Deep learning and Vegetation indices based approach for leaf diseases classification in RGB images

Abstract

Foliar diseases are one of the factors that impact corn crop productivity. Automatic detection of corn diseases can play an important role in addressing the issue of diseases management and productivity enhancement. In this regard, an experimental study was conducted on the corn set of the Plantvillage dataset following a flowchart that consists of three main steps: preprocessing, feature extraction and classification. Vegetation indices (ExG, ExR, ExGR, GRVI, NDI, RGI, CIVE) and color spaces (HSV, YUV, LAB) were used for preprocessing. VGG16 was used for feature extraction through transfer learning and three different algorithms were tested for classification: K-Nearest Neighbour (KNN), Support Vector Machine (SVM) and Random Forest (RF). Results reached an accuracy of 92.39% with RGB color space and the SVM classifier.