Abstract: To achieve precise application of pesticides and improve yield and quality of rapeseed, rapid and accurate detection of disease occurrence is very important. In this paper, a machine vision-based classification detection method for sclerotinia on rape leaves and stems was proposed, which was mainly based on the proportion of leaf diseased spots and the vertical expansion length of stem diseased spots, and which used the color difference between diseased spots and healthy areas. The HSV color space model was used to segment the target area. First, the image was converted from RGB image to HSV image, and then the HSV component was used to traverse all the pixels in the image to extract the area of interest. Rape leaves were mainly drawn by ROI and completed leaves. The contour was then used to calculate the area. Owing to the complex environmental background of the stalk image, the entire stalk area was obtained from the complex background through the Gaussian mixture model as the target area prior to HSV color model segmentation, after which the lesions in this area were segmented. The contour drawing method of the minimum circumscribed rectangle obtained the length of the longitudinal expansion of the lesion, and then graded the degree of infection. Experiments showed that this method can effectively classify the disease degree of leaves and stems, and its recognition accuracy was 94.25% and 92.5%, respectively, with high accuracy and robustness, which can provide a theoretical basis for precise pesticide application.