Method for intelligently segmenting film characters of soft pharmaceutical bag

Abstract

The invention relates to a method for intelligently segmenting film characters of a soft pharmaceutical bag. The method comprises the following steps: collecting a film character image of a soft pharmaceutical bag; calculating the width of a single character and rough values of character heights; segmenting characters according with a feature vector by using a mode matching method; comparing whether the number of segmented characters is equal to that of practical characters or not, if not, correcting the character widths and the character heights in d), reducing the character widths if the number of the obtained characters is less than that of the practical characters in the product batch number, and otherwise, increasing the character widths; re-suing mode matching till each character is completely segmented so as to obtain precision values of the character widths and the character heights; matching characters of the whole image by using the precise character widths and the precise character heights; and segmenting all the characters according with requirements. By means of the invention, character segmentation can be carried out precisely.

Description

technical field [0001] The invention relates to a method for intelligently segmenting pharmaceutical soft bag film characters. Background technique [0002] Otsuka Pharmaceutical's soft bag film is a packaging bag for infusion medicine. Before filling and sealing, it is necessary to detect the characters printed on the film to prevent unprinted characters, wrong characters, reprinted characters, and unclear characters. Before recognizing characters, it is necessary to extract character templates to train artificial neural networks (hereinafter referred to as neural networks). [0003] At present, one method of extracting character templates is to use morphological algorithms such as opening and closing operations, but the algorithm has poor stability, poor applicability, and is easily affected by the quality of character images. The other is manual segmentation, that is, to manually delineate a single character on the image. This method is simple to implement, but the delin...

AI Technical Summary

Problems solved by technology

[0003] At present, one method of extracting character templates is to use morphological algorithms such as opening and closing operations, but the algorithm has poor stability, poor applicability, and is easily affected by the quality of character images.

The other is manual segmentation, that is, to manually delineate a single character on the image. This method is simple to implement, but the delineated character area varies from person to person. Although personnel have been trained before operation, it still leads to templates caused by inconsistent standards. Failure and other issues

Images