Study On The Surface Defect Detection Method Of Tubular Materials Based On Machine Vision

The surface defects of tubular materials(This paper later called surface defects)is the key index to evaluate whether the products are qualified or not.However,in the production process of tubular materials,due to the influence of production process,production materials,production equipment and other factors,it is inevitable that there will be spots,scratches,cracks,pits and other surface defects,but also seriously reduce the quality of products.The traditional defect detection method is to select the defective materials manually,which is subject to human factors and low efficiency.In order to meet the requirement of high accuracy and high speed in the detection of tubular materials,this paper presents a defect detection and classify scheme for tubular materials based on machine vision.This paper focuses on the following contents:(1)In this paper,an implementation scheme of detecting,identifying,judging,classifying and selecting the ground for the tubular material detection system is presented.Firstly,the materials are detected and identified by server A to determine whether they are defective materials.Then,the defective materials are summarized and classified by server B.Finally,the defective materials are selected and removed by the mechanical mechanism controlled by PLC.The common defects of material surface are introduced through the collected images and the causes are analyzed.(2)In order to solve the problem of material image preprocessing,an improved median filtering mixed noise image processing algorithm is proposed,which can effectively remove salt and pepper noise and suppress gaussian noise.The filtered image is enhanced by piecewise linear grayscale to improve the image quality.The method of Hough transform was used to correct the inclined material.(3)According to the situation of material rotation,small camera angle and long material length,this paper adopts the scheme of rotation splicing first and piecewise splicing later.Firstly,cylindrical backprojection algorithm is used to expand the surface of the material,Then,SIFT algorithm is used to extract feature points to achieve image mosaics,Finally,weighted average algorithm is used to carry out stitching operation on mosaics image to achieve uniform image transition.(4)Canny edge detection algorithm was used to segment the image of defects after stitching,so as to realize the separation of defects from the background,and a keep a large area and cut off a small area algorithm was proposed to complete the localization of defect areas.Aiming at the problem of parameter optimization of support vector machine(SVM),this paper adopts the genetic algorithm(GA)to optimize the parameters,and then takes the feature vector after the dimension reduction of principal component analysis as the input parameter of SVM,so as to realize the accurate classification of defects.