Research On Extraction Method Of Glass Detect Characteristics Based On Double Threshold Segmentation

Because of restriction of technology and production environment, it is inevitable to makedefects in glass production. The degree of glass defect affects the grade of glass quality. It playsan important role in adjusting process and guiding production for judging the type of glassdefect accurately. Thus it will have important practical value for improving the quality andproduction process of glass to research glass defect’s extraction and recognition technology.The paper make glass defect as object of study. Firstly it analyses national current standardof glass appearance so as to ensure the detecting standard of glass defect. Then it lists theoverall structure of testing device for glass defect and determines the selection plan of mainhardware. Last, through the testing device, it gains typical glass defect image whose is obvious.Based on analyzing typical defect image, the paper adopts the top hat transformation method toeliminate the influence of trend background from area free from defect. Meanwhile, the paperadopts two kinds of image enhancement technology to improve the contrast of glass defectimage, which is based on gray level transformation of the image enhancement technology andbased on the morphology of the image enhancement technology. The contrast result showsenhancement technology is based on morphology of the image, which not only maintains thecharacteristics of glass defect, but also enhances the image effect of defect edge.In order to calculate the degree of the defect and recognize the type of defect accurately, onthe base of analyzing the iterative method and difference between maximum cluster methods,the paper puts forward a kind of improvement method based on glass defect characteristics,named double threshold defect segmentation. The method divides the gray level image of glassdefect into Ternaryzation image with background, defect and edge. The paper realizes thecalculation of defect’s degree through counting the pixel number of connectivity defect, and adopts scanning method to mark the quantity of defect. Then according to the area of eachdefect and edge, the paper realizes classification for typical glass defect and puts forwardclassification method.