Image Processing Methods Of Blister Packaging Drugs Visual Inspection System

Blister packaging drugs have been welcomed by the consumers with theadvantages of simple manufacture and convenient carrying. It is also named pressthrough packaging. However, traditional detection of PTP by labor is time-consuming,tiring and inaccurate. In this paper, a set of system about PTP based on visioninspection is designed. This paper studies the image processing methods of thissystem for the purpose of lower consumption, reduce cost and improve the quality.Firstly, according to the design requirements of blister drugs visual inspectionsystem, and combined with excellent image processing methods of both at home andabroad, this paper sets up a visual inspection system, which is composed of LED lightsource, industrial camera, industrial computer and PLC. Also this paper describes thesystem software processes in detail.Secondly, this paper studies the image process methods of Aluminum-plasticblister packaging and Aluminum-aluminum blister packaging on the market, For theAluminum-plastic image, the image preprocessing has been used at first in this paper,and then using Otsu algorithm for the global threshold segmentation. After thatprocessing, we use projection to get the location of single tablet, and then using thefeature extraction, template matching technique to determine whether the tablet isgood or not. For the Aluminum-aluminum image, this paper proposes a new localthreshold segmentation algorithm to get binary image, and then using projection to getthe location of single tablet. After then, we use seed filling algorithm to fill the singletablet. After that processing, we extract the area and perimeter characteristics of thesingle tablet, and compare with the standard template to judge whether the tablet isgood or not. The new local threshold segmentation algorithm of this paper gets rid oflocal standard deviation factor of the traditional threshold model, and uses integralimage to calculate the local average, using the obtained local average and grey valueof pixel to get local threshold value. Compared with the traditional local thresholdsegmentation, the new local threshold segmentation algorithm overcomes thereflective effect, reduces the time of segmentation. And the segmentation effect isperfect. It can suitable for several kinds of drugs. Finally, this paper finishes the visual inspection software and dynamic linklibrary design using VC++and OpenCV, and shows the interface of the software. Theexperiments indicate that the percentage of detecting the defects of missing, brokentablets and tablet fragments has exceeded99.2%with wide application prospect.