A Fusion Method for Detection and Classification of Diseases in Tomato Plants Using Swarm-based Deep Learning

Abstract

Precise identification and detection of ailments in tomato plants are essential for preserving crop vitality and optimizing agricultural productivity. This promotes the use of agricultural methods that can be maintained over time and decreases financial losses caused by plant diseases. Detecting and classifying diseases in tomato plants is critical for ensuring crop health and maximizing agricultural productivity. Utilizing advanced computer vision techniques for this purpose enhances precision in monitoring plant health, ultimately leading to more efficient and targeted agricultural interventions. This research work presents a novel framework for Tomato Plant Disease Detection and Classification (TPDDC) using a fusion of swarm-based methods and deep-learning techniques. Our approach leverages K-means clustering with Grasshopper Optimization (GO) for segmenting Regions of Interest (ROI) from tomato leaf images, followed by feature extraction and optimization using Maximally Stable Extremal Regions (MSER) and GO. The optimized features are then classified using a Convolutional Neural Network (CNN). The proposed TPDDC model was evaluated using the Plant Village Dataset, encompassing ten different tomato leaf diseases. Experimental results demonstrate significant improvements in detection and classification accuracy, achieving an average accuracy of 97.6% with the GO-based approach compared to 92.7% without GO. These results underscore the effectiveness of integrating swarm-based optimization with deep learning for robust and precise disease detection in tomato plants.