Aiming at the requirements of intelligent production of deep processing of fresh shiitake mushrooms (Lentinus edodes), the research proposed an intelligent grading method of fresh shiitake mushrooms based on machine vision and designed an automatic grading system. The image of fresh shiitake mushrooms can be divided into three areas, including pileus, turned edge and lamella. The geometric characteristics of each area are the main basis for classification. Firstly, the manual intervention was used to divide and identify a limited number of samples, and establish a Gaussian mixture model; Secondly, the system was trained to identify samples and the identifying model was established, which the parameters could be used to distinguish the intelligent segmentation of shiitake mushroom areas from background. Finally, the geometric features, including roundness, diameter and uniformity of each area were quantified, and different grading standards were set, and then the classification of shiitake mushrooms was realized. The simulation results showed that the reliability, speed, stability and other indicators of the newly-designed automatic classification system had glaringly obvious advantages, compared with the traditional manual classification method.