Research On Key Image Processing Techniques For Automotive Surface Defect Detection

During the production and transportation of automobiles,defects and damages on the automobiles surface are easy to occur,which has become an important factor affecting the quality of automobiles.However,most of the current automotive surface defects are detected by manual methods which has the disadvantages of high miss rate and low efficiency.This thesis studies a defect detection scheme for automobile appearance based on machine vision.The main contents include the design of image acquisition scheme,and research on image registration method and defect detection algorithms.First,based on the morphological characteristics of the car-surface defect and the strong specular reflection effect of the car surface,an image acquisition scheme based on the deflectometry technology and image fusion technology was designed.This solution can obtain an image of uniform illumination on the surface of the vehicle body,avoiding the phenomenon of light reflection on the surface of the vehicle body,and makeing small defects captured by the camera.Secondly,in order to improve the speed and accuracy of image registration,an image registration algorithm based on SURF dimensionality reduction and FSC is proposed.PCA was used to reduce the dimension of SURF’s 64-dimensional descriptor vector,and then the matching point pair was purified using FSC algorithm.Experimental results show that compared with the SURF-RANSAC algorithm,the PCA-SURF-FSC algorithm proposed in this paper improves the speed and accuracy of registration,and the quality of image stitching is better.Finally,based on a comparative analysis of several typical defect detection algorithms at present,an adaptive defect detection algorithm based on the joint mapping of VAR contrast image construction and Sobel edge detection is proposed for automotive surface defect images with environmental background.The defect detection algorithm requires few artificially set parameters.The experimental results show that the defect detection algorithm proposed in this paper can effectively detect defects in the appearance of the car,and the defect detection effect for small areas is obvious.In this paper,the research work and achievements of automobile appearance defect detection can be used in the detection of surface defects of aerospace products...