Research On The Surface Defect Detection Method Of Decal Ceramic Dish Based On Machine Vision

With the continuous improvement of science and technology,the automatic detection of surface defects of industrial products is bound to become the future development trend.At present,manufacturers mainly adopted the traditional manual visual inspection and sampling inspection method to detect the surface defects of the products,which has great labor intensity,high false inspection rate and greatly affected the production efficiency and the improvement of product quality.In this paper,after researching the full investigation of the surface defect detection methods at domestic and abroad,the surface defect detection method of decal ceramic dishes based on machine vision was proposed,and the corresponding defect detection system was designed.The research content is as follows:(1)With field investigations and literature surveys,the production process of decal ceramic dishes and the common types of surface defects in the production process were understood in detail.In this paper,a defect detection system based on image difference is designed in combination with product characteristics.The spatial geometric coordinate alignment of the image to be tested and the template image(without defects)is realized through image registration technology.The image to be tested after registration and the template image are processed by image difference,and the surface defect intelligent detection algorithm is used to determine whether the product is qualified.(2)The accuracy of image registration is a key factor to affect the accuracy of defect detection.Therefore,by focusing on the three main registration algorithms,this paper proposed an improved SURF+BBF fast image registration algorithm to meet the requirements of the system.The algorithm uses FAST(Features from Accelerated Segment Test)feature point detection+SURF(Speeded Up Robust Features)feature descriptor generation+BBF(Best Bin First)two-way matching algorithm to gain the initial matching point pairs,and then uses RANSAC(Random Sample Consensus)algorithm to eliminate the mismatched point pairs,calculates the homography matrix to obtain the affine variation parameters of the image to be tested,and completes the registration of the image to be tested and the template image.This algorithm has the advantages of high speed,high precision,good robustness to rotation,translation,illumination and noise.lhrough the verification of a large number of samples of decal ceramic dishes to be tested,the accuracy rate of defect detection was 95%and the false detection rate was 2.5%,reaching the expected target.(3)According to the attribute characteristics of decal ceramic dishes surface defect detection,it can be divided into three kinds of defects,which contains point defects,line defects and area defects.By comparing the advantages and disadvantages of each classifier and the suitable conditions,this paper finally used the S VM(Support Vector Machines)classifier to identify the defect types,and the accuracy of this classifier is only 86.7%because of restriction of the number of training samples,so the accuracy of the classifier can be improved by increasing the number of training samples and the kind of feature attributes.(4)The application program of surface defect detection of decal ceramic dishes was compiled based on MFC(Microsoft Foundation Classes),and many scattered algorithms were integrated into the graphical user interface to facilitate the operation of users.Finally,the application program could be operated stably.