DEEP TRANSFER LEARNING WITH OPTIMAL KERNEL EXTREME LEARNING MACHINE MODEL FOR PLANT DISEASE DIAGNOSIS AND CLASSIFICATION

Presently, automated plant disease identification and classification is a major research area, which has received popularity in the farming sectors. Conventional disease diagnosis models are mainly based on the handcrafted features obtained from the captured images to determine the infection type. In addition, the outcome of the models is mainly based on the characteristics of chosen handcrafted features. This issue can be resolved by learning the features automatically by the use of deep learning (DL) models. This paper presents a new Deep Transfer Learning with Optimal Kernel Extreme Learning Machine (DTL-OKELM) model for Plant Disease Diagnosis. The presented DTL-OKELM model involves Kapur's thresholding based image segmentation technique with grey wolf optimization (GWO) algorithm to detect the infected regions of the plant image. Besides, DTL based Inception v3 technique is employed as a feature extractor for the derivation of optimal features. The DTLOKELM model also makes use of KELM as a classification model in which the parameter tuning of KELM takes place using the GWO algorithm. The application of the GWO algorithm for Kapur's thresholding and hyperparameter tuning of KELM results in effective classification performance. The simulation analysis of the DTLOKELM model takes place using two datasets namely the citrus dataset and tomato dataset. The resultant experimental values portrayed the effective outcome of the DTLOKELM model with the maximum accuracy of 98.58% and 99.32% on the applied citrus and tomato dataset respectively