The Detection System Of The Cylindrical Roller’s Surface Defect Based On Machine Vision Technology

As a key component of bearing, the surface quality of cylindrical roller is animportant indicator to assess the roller. The presence of surface defects seriouslyaffects the quality and performance of bearing. During a long time, the most domesticbearing manufacturers mainly rely on manpower to complete defect on the surfacedetections of cylindrical roller. But the testing results have the poor consistency ofproduct quality, low-efficiency of detection, and missed-detection and false-detectionof defections, etc. Therefore, it is an urgent need to ameliorate the means of detectionto alleviate the contradiction between the high speed fabrication and low-speeddetection. Machine vision technology as an image engineering developed rapidly inrecent years, in this paper, the research about the real-time detection system of roller’ssurface defects and the image processing algorithms of roller’s surface defects havebeen studied which based on machine vision technology, the purpose of identifyingroller’s surface defects has been achieved through the decent algorithm.Based on the research of the surface defect about the detection technology ofroller, aimed to the shortcomings of artificial visual inspection, the real-time detectionmethod based on machine vision about the surface defect of bearing roller has beenproposed. According to the testing requirements，the automatic flatten system aboutthe surface of roller has been developed, and the CCD camera is installed on theexpand track to acquire the image of roller surface. In order to overcome the reflectiveeffect of roller surface and the bad effect of the collected surface image, combinedwith experiments, the red LED is selected as the detection light. The illuminationmethod which the two light source obliquely light the roller from the top has beentaken, then the clear and true surface image of roller were acquired, high surfacequality of roller image has been acquired.According to the real-time detection environment and the characteristics of theroller, by analyzing and studying the pretreatment operators of image, the median filtering algorithm has been selected for the image processing. This can effectivelyremove the noise on the influence of the image and lay the foundation for the next stepof image segmentation. The advantages and disadvantages of kinds of edge detectionoperator has been introduced, by analyzing and studying the image handled by variousoperators, the Robert operator was selected for the edge detection operator of thissystem, the availability and rapidity were ensured. Finally the defect featureparamenters of the roller’s surface images were extracted, the single and whole area ofdefect in the roller’s surface were calculated, which provided information for roller’sclassification.