A Smart Solution for Tomato Leaf Disease Classification by Modified Recurrent Neural Network with Severity Computation

Abstract

The artificial intelligence-assisted deep learning approach plays a significant task in identifying diseases through a large set of plant leaf images. Thus, the major aim of the designed model is to design and develop the tomato leaf disease classification model with intelligent approaches. Initially, the images are gathered from standard datasets and real time datasets. Then, the pre-processing is conducted for cleaning, shadow removal, and enhancing the images using the image enhancement approach. Then, the spot segmentation from the leaf is performed by the Optimized K-Means Clustering (OKMC) using Standard Deviation-based Grasshopper Horse Herd Optimization (SD-GHHO). Further, the deep feature extraction is extracted from the segmented spot using Convolutional Neural Network (CNN), VGG16, and Residual Networks (ResNet). The gathered features are further concatenated. Then, the optimal features are selected using a new meta-heuristic algorithm called SD-GHHO. Finally, the disease classification is performed with the help of the Modified Recurrent Neural Network (MRNN), where the weight of RNN is optimized using the same SD-GHHO technique. Once after classifying the spots, the disease severity computation is performed. The proposed method has enhanced classification efficiency in high accuracy, specificity, and sensitivity values.

Keywords:

• Convolutional neural network
• modified recurrent neural network
• optimized K-means clustering
• residual networks
• severity computation
• standard deviation-based grasshopper horse herd optimization
• tomato leaf disease classification
• VGG16

Acknowledgment

I acknowledge “KLEF –2012030001” Full Time fellowship awarded for the successful completion of the work.