Research Of Pharmaceutical Packaging On-Line Detection System Based On Machine Vision

In recent years, the development of China’s pharmaceutical packaging is rapidly.Different drugs have different forms of the packaging.Among this,the forms of blisterpackaging are more and more loved by the people and it has been widely used anumber of fields.In blister drugs production,packaging quality inspection is anextremely important part.However,in the domestic,many companies accomplishpackaging quality inspection by manual operation.However,this method istime-consuming and can not guarantee the reliability of test results.Therefore,it needresearch the pharmaceutical packaging on-line detection system.The rapid development of machine vision technology provide an effectivetechnical support for non-destructive testing of drug quality.The paper use lack of grainand defects of pharmaceutical packaging as the study object as well as describe themethods and principles of the detection system.It focused on the tablet packagingdetection algorithm.The detection system include image acquisition module、an imageprocessing unit and the display module.The image acquisition unit use high-definitionCCD camera to complete the acquisition and transmission of live images.The imageprocessing unit through the industrial computer that use the specific algorithm toprocess the detected image.The control unit is responsible for the implementation ofcomputer instruction and remove substandard pills.The image processing part is the key to the whole system,but also the focus of thisstudy.In the first,it use the algorithm to prosess original image in order to enhanceimage and highlight the region out of interest and weaken the characteristics of otherregions.The key that real-time detection system can be accurate detection is the goodimage.In the image enhancement,it introduce the image graying,histogram equalization,median filtering algorithm and use this algorithm to prosess image toprovide a good image data for subsequent image segmentation.Secondly,It is imagesegmentation processing.Image segmentation is the basis of the successful extractiontarget.This article summarizes five categories of image segmentation and focus on thecommonly used threshold segmentation algorithm.The paper through study theclassical algorithm of threshold segmentation to propose the multi-zone algorithm thatbased on differential evolution.Using this algorithm cut up the image and extract pillsedge to calculate the geometric characteristics of the tablets.Finally compare to thegeometric characteristics of the test image and the standard value of the statistics tocomplete the template matching.Thus,it can achieve tablets defect detection.In thispaper,it use matlab to simulate the multi-zone algorithm and compare to the OTSUalgorithm.The results prove that the algorithm can achieve a better image segmentationas well as meet the detection requirements of the system.But this algorithm have theinadequate in real-time performance and this is also need to further study andimprovement.