A Spider Wasp Optimizer-Based Deep Learning Framework for Efficient Citrus Disease Detection

Managing citrus diseases is important for lowering crop losses and raising the economic value of citrus output. To provide a novel approach for the identification and classification of three significant citrus diseases—Citrus Canker, Citrus Greening, and Citrus Black Spot—this study uses a Deep Convolutional Neural Network (DCNN) optimized using the Spider Wasp Optimizer (SWO). Traditional disease diagnosis methods heavily rely on expert visual inspection, which is often subjective and time-consuming. To overcome these drawbacks, the proposed SWO-DCNN model automates hyperparameter tuning, improving classification accuracy and reducing computation time. Citrus image datasets containing both healthy and infected samples were pre-processed using grayscale conversion, normalization, and augmentation, and then trained using a 10-fold cross-validation technique. Performance evaluations based on sensitivity, specificity, false positive rate, accuracy, and identification time show that the SWO-DCNN outperforms the conventional DCNN in every disease category. With accuracies of 96.22%, 96.51%, 95.70%, and 97.04% for the classification of Black Spot, Greening, Canker, and overall healthy/non-healthy, respectively, the SWO-DCNN significantly reduced false positive rates and recognition times. This paper contributes to knowledge by presenting the Spider Wasp Optimizer, a hyperparameter tuning technique for deep learning models used in the identification of agricultural diseases. The SWO-DCNN framework offers a dependable and scalable approach for automated citrus disease classification by enhancing model performance and computational efficiency. This innovation supports precision farming initiatives and provides a reliable alternative to traditional diagnostic methods, which may improve export quality control and reduce citrus farming's financial losses.