Research On Surface Defect Detection Of Bearing Parts Based On Polarization Vision

As a common rotary standard part of mechanical equipment,the parameters and performance of parts processing accuracy,rotary accuracy,bearing capacity and rated speed affect the operation efficiency and service life of mechanical equipment at all times.With the popularization and development of a variety of bearing surface defect detection methods,it is urgent to find a non-contact,non-destructive and efficient bearing defect detection method.For this problem,this paper takes the outer surface defect of the inner ring of centripetal cylindrical roller bearing as the research object,and designs a set of visual inspection system for the outer surface defect of the inner ring of bearing by using the combination of polarization visual inspection technology and traditional visual technology.In this paper,the common surface defects and smooth surface imaging characteristics of metal bearing inner ring are analyzed and studied,the optical imaging theory research and detection methods are summarized,the overall design scheme of polarization vision detection system is improved,and an effective surface defect detection optical path of bearing parts is established.Firstly,using the imaging principles of orthogonal polarization imaging and scattering depolarization as the important theoretical basis of polarization imaging,an optical path system for detecting the surface defects of metal bearing inner ring is designed.Based on the analysis of the traditional visual imaging method,a forward image acquisition method is proposed,which selects the linearly polarized light with short wavelength（450nm）to irradiate the outer surface of the bearing inner ring in a grazing incidence mode under the condition of dark field,and makes the grating axis corresponding to the polarizer and the polarizer orthogonal.Through the imaging theory,the polarization detection system is designed to ensure the accurate extraction of bearing surface defect information.Secondly,the optical elements and imaging equipment in the optical path are reasonably selected to determine the relative position of each system in the detection optical path.According to the information complementary principle of polarization component image and light intensity image,the polarization camera and linear CCD camera are used to collect the image of the defect position of the same curved surface,and then the experimental platform is built.Zhang Zhengyou calibration method is used to calibrate two industrial cameras to obtain one-to-one corresponding pixel level images.Finally,according to the characteristics of surface defect images of bearing parts collected by two cameras,an image processing algorithm for surface defect detection of bearing parts is designed.The polarization image is preprocessed by combining morphological filtering and differential operation,and the linear array camera image is processed by median filtering.The two images after preprocessing are fused and preprocessed with the same polarization image.The second level stationary wavelet transform is decomposed,and the multi-form wavelet coefficients are processed by bilateral filtering.Then the image is reconstructed by inverse wavelet transform,and then the reconstructed image is segmented by Otsu image to obtain a clear and complete defect feature image.Using this method to identify the surface defects of centripetal cylindrical roller bearing,the accuracy is more than 95%.