Abstract: To detect the roundness and the black core defects of sintered spherical abrasive block, a defect detection method based on machine vision was proposed. Firstly, an image acquisition system of the abrasive block was built by using single chip microcomputer, stepping motor, sampling disc, digital microscope and upper computer to realize the continuous acquisition of the abrasive block images. Secondly, the image graying, the threshold segmentation and the morphological processing were used to extract the abrasive block area and the black core defect area. Thirdly, the roundness of the abrasive block area and the size of the black core defect were calculated. Finally, the detection threshold of the abrasive block defect was determined through the detection test. The results show that the method can digitally detect the roundness and the black core defects of sintered spherical abrasive block. It can also analyze the problems of in the preparation process of the abrasive block, thus providing feedback basis for the improvement of the abrasive block preparation method.