Research On Parts Quality Inspection Based On Machine Vision

In recent years,machine vision technology is the hotspot in many fields.With the development of machine vision technology,non-contact,high precision and fast speed are the advantages of machine vision technology,so it is widely used in many fields,especially product defect detection and product size detection.After machining the parts,the quality inspection is an important part of the work,the current inspection of the thread defects of the bolt and the size of key parts of valve are still mostly in the form of visual inspection or traditional inspection,it results in low detection efficiency and low precision.In this paper,we focus on the image processing technology to build a detection platform based on machine vision and do some research on hardware selection,camera calibration and image processing.The bolt and valve are used as the research object,at last the thread defects of the bolt and the size of key parts of valve are detected and the actual requirements are achieved.For the detection of the thread defects of the bolt,the characteristic parameters of the thread texture are mainly studied,and the determination is established by texture.According to the three states of the thread(normal,curl and scratches),the thread images are corrected with the determined camera parameters,and the gray image is processed by gray scale and median filter.15 characteristic parameter of the thread feature are extracted by gray gradient co-occurrence matrix and the typical characteristic parameters are selected by hierarchical R-type cluster,and finally the Bayesian discriminant function and the BP neural network system are used to detect the thread defect of the bolt based on the typical characteristic parameters.The detection rate of the two methods has reached more than 92%,the two methods meet the actual testing needs.According to the actual situation,BP neural network is easier to achieve.For the valve geometric parameter detection,the sizes which include the total length,stem diameter,disk diameter,groove diameter and cone angle of the valve are mainly studied.The valve image is corrected with the determined camera parameters and the calibration of the measurement system is completed.The image preprocessing of the valve image is carried out by image filtering,enhancement and binarization,and the image is enhanced by fuzzy operatorand binarized by OTSU.The canny operator is used to detect the edge of the valve image,and the edge is improved by binomial interpolation.Finally,the least squares linear fitting and least squares circle fitting are used to detect the total length,stem diameter,disk diameter,groove diameter and cone angle of the valve.Tests show that the proposed method can meet the requirement of the size detection of key parts of the valve.