Abstract: The method of image recognition was used to distinguish the common diseases of millet leaves and provide scientific basis for the development of reasonable disease control measures.The original images of four kinds of leaf diseases including millet blast disease, white hair disease, red leaf disease and rust diseasewere were collected.The maximum variance method based on super green characteristics was used to segment the leaf diseases of millet. A total of 19 features of millet leaf disease were extracted, including color, shape and texture. The experiment adopted ant colony optimization algorithm to select 8 features. Cuckoo search (CS) was used to optimize the penalty factor c and radial kernel function g of SVM, and SVM was used to automatically identify millet leaf diseases. When c=80.2662 and g=1.8467, the average identification rate of millet leaf disease and leaves was reached 99%. These results showed that the method of image recognition based on support vector machine was feasible for the classification of four millet leaf diseases.

Key words: millet leaf disease, super green feature, ant colony optimization, cuckoo algorithm, support vector machine