Implementation of Optimal Leaf Feature Selection-Based Plant Leaf Disease Classification Framework with RNN+GRU Technique

Abstract

Agricultural productivity is the most important factor for supporting the food sources in the populated economy. Therefore, it is highly necessary to solve the disease detection problems that affect the plant agricultural production since it occurs in the natural way. The early disease identification is required for increasing the crop yield in the agricultural sector. The leaf diseases like yellow curved, septoria leaf spot, late blight and bacterial spot are mostly reduced the quality of the crop. Automated techniques have to be developed for classifying the plant diseases and also it supports to take preventive measures to identifying the leaf disease symptoms. Hence, this chapter aims to design the plant leaf disease identification model with efficient deep learning architecture integrated to leaf features to provide the accurate detection results. The input images are collected and utilized for primary stage of processing using the CLAHE technique. Then, the pre-processed images are given into the Leaf segmentation phase, where the appropriate leaf regions are segmented to get accurate classification results. The segmented images are used in the feature extraction phase. Further, the extracted features are considered for choosing the optimal features for classification phase through Fisher Discriminant Analysis (FDA) technique. The optimal features are given into classification phase, in which Recurrent Neural Network (RNN) and Gated Recurrent Units (GRU) are used for performing the leaf disease classification. Experimental analysis reveals that the proposed approach attains better classification performance when considering the analysis with various performance measures.