Comparision of Performance of Classifiers - SVM, RF and ANN in Potato Blight Disease Detection Using Leaf Images

Abstract

In agriculture, timely disease control and management plays an important role in optimizing yield, improving harvest quality and minimizing loss to farmers. Automated disease management tools turn out to be of significant aid to farmers. In our paper, we present efficient automated disease management techniques in potato. Potato is world's fourth largest food crop, cultivated in many parts of the globe. Potato crops are majorly affected by fungus infections namely early blight and late blight in different stages of growth. These appear as brownish spots and black lesions on the leaves. The images of potato leaves are captured and analyzed to detect disease symptoms and further classified as diseased or normal. Here, we use image processing and machine learning techniques. The comparison of performance of classifiers support vector machine (SVM), random forest (RF) and artificial neural network (ANN) is performed on the same test dataset of potato leaves. From results, we observe ANN scores highest of 92 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> accuracy followed by SVM with 84 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> accuracy and RF with 79% accuracy.