Abstract:In order to solve the practical problems of low accuracy and low efficiency of metal hose joint surface detection, a set of joint surface defect detection system was designed based on machine vision.In order to improve the contrast of the image and extract the target detection area effectively, image filtering and denoising, Otsu algorithm binarization and image morphological analysis were used.Then, based on Canny edge detection algorithm, the edge contour of the image was accurately identified, and the fast Fourier transform method and R-FCN algorithm were used to calculate and analyze the defect feature information for fast matching extraction 〖JP2〗and defect classification processing.The test results show that the defect detection system can effectively improve the detection efficiency,〖JP〗 ensure high detection accuracy rate and precision, meet the needs of actual industrial detection, and has good practical value.