Study On Sandpaper Defect Detection Method Based On Computer Vision

Sandpaper is a frequently used grinding material.Sandpaper defects have many kinds,shapes and positions.At present,sandpaper defect detection mainly depends on manual detection.This method has many disadvantages,such as low detection accuracy,high labor intensity,high labor cost,limited defect detection type,low degree of automation,etc.Because the visual detection has the advantages of non-contact,high speed,high precision,good automation and so on,sandpaper defects can effectively be detected by visual inspection technology,therefore this paper study the sandpaper defect detection method based on computer vision.This paper carries on the analysis and classification of sandpaper defects,and sandpaper defects are divided into easy identify sandpaper defects,difficulty identify sandpaper defects,and sandpaper printing defects.Study on the influence of sandpaper defect image from the change of lighting direction and shooting direction,build a sandpaper visual defect detection system,hardware selection and software design are carried out,the vision system is calibrated and the accuracy of the calibration method is verified.Easy identify sandpaper defects include scratch,flod mark,rubber spot,lack of sand.The characteristics of sandpaper image noise is analyzed,the median-mean filter algorithm is improved,and local images containing defects are obtained.The length,width and area of sandpaper defects are obtained by adaptive enhancement,defect segmentation,morphological operation and defect size calculation.Difficulty identify sandpaper defects include sand mass,slight seal,dark mark.The statistical characteristics and Gabor features of small sandpaper image are analyzed,feature vector based on statistical features or Gabor features is constructed.Through support vector machine and two kinds of feature vector,the classifier is trained,recognized and classified.Experimental results show that good image recognition and classification results can be obtained by Gabor feature vector and nonlinear support vector machine algorithm.Sandpaper printing defects include printing characters unclear,ink pollution.The characteristics of sandpaper print image is analyzed,sandpaper print defect detection algorithm based on template matching is designed,the operation speed is improved by the image pyramid method,using the edge image as matching image,the template image is rotated and scaled to fit the rotation and scaling of image matching.Sandpaper print defects are extracted by image affine transformation and image subtraction method.The experimental results show that the sandpaper print defects detection algorithm based on template matching has high detection accuracy and speed.