Abstract: In order to improve the recognition accuracy of wheat leaf disease, the Gaussian mixture
model combined with EM algorithm was used to extract the wheat leaves and obtain the bigger target,
which made the segmentation accuracy higher than the direct segmentation disease area. And the
roughness, the degree of orientation and the contrast were selected by combining the HSV main color
histogram and the Tamura texture feature. The images of wheat healthy leaves, powdery mildew, leaf
blight and leaf rust were identified by random forest method and recognition accuracy is up to 95%.
Experiments show that this method is effective and superior to the support vector machine (SVM)
method under the same conditions.

Key words: Gaussian mixture model, EM algorithm, HSV main color histogram, texture feature, support vector machine