Research On Surface Defect Detection Of Bearing Roller Based On Machine Vision

Bearing roller as an important foundation component in rolling bearings,its surface quality affects the performance of mechanical equipment.At present,most bearing roller manufacturers are still using traditional manual visual inspection methods to detect the bearing rollers surface defects.However,it has low detection efficiency and high false detection rate and can't no longer meet the needs of enterprise production development;The existing nondestructive testing schemes,such as ultrasonic and eddy current,have the disadvantages of complex operation and low calibration efficiency machine vision,as a new detection technology,which has been widely applied in various fields of industry and has made some progress in the field of bearing roller surface defect detection,but it still has some shortcomings: Blurred image features,Low accuracy image processing method,and less detected defects.Because of the above reasons and on the basis of studying the existing detection algorithms of bearing roller defects,the paper designs an image acquisition and acquires the bearing rollers image.Then the Niblack improvement method,the gray scale compensation method and the image optical flow method are applied to the bearing roller surface defect detection.The main contents of the work are as follows:(1)This paper studies the characteristics of the camera,lens,light source and considers the actual defect detection requirements,we design a set of bearing roller image acquisition device,which uses the backlight source to illuminate the bearing roller side to get the more obvious image in the defect area and overcome the limitation of the traditional light source.Then the bearing roller area is extracted from the original image to reduce image processing time.Meanwhile,in order to reduce image nosie interference,the median filter is applied to the extracted image.(2)An improved method,which is based on the characteristics of the image that has the similarity of the graysacle in the vertical direction and the adaptive Niblack algorithm,is proposed.This method,which only need one parameter can complete the segmentation of image defect area,is efficient and simple.Experiments show that this method can detect all kinds of defects of bearing roller effectively,meanwhile its efficiency and precision are higher than the traditional image threshold segmentation method.(3)In order to solve the problem of uneven global gray value of tapered roller image,an improved gray compensation method,which compensates the image grayscale in horizontal direction and vertical direction by the projection of ideal gray value of image,is proposed.Then the histogram clustering segmentation is applied to the compensated image,it overcomes the shortcomings of the traditional threshold segmentation method.Experiments show that the improved gray compensation method makes the image more balanced and improves the contrast between the defect region and the background area.Compared with other segmentation algorithms,the improved method has a higher accuracy rate.(4)A detection method for surface defects of bearing roller based on image optical flow was proposed.First of all,this paper describes the theoretical basis of image optical flow for surface defect detection of bearing roller.Secondly,in order to eliminate the influence of abnormal gray value on optical flow computation,singular value decomposition is applied to the original image.On the selection of the estimation model of the optical flow calculation,this paper uses smoothing strategy combining local constraints and global constraints,to improve the robustness of the algorithm.In the solution strategy of optical flow,the Pyramid stratified thinning method is introduced to improve the accuracy of the large displacement of the optical flow calculation.Through the pseudo-color images generated by the optical flow,the rough location of the defect area is completed.At last,the threshold method is used to divide the defect area.The experiment shows that the proposed method is effective and feasible for the detection of the surface defects of the bearing roller.