Research Based On Machine Vision Of Plate Surface Defects Detection And Recognition Algorithm

With the development of science and technology and the market competition becomesincreasingly fierce, the demand of the appearance and the quality of the products becomeshigher and higher. The surface defect of plates is an important factor affecting the appearanceand the quality of the products, how to effectively detect the surface defects of the platesbecame an important issue of the nondestructive testing. The development of the machinevision and the defects testing that applied in the surface of the plates make the surface defectstesting more effective and more intelligent. Controlling of surface qualities of the productshas profound significance to reduce production energy consumption of the enterprises,improve the production efficiency, reduce the trade disputes, and maintain the enterprises’credits.This paper will introduce the principle and system construction of machine visiontechnology, make the defects testing of the surfaces of the plates and the automaticclassification based on machine vision technology, analysis the filtering of the images, thegrey level transformation, the edge detection of the images and the automatic classification ofBP neural network, the main research work are as follows:1. Drawing the surface defect images of6main plates, and contrasting image filteringprocession of the median filtering and mean filtering through the simulation experiments inorder to compare the advantages and the disadvantages of these two kinds of filteringprocession, and effectively reduce the Gaussian noise and salt and pepper noise, and protectthe detail of the images and the defective edges.2. Making gray transformation and gray equilibrium of the defect images, greatlyimprove the distinguishabilty of the defective images. Then combined the edge detection ofthe images and adopted five kinds of edge detection operators to carry out the edge extractionin order to get the most suitable Canny operator in the edge detection through comparison,which can display the edge details of defective images better. 3. Through drawing the morphological characteristics value, the gray characteristicsvalue and the texture characteristics value of the defective images, and identified the surfacedefects of plates automatically through BP neural network, the author designs the classifierstructure of BP neural network, and selects the samples to practice and test the classifier.Through the experiment, we can prove that the BP neural network can effectively identifiedsurface quality defects of the plates, and the surface defect recognition rate of the six mainplates can reach83.4%, which can better complete the defective image recognition.In this paper, the author applied the image processing technology of machine vision andneural network technology to the application of surface defects detection, which can betterrealize the automatic test of plates’ surfaces, and meet the demand of plates’ detection, so it isworth using and popularizing.