Research On PCB Defect Detection Algorithm Based On Machine Vision

In recent years,the PCB industry has achieved rapid development.PCB is developing towards miniaturization,multi line and high accuracy,so the quality of PCB has become the hot spot of attention.The traditional detection method can not meet the development needs of PCB.With the development of image processing and machine vision,Automatic defect detection technology of machine vision has replaced the traditional detection technology.The PCB defect detection system based on machine vision includes illumination system,image acquisition system and image processing system.According to the actual production situation,the thesis is studying image preprocessing algorithm,image registration algorithm and defect detection algorithm based on the application of PCB defect detection.The main work and achievements are as follows:(1)The system composition and image preprocessing of PCB defect detection are analyzed.According to the characteristics of the image collected,the image preprocessing is proposed by using median filter and gray transformation because of the unique production process of the PCB and the noise interference when collecting the images.The images contrast get better.(2)The basic arithmetic of image registration is analyzed,then the existing techniques of image registration and the traditional registration algorithm are compared and analyzed.The feature points extraction algorithm of improved Harris is proposed,and the matching efficiency is improved.There are many error matching problems for the feature points extracted,so the two image matching algorithm is proposed after feature extraction.First,rough matching with NCC algorithm,then use the exact matching of RANSAC and the exact feature points matching are obtained.According to the affine relations of two images,the size and rotation deviation of the image are corrected,so the image registration is completed.(3)A PCB defect detection algorithm based on difference detection is proposed,In this paper,according to the different the defects are spread all over the circuit board and the different size of the defects.First,the standard image is subtracted to detect the defect image after matching,then,the two value images are obtained by using the threshold segmentation algorithm,and then the defecsts are processed accurately according to the correlation and morphology filtering of the defect region,filtering false defects to achieve accurate positioning.Finally,a defect recognition algorithm based on the connected domain features is proposed.In view of the characteristics of different defects,the comprehensive method,.the connected domain method and the Euler number method are used to classify the defects.