Bearing Defect Detection Based On Machine Vision

Visual inspection technology has been widely used in industrial production and human life depends on its unique non-contact and non-destructive testing methods.As an important part of the machinery industry,the factory quality of the bearing is crucial.But in the process of production and assembly,pits,scratches and other defects will exist in the bearing surface inevitably.Because of the disadvantages of the current detection methods which consist of low efficiency and high rate of false detections.So in this paper,a set of bearing defect detection system was developed by the bearing surface defect detection algorithm.The main research of this dissertation can be summarized into the following aspects:(1)This thesis has established the camera imaging model and analyzed the cause of lens distortion.In addition,the camera's internal and external parameter matrix and the position relationship between the Cartesian coordinate system and the world coordinate system ware obtained through calibration experiment.(2)For bearing shape features and standard dimensions,this paper provides a new combination algorithm for rapidly extracting from the area of bearing dust cover.And the algorithm is composed of three steps.Firstly,to get the contour of the bearing's outer circle by using image binaryzation,morphology filling processing and Canny operator edge detection.Secondly,Least squares method was used to calculate the center coordinates and its radius.Finally,contacting the relationship of bearing's image and the actual size,the bearing dust cover was separated from the original image successfully.(3)The image preprocessing algorithm was designed by the gray level transformation and the median filter,which was used to enhance the contrast of the defect in image.Furthermore,In order to solve the problem that the OTSU method can't distinguish between the texture details and defects completely,the improved wavelet transform was adopted to weaken the texture details.Then,the output image is threshed by way of the OTSU method to complete the extraction of defects.Finally,to determine whether the defects exist,it can identify the number of connected domain in the area of bearing dust cover through the eight neighborhood marker method.(4)This paper builds a hardware platform of visual inspection system.The algorithm program and the host computer interface was written by Matlab.And the PLC achieves the process of controlling the Cartesian coordinate robot to sorting defective bearings.Designing experiment to analyze the experimental results combined with a given evaluation standard.