Leaf Image Analysis-Based Crop Diseases Classification

Eminently, the countries of developing state have their economy based on agricultural crop yieldings. To retain the economic growth of these countries, the agricultural plants’ disease detection and proper treatment are a leading factor. The work available in the literature basically features pull out to classify the leaf images due to which the classification performance suffers. In the proposed work, we tried to resolve this rough image dataset problem. The proposed technique initially localizes the leaf region by utilizing the color features of the leaf image followed by mixture model-based county expansion for leaf localization. The classification of the leaf images depends on the features of discriminatory properties. The characteristics features of the diseased images show various types of patterns into the leaf region. Here, we utilized the features discriminable property using the Fisher vector in terms of different orders of differentiation of Gaussian distributions. The performance of the proposed system is analyzed using the PlantVillage databases of common pepper, root vegetable as potato, and tomato leaf images using a multi-layer perceptron, and support vector machine. The implementation results confirm the better performance measure of the proposed classification technique than the state of arts and provide an accuracy of 94.35 % with an area under the curve 94.7 % .