Research On Surface Defect Inspection Technology Of Electronic Components Based On Machine Vision

With the rapid development of science and technology,the automation product line in industrial production is maturing.However,most of the surface defect detection still depends on manual inspection.The subjective detection is high,the cost is high,and the visual fatigue is easy to occur.Surface defect detection technology has not yet reached a complete solution.Therefore,it is necessary to study an efficient and objective detection method to meet the needs of automated,high-efficiency,and high-precision battery surface defect detection.This topic puts forward a method of surface defect detection for electronic components based on machine vision based on a thorough investigation of domestic and foreign surface defect detection methods.The method not only satisfies the requirements of automatic detection but also can effectively solve the difficulties of traditional automatic detection.Firstly,the image preprocessing algorithm is proposed for the effects of the collected illumination and noise,including the mean,median,Gaussian,and bilateral airspace filters,as well as the ideal low pass,Butterworth,and Gaussian low pass frequency domain filters,respectively.The filter is compared with the horizontal and vertical effects,and the Gaussian filter is used to perform smooth denoising.The image enhancement algorithms such as histogram,CLAHE and wavelet contrast enhancement are introduced.The self-adaptive threshold and Otsu binarization algorithm as well as morphological analysis and processing algorithms are introduced.Then the multi-objective detection algorithm was studied.The Canny edge detection algorithm and contour representation were used to extract the region of interest.According to the different features of scratch defects and surface defects,a scratch defect extraction algorithm based on Gabor filter is proposed.The image registration algorithm based on Fourier transform is studied.The matching between the target image and the template is completed.The template feature subtraction is used to extract the defect feature region.The design of the defect area area,area perimeter area ratio,area minimum circumscribed rectangle aspect ratio,area gray mean value,and contour area and convex area ratio are represented by five defect area features.According to the feature image,a feature classifier based on SVM and CNN is designed to compare the two recognition effects.In the end,this paper designs a machine vision inspection system for electronic component defects.Based on the difficulties of imaging and the types of defects,the overall structure design is carried out from both the hardware system and the software system.The effectiveness of the inspection system is verified with the button battery.