In order to realize automatic recognition of the jewelry，a set of jewelry recognition system based on machine vision is designed，and the corresponding algorithm is proposed. In this study，8 different kinds of jewels are used as experimental samples.The color threshold segmentation method is used by the componentsofG+R. Binary image is filtered with a template of 27×27 pixel. A method of edge extraction is proposedbased on values of pixels combined with the threshold by the 2×2pixel window.The features of areaand the distancefrom image centroid to edgeare extracted.The values of area are judgedwithin the limits of given values，and the cross-correlation algorithm is applied to the values of distance feature. The experimental results show that the precision rate of the method is 100% and average recognition time for each is about 3 seconds，which meets the requirements of practical application.