Study On Polished Surface Optical Imaging And Image Processing Algorithm Based On Machine Vision

Under the trend of industry system shifting from labor-intensity to technology-intensity,the traditional product quality control method which mainly depends on artificial visiondetection is facing more and more problems, such as low reliability and longtimeconsumption. The product quality detection method based on machine vision system hasbecome one of the main developing directions of industrial automation which can realize leanproduction and total quality management.There have been a variety of visual systems for appearance quality detection at home andabroad, but most are for non-reflective parts and few only concern highly reflective ones.Based on this situation, we conducted a deep research on the structural features and opticalimaging characteristics of coating plastic parts. In this paper, a vision system for imaging,detecting and characterizing defects onto highly reflective surface is presented. Thecornerstone of this inspection system is the innovative lighting system. Compared withoff-the-shelf lighting systems, the illumination condition designed in this paper can revealdefects visibly which appear white on a mid-gray background. Thank to this, the defectsegmentation of aimed image which is synthesized from a series images collected of one partsis straightforward and without any image preprocessing.In-depth survey and evaluation of commonly used thresholding technique are givenduring image segmentation and then a revised thresholding method that based on iterativemethod has been developed, which cannot just only segment the micro-defects laid on thehuge background successfully, but also has a strong robustness effect against noisedisturbance. After defects are successfully segmented, use the method that combines thearea shape feature and Hu moment invariants to describe the characteristics of the defects, andat last use the minimum distance classifier to identify and classify them.