Abstract: A seed's graphic usually has a quasi-convex boundary and symmetry, the development of its simple and rapid identification methods have practical significance. Using the pseudo minimum bounding rectangle (PMBR) and centroid, 8 of a object's feature informations were extracted. Based on the coefficient of shape difference (COSD), the similarity between multi-feature objects were measured. Based on the COSD in a class couple, a multi-feature graphic object alternative identification method was developed. Based on the mean of combination coefficient of shape difference (MOCCOSD), a multifeature graphic object one of many identification method was proposed. Identifications of rice and other seeds validated the methods.