The Study Of Detecting System For Verticality & Foreign Particle Of Glass-Bottle Based On Industrial Computer Vision

The detection of verticality and foreign particle of glass-bottle is veryimportant in the course of production of glass-bottle. At present, many domesticfactories adopt the artificial method to detect the quality of their productions,which not only is inefficient and inaccurate but also lacks of uniform inspectionstandard. A few domestic factories introduce into foreign machines which adoptmechanism-touched mode, but these machines are low agility and require high cost.Especially, there isn't a mature machine that can detect the verticality ofglass-bottle by far. In this paper, a new on-line inspection system of glass-bottlehas been studied and developed by author, which synthesizes none-touched sensorsand modern pattern recognition theory and can accomplish the non-contactautomatic inspection of product quality. On the other hand, the system has virtuesof high cost performance and high applicability as well as convenient operationinterface. The measure of verticality of glass-bottle is based on the computer stereovision technique, and the step of measure is following: Firstly, capturing twoimages of glass-bottle with two cameras from the vertical directions at the sametime. Secondly, we use Canny arithmetic operator to get the edge of image, thentrack and link the contour of glass-bottle. Thirdly, based on the pure contour, wecalculate the bottle mouth center, bottle bottom center and bottle center axis.Fourthly, after calculating the angle between center axis and the line of bottlemouth center to bottle bottom center, the one verticality(V1) of glass-bottle equalsto the product of bottle normal height and tangent of the angle, with the sameprocess, we can calculate the other verticality(V2 ) of glass-bottle. Lastly, the totalverticality equals to V1 +V2 . In the paper, we analyze two primary measurement 2 2errors and their influence on the accuracy of survey. Experiment result showed thatthe result of survey is accurate and its relative deviation is 6.1%～7.7%. Together with the reality, through the analyzing and concluding a lot of imagesof bottle bottom with foreign particle, the author put forward a divisionalself-adapting solution to extract the foreign particle of bottle bottom, in which thebottle bottom is partitioned into three regions and the threshold of foreign-particle-growth is different in different region. The image is captured through thebottle mouth while the bottle bottom is illuminated. Some algorithms, such asimage segmentation, Hough transform and region growth, were used to process the IIAbstractimage captured. After separating the suspectable foreign particle, we extract itsfeatures and judge true or false by these features. Experiment result showed thatthe lowest rejection criterion is 1.5mm×1.5mm and the highest precision is 95%. In order to improve the processing speed of inspection system more, optimizerof compiler and MMX(Microprocessor Media Extension) technology are appliedto optimize the kernel arithmetic and the whole processing software. VC++6.0 is adopted as a development platform. All the algorithms andfunctional modules include in this paper have been implemented by programming.All results have been analogized with computer.