Recognition of rice leaf diseases based on salient characteristics

The paper takes the advantage of colorimetric model, ellipse fitting and support vector machines to recognize the chronic type, acute type and white type of rice blast. First of all, with the help of the difference of lesion manifestation in HSV space, the images which are converted in HSV color space are respectively segmented by a green detection method and an OTSU algorithm, and the best segmentation result is selected for each sample by an evaluation mechanism. Then, five feature components based on morphology and color are proposed as the basis of classification model. To test the validity of these features, one of the three types of lesions is used as a positive example and the other is used as a counterexample for statistical validation. The results show that the differences between the three types of samples are significant. After that, a method of multi-classification SVM based on radial basis function is chosen to classify the three types of lesions. Experiments show that when compared with other methods, the proposed algorithm has a high recognition accuracy and can effectively identify three types of rice blast diseases.