In order to solve the problems of low manual detection efficiency and difficult extraction of rubber ring surface defects,and improve the on-line detection speed and accuracy of rubber ring defects,a machine vision based rubber ring surface defect detection method was proposed.In this method,the multi-camera multi-thread image acquisition mode was used to collect local images at different positions on the upper and lower surfaces of the rubber ring.The image was adaptively median-filtered and then edge-enhanced.The Gaussian difference operator was used to extract the rough edges of the contour,and the Zernike moment was used to obtain the sub-pixel edge position.The Ceres library polynomial fitting was used to estimate the position of the breakpoint and update all the edge positions for the problem of discontinuous points on the edge.The image area of the rubber ring surface in the whole image was obtained according to the edge position,and the ring region of interest was mapped into the rectangular region.Singular value decomposition (SVD) was performed on the obtained image,and through connected domain analysis,the singularity in the image,that is,the location of defects,was extracted.The experimental results show that the rubber ring surface defect detection method based on singular value decomposition is robust and efficient,which can quickly and accurately find the surface defect information of the rubber ring.