Enhancing Tomato Fruit Disease Detection using Dung Beetle Optimization with Deep Transfer Learning based Feature Fusion Model

Abstract

Tomato fruit, scientifically known as Solanum lycopersicum, is economically important and the most widely consumed fruit worldwide. Tomato fruit disease encompasses a range of ailments that may damage the yield, marketability, and quality of tomatoes. Several pathogens, such as viruses, fungi, bacteria, and environmental conditions cause this. Plant disease is most major aspect restricting the sustainable process of agriculture and has often been a challenging conundrum in agricultural production. Current developments in technology, including the usage of Deep Learning (DL) methods for disease detection, provide promising avenues for timely intervention and target management of Tomato Fruit Disease. These systems can precisely diagnose diseases through analyzing images of diseased plants, enabling growers to perform early detection and diminish crop loss. In this study, a new Tomato Fruit Disease Detection using Dung Beetle Optimization with Deep Feature Fusion (TFDD-DBODFF) model is introduced. The TFDD-DBODFF model aims to improve tomato fruit disease detection. Primarily, the CLAHE-based preprocessing is performed. Besides, the TFDD-DBODFF technique follows a deep feature fusion process containing 3 DL approaches such as Residual Network (ResNet), Capsule Network (CapsNet), and SqueezeNet. In addition, the hyperparameter tuning of these DL models can be performed using the DBO technique. Finally, the classification and detection of tomato fruit diseases take place using a Stacked Autoencoder (SAE). The investigational outcome of the TFDD-DBODFF approach can be examined using a benchmark dataset. The experimental validation indicates the optimum performance of the TFDD-DBODFF approach with existing methods under diverse measures.