Research On Surface Defect Imaging And Identification Technology Of Metal Shaft Parts

Fluorescent magnetic particle detection is one of the nondestructive testing methods which is often used to detect the surface flaws of ferromagnetic material.Aiming at the requirements of the defect detection for the metal axial’s surface,the fluorescent magnetic particle imaging surface defect inspection system based on machine vision and image processing technology was designed.For flaw recognition,several key algorithms of the image processing,including image preprocessing,image segmentation and defect extraction were studied.Firstly,the perspective distortion of the step shaft was corrected by using the imaging model based on the vanishing point algorithm.Then,the cylinder surface of the shaft was unfold in order to correct the distortion introduced by inconsistent depth resolution.Different unfold images were stitched by making use of the phase correlation algorithm.In order to achieve a higher SNR and contrast,the wavelet threshold image de-noising algorithm and multiscale Retinex image enhancement algorithm were studied.For the aim of the shaft extraction in the original image,image segmentation algorithm based on edge detection was studied.In consideration of the drawbacks of the general algorithms,the paper proposed the combination of the adaptive threshold Canny operator and the Hough transform,and effectively extract the upper and lower edges of the shaft.Then,the improved Otsu threshold segmentation algorithm was studied to solve the problem of the smaller foreground.The characteristics of the defects were summarized,and the search for minimum circumscribed rectangle and the calculation of the length,width and direction angle was completed using center spindle algorithm.Since there are branches and cross in cracks,a crack recognition algorithm based on node characteristics was proposed,and use it to implement the classification of the defects.Finally,the inspection system was formed on the basis of the former design,and this system was used to do the experiments for an artificial shaft and the real products.The results are consistent with artificial results,achieving the goal of the identification for the Φ 0.5 mm circular defect and the 1.5 mm × 0.5 mm linear defect.These experiments verify the effectiveness of the system.