Aiming at the problem of micro tooth shape geometric parameters of rubber roller cannot be detected on the machine,a visual inspection method for rubber roller was proposed.Through a reasonable measurement device design,an on machine inspection system was built based on the principle of machine vision detection.The rubber roller detection was achieved through an on machine visual detection device cooperating with a linear reciprocating motion mechanism,a swing arm mechanism and a visual detection mechanism.The images of the rubber roller were obtained by the device,the image filtering,morphological analysis and other image preprocessing were performed for the rubber roller image using visual processing software.Subpixel edge detection method was used for precise positioning of thread contours.Then Harris corner detection algorithm was used to extract feature points.Finally,the least squares segment fitting method was used to obtain the rubber roller thread contour,and the main parameters of the rubber roller thread tooth shape were calculated to determine the qualification of the rubber roller.The experimental results prove that the proposed on machine vision detection method for the geometric parameters of the rubber roller tooth shape can achieve efficient and high-precision detection of the rubber roller tooth shape.