Abstract: Fruit shape is an important index of mango grading. In order to avoid the influence of human factors on Mango grading, image processing technology was used to detect mango shape. In this paper, the evaluation index and method of mango shape were put forward. Cluster analysis was carried out to determine the grading basis of each evaluation index. The evaluation method was verified by experiments and the results show that the accuracy of mango shape evaluation by combining shape index and plumpness is up to 92%.