Font Size: a A A

Algorithm Research On Detection Of Drugs Blister Packaging Defects Based On Machine Vision

Posted on:2015-07-04Degree:MasterType:Thesis
Country:ChinaCandidate:W WuFull Text:PDF
GTID:2298330467490114Subject:Electrical engineering
Abstract/Summary:PDF Full Text Request
At present, aluminum-plastic blister packaging way has become themainstream medicine packing way because it is convenient to carry,taking medicine health, and long storage period, etc. Due to the influenceof various factors in the process of drug manufacture may give rise toquality problem, so the quality test of drug packaging is an important partin the process of the pharmaceutical industry production.The detection technology of drug packaging defect based onmachine vision has overcome many shortcomings, for example, the lowartificial detection efficiency and high labor intensity, etc. Therefore, thistechnology becomes an important developing direction of drug qualitytest. This paper is main to analysis and study on the detection ofmedicines in aluminum-plastic blister package based on machine vision.On the basis of comprehensive and in-depth study of related theory andalgorithm about drug defect detection technology, choose the differenttesting methods contraposing the two different types of tablets andcapsules drugs, and then compare the treatment effect of the algorithmthrough the simulation test. The main research work is as follows:(1) In order to deal with real-time acquisition of the medical image,this paper has adopted image medicine plate region extraction algorithmbased on the least squares fit.(2) in order to distinguish the type of drugs automatically, this paperset up the color histogram of every drug image in RGB color space. Usethe similarity of color histogram to judge the drug types. For differentsegmentation algorithms are analyzed, put forward a new thresholdsegmentation output function on the basis of two-dimensional Otsumethod, and improve the algorithm. In addition, the paper improvesCanny algorithm through using median filter, optimization of gradientamplitude and the self-adaption of increase of strengthening doubleadaptive threshold selection. And complete the pill image segmentationby using the improved Otsu threshold segmentation algorithm andmorphological operation. Capsule image segmentation was realized byusing color information and projection. And all of them have better effect.(3)Identify the presence of defects by extracting the color characteristics of each capsule blister area. And the way to identifyTablets’defects is to use the connected domain tags and the boundarytracking to get the geometrical characteristics of bills to build as thefeature vector. And then according to the vector distance between thefeature vector and complete tablet to identify whether there is a defect.
Keywords/Search Tags:Machine vision, Drug defect detection, Image segmentation, Defect recognition
PDF Full Text Request
Related items