and the fact that the early printed books did not have the author's electronic manuscripts, it is necessary to digitize paper books. [0003] Optical character recognition software is a powerful tool for converting paper books to electronic documents. It mainly uses a large number of character samples to generate corresponding model file...

AI Technical Summary

Problems solved by technology

However, when the text in the image is glued together and the image contains Chinese characters with a left-right structure, it is difficult to achieve a good segmentation effect with a simple projection method; it is for this reason that segmentation has always been a difficulty in OCR recognition , the quality of segmentation will directly affect the text recognition effect

[0005] In addition, the main function of optical character recognition software is to recognize characters in photographed and scanned pictures. For scanned copies of some special fonts, official seals, photographs, such as early printed books, certificates made by government units, etc., due to historical reasons and confidentiality and security needs , the fonts are often specially made, the existing optical character recognition software mainly focuses on machine learning methods, the model has a large amount of computation, and because the training font samples do not cover the special fonts, the recognition accuracy of special fonts is not high, serious Influencing the digitization of paper documents

[0006] Most of the existing technologies use neural network machine learning algorithms to recognize characters, which requires the production of a large number of samples and takes a lot of time for training, and the generated model files are very large, and the recognition rate is not the same for characters of different fonts. The recognition rate of some special font characters is relatively low, and it is difficult to meet the character recognition in some special scenarios

Images