Penicillin bottle drug label identification method based on machine vision

Abstract

The invention relates to a penicillin bottle drug label identification method based on machine vision. By using a SFDA (State Food and Drug Administration) approval number as a drug label identification basis, drug information is extracted and identified from an imaged picture of a cylindrical drug label so as to complete automatic identification of a drug variety; secondly, SFDA approval characters are rapidly positioned, positioning on a label image SFDA approval region is completed, and a to-be-identified SFDA approval number character string block is extracted; and finally, the SFDA approval characters are identified, candidate template characters are screened by adopting candidate character extraction based on statistical characteristics, and the identification of the SFDA approval number is respectively carried out on the basis of a BP neural network algorithm and XOR (Exclusive OR) enhancement algorithm. Only when identification results of both algorithms are the same with the SFDA approval number in a dispensing list of a patient, a case that a drug label is correctly identified is determined. The penicillin bottle drug label identification method is applied to a full-automatic dispensing machine, and can greatly reduce working intensity of medical personnel, improve infusion dispensing efficiency and reliability, implement the sterile operation, avoid potential hazardof drugs to medical personnel and the like.

Description

technical field [0001] The invention relates to the field of machine vision and image processing, in particular to a machine vision-based recognition method for vial drug labels. Background technique [0002] The workload of infusion and drug dispensing in medical and nursing work is huge, which can easily cause operational fatigue of medical staff, lead to errors in drug dispensing, and cause medical accidents. Compared with traditional dispensing methods and processes, automatic dispensing machines have obvious advantages in terms of dispensing accuracy, dispensing cleanliness, and health protection for medical staff. However, the automatic identification of drug types faces the following major problems: ①Most drug labels do not have identifiable barcodes, which brings inconvenience to label identification and drug management; ②Due to the lack of industry standards, the label specifications produced by different drug manufacturers Inconsistency, the position of informatio...

AI Technical Summary

Problems solved by technology

However, the automatic identification of drug types faces the following major problems: ①Most drug labels do not have identifiable barcodes, which brings inconvenience to label identification and drug management; ②Due to the lack of industry standards, the label specifications produced by different drug manufacturers Inconsistency, the location of information such as drug names and national drug standards varies greatly; ③The drug label is a cylindrical surface, and the edge image of the label is seriously deformed, making it difficult to identify the label character information by conventional methods

The invention patent "A Method for Recognition of Curved Written Characters" corrects the distortion angle of the character line by fitting the center of each line of characters with a curve fitting method, and realizes the recognition and understanding of the book layout information, but the curve fitting process is complicated, and it is difficult for the recognition unit Text with less distortion in this picture has better recognition effect

This method has high static recognition accuracy, but the dynamic recognition reliability needs to be further improved

Images