Study On Visual Detection Method Of Aviation Bearing Assembly Based On Local Characteristics

Bearing assembly detection is an important part of aviation bearing safety.However,the accurate detection of bearing balls based on limited local information is the difficulty of bearing assembly detection.In this paper,based on the study of the segmentation and identification method of the bearing ball based on local characteristics,mistaken and leaky balls of the aviation bearing assembled are detected.Firstly,the image surface characteristics of the bearing are analyzed by using the threshold segmentation method of image processing and the optical principle.And the image acquisition system and the ball detection method of the aviation bearing are designed.In order to solve the problems of the interference of lubricating oil on the bearing surface and the halo on the bearing surface,the light source of the acquisition system was designed from the light source,light color and lighting mode to obtain high-quality bearing images.Then,aiming at the ball detection method of aviation bearing,according to the high-precision requirements of aviation bearing detection,combined with the characteristics of local dynamic distribution of bearing ball,U-Net network is used to realize the automatic segmentation of aviation bearing ball.However,the original network is not clear about the boundary of ball bearing segmentation.According to the results,the reason is analyzed.A dynamic local ball segmentation network model based on U-Net network is designed to achieve the accurate segmentation of the local ball area of aviation bearing.On the basis of the segmented local ball image characteristics,Hough circle detection method is used to detect the local ball.According to the gradient information of edge detection and the known ball size information,Hough circle detection algorithm is improved to improve the Hough circle detection accuracy and precision.Finally,the experimental verification of the above algorithm shows that the dynamic local ball segmentation model based on U-Net network achieves more than 99% accuracy in local ball segmentation of bearings.At the same time,based on the precise segmentation of U-Net network,the Hough circle algorithm was designed for circle detection.And the error detection rate of mistaken and leaky balls in aviation bearings was less than 3%.This method can provide a new method and idea for the detection of mistaken and leaky balls after the aviation bearing assembly.