In view of the complexity and fuzziness of the diseased leaf image, a fast detection method of diseased area of plant leaf based on gravitational kernel density clustering algorithm is proposed. Firstly, the R-channel value of the diseased leaf image is extracted in RGB color space. According to the characteristic histogram characteristics of the diseased leaf R value, the characteristic histogram curve is fitted by polynomial. The peak point and the peak area of the fitting characteristic histogram curve are determined according to the derivative property. Then, the clustering number and the initial clustering center of diseased leaf images are determined according to the peak area and the peak point. Secondly, according to the preliminarily determined clustering center of the diseased leaves’ image, the gravitational kernel density clustering algorithm proposed in this paper is used to quickly segment diseased leaves. The experimental results show that the average segmentation precision based on gravity kernel density clustering algorithm is more than 80%, and the average detection time is 4.912 s, which is better than the performance of existing disease region segmentation algorithm K-means and Meanshift.