Smart Citrus Farming: Deep Learning and Swarm Optimization for Leaf Disease Diagnosis
Аннотация
Citrus crops play a vital role in global agricultural economy, but their productivity is severely impacted by many leaf diseases such as canker, greening, besides scab. Citrus plant disease detection involves identifying infections such as citrus canker, greening (HLB), and black spot using advanced techniques like image processing, besides deep learning. Early detection helps prevent widespread crop damage, reduces the need for excessive pesticide use, and ensures better yield and fruit quality. Automated models using CNNs, besides spectral analysis, can efficiently classify diseased and healthy leaves, enabling timely intervention. Timely besides accurate proof of identity of these diseases is crucial for effective management and mitigation. In this study, a novel deep learning-based framework is proposed that integrates graph convolutional networks (GCNs) with perpetual pigeon galvanized optimization (PPGO) to enhance citrus leaf disease classification. GCNs are employed to exploit spatial besides contextual relationships in graph-structured representations of citrus leaf images, while PPGO optimizes hyperparameters and selects the most discriminative features to improve classification performance. Our architecture includes a preprocessing pipeline, graph construction, feature selection, and deep learning classification stages. The proposed system is evaluated on benchmark citrus leaf image datasets with four disease classes: Canker, Scab, Greening, and Healthy. Experimental results prove that the proposed GCN-PPGO model outperforms traditional classifiers like SVM, KNN, and Random Forest in score, and false positive rate. Specifically, integration of PPGO significantly enhances model convergence and detection precision. This framework is robust under varied environmental conditions, making it suitable for field deployment and mobile-based applications for farmers. Our results suggest that combining graph-based learning with bioinspired optimization can serve as an effective strategy for real-time agricultural disease management. This model has the potential to be extended to other crop types, contributing to the broader goal of intelligent precision farming.