Image and text embedded medicine label identification method based on deep learning

Abstract

The invention discloses an image and text embedded medicine label identification method based on deep learning, and the method comprises the steps: obtaining medicine label information data which comprises medicine picture information; extracting the input medicine picture information in combination with an optical character identification mode, and using a Tesseract OCR engine to complete identification and extraction of image texts in optical character identification; processing the extracted picture text information based on a similarity module; processing the extracted picture content based on a similarity module; and enabling the picture content text recognition result integration module to compare the K pictures which are obtained respectively and have the highest similarity ranking, and carrying out model training after comparison. According to the method, the judgment accuracy is improved by combining the picture content information and the image text recognition information, and the working efficiency of investigators is improved.

Description

technical field [0001] The invention relates to the field of medical image text recognition, in particular to a drug label recognition method based on image and text embedding of deep learning. Background technique [0002] Illegal, unapproved, counterfeit and potentially risky medicines can cause serious harm to medical patients, and the role of law in regulating medicines is limited and dependent on people's compliance with the law. At present, the verification of illegal and prohibited drugs requires manual input of drug names for inspection and relies heavily on the historical illegal inspection record database. In addition, it is impossible to give immediate judgments on drugs other than the illegal drug name record database. Investigations on drugs outside the record database are often More time is required, which can greatly reduce the efficiency of investigators. For regulatory investigators, how to make an immediate decision on whether to investigate drugs that are...

AI Technical Summary

Problems solved by technology

At present, the verification of illegal and prohibited drugs requires manual input of drug names for inspection and relies heavily on the historical illegal inspection record database. In addition, it is impossible to give immediate judgments on drugs other than the illegal drug name record database. Investigations on drugs outside the record database are often Requires more time, which can greatly reduce the efficiency of investigators

For regulatory investigators, how to make an immediate decision on whether to investigate drugs that are not in the illegal drug record library is a very complicated issue

[0003] At present, the image recognition technology based on deep learning is relatively mature, and the pictures of illegal drugs can be identified by comparing the similarity of the picture content. High, it is not suitable to only use the method of image content retrieval. It is more feasible to use text recognition to identify the text information such as the name of the drug in the image and compare the text similarity to make a decision whether to investigate. Currently, it is mainly used by the industry. The text recognition engine is Tesseract OCR, but there are still big problems in the case of text recognition in the face of different backgrounds

Images