In this paper, a soybean classification method based on histogram and VS(valley size) of gray-level profile is proposed. Deformity of soybean can be classified as color change on surface and modification of external form. We sorted soybeans into groups of four types, good beans, surface-modified beans, entirely discolored beans and partially discolored beans and classified good beans from three modified beans using thresholds based on histogram and sorting algorithm based on VS. In order to obtain real-time processing for crops with huge population, proposed sorting algorithm was implemented based on integer addition. As a inspection result for 4 types of beans, sorting accuracy was 94.5%. Therefore, the proposed method showed equivalent performance with low computation loads compared to the previous sorting algorithm.