The Label Verification Method Research Based On Panoramic Stitching Technology

At present,intravenous infusion is an important part of infusion,but still more than 98% is by medical staff to use manual dosing method to achieve.In the process of dispensing,nurses repeatedly check drug name,dose,time and other information,the task is large and prone to flaws.In view of these problems,this paper studies a method of checking the bottle label based on panoramic splicing technology.This paper introduces the difficulties and problems in the calibration of the bottle label,the development status of the label recognition,the development of the panorama stitching technology,and the image processing method of the tag image recognition.The object of this study is the surface of the RGB intravenous drip bottle label image,the image will appear on both sides of the distortion of the phenomenon of compression,the identification of the bottle directly on the label will cause bias.Due to the camera to collect the label image angel is small,can only contain some information for the label,change the angle of rotation and the bottle,acquisition of image information change effectively.Therefore,this paper studies a method based on MATLAB image processing for surface tag recognition.The method mainly includes four parts,one is to extract the surface of the bottle label distortion correction,first through the image preprocessing,into a gray image,the use of Hough transform to detect vertical lines to correct the tilt angle of the bottle Tag image,and then the tilt correction of the image using the cylindrical back projection algorithm will be distorted curved surface of the bottle label to correct.Second,the collection of the vial label image is a multi-angle overlap area of the image,the distortion of the image after the stitching to get the label containing the full name of the image.Third,the name of the drug label to accurately cut the characters.Fourth,the drug name tag segmentation character matches the template character,where the specific unit area of the character image is selected to obtain the characteristic matrix of the pixel value per unit area,calculate the feature matrix of the cut character and the template character respectively,and calculate the feature Matrix similarity,the highest similarity is to identify the characters,so as to complete the calibration of the bottle label.In this paper,a number of different vial label images were used to test the distortion correction algorithm,and the effects of improved distortion correction algorithm and improved pre-distortion correction algorithm on image splicing and character recognition rate were compared.And the paper puts forward the requirement analysis of the bottle label algorithm in this paper.The experimental results show that the above method can accurately identify the bottle label information,with practical value.