Classification of Wheat Disease Using Model-Level Graph Network with Parrot Optimizer

One of the most vital staple crops in the world, wheat plays a major role in both economic stability and human sustenance. However, a number of diseases pose a threat to wheat productivity, which can severely reduce yield and quality. Existing detection methods face limitations such as imbalanced datasets, noise in images, and high computational requirements, restricting timely and accurate disease identification. To overcome these challenges, the research presents a Model-Level Graph Network with Parrot Optimizer (MoLGNNet-ParO) for wheat disease classification. The approach begins with data collection from the wheat leaf disease dataset, ensuring representation of both healthy and diseased samples. Collected images undergo preprocessing using a Modified Square Root Sage-Husa Adaptive Kalman Filter (MSSHAKF) to reduce noise and enhance image quality. Feature extraction is performed using a Modified ResNet-152 with Hinge Attention Network (MR-152-HiAN), capturing discriminative and hierarchical disease features. Then, the Model-Level Graph Network (MoLGNNet) leverages structural relationships between features to classify wheat leaf diseases accurately, while the Parrot Optimizer (ParO) optimizes the loss function, improving convergence and overall model performance. The experiment yields the better results, which are out of the F1-score (99.79%), precision (99.62%), accuracy (99.84%), specificity (99.57%), and recall (99.45%); the proposal framework performs better than the traditional models. The suggested methodology offers an effective, highly accurate and efficient method of detecting early wheat diseases, promoting sustainable crop yield, and efficient management of the disease.