Duck optimization with enhanced capsule network based citrus disease detection for sustainable crop management

Abstract

Sustainable agriculture methodology seeks to apply natural resources whereby they could reconstruct their productive capacity and minimise adverse effects on the ecosystem beyond a field edge. In agriculture, plant diseases are mainly responsible for production degradation, which causes economic loss. In plants, citrus is widely utilized as a critical resource of nutrients such as vitamin C. But Citrus disease negatively impacts the quality and production of citrus fruits. Over the last few years, image processing and computer vision technologies have been extensively applied to detect and classify diseases in plants. Mostly, plant disease has clear signs, and today’s recognized model is for an expert plant diagnostician to recognize the disease by observing a diseased plant leaf through a microscope. Manual disease detection can be labour-intensive and the efficiency of the diagnoses is associated with the pathologist's skills, which makes this a greater application field for computer-aided diagnosis systems. This work develops an effectual Duck Optimization with Enhanced Capsule Network Based Citrus Disease Detection for Sustainable Crop Management (DOECN-CDDCM) technique. The presented technique majorly focuses on the detection and classification of citrus diseases. In the presented approach, several stages of preprocessing are performed. In addition, the duck optimization algorithm (DOA) with the Enhanced Capsule Network (ECN) model is exploited as a feature extractor, and the optimal hyperparameter tuning process helps obtain enhanced performance. Finally, the deep stacked sparse autoencoder (DSSAE) model is utilized for citrus disease detection and classification. A series of simulations were performed to ensure the enhanced disease detection performance of the DOECN-CDDCM algorithm. The simulation results demonstrate the improvised outcomes of the DOECN-CDDCM algorithm than other current techniques.