Metaheuristic optimization based improved neural network for the timely prediction of paddy leaf diseases.

Rice is a fundamental dietary staple that embodies cultural identity, culinary diversity, food security and economic stability, providing a significant portion of daily caloric intake for considerable portion of world populace. However, vulnerability of rice crops to various leaf diseases poses a considerable concern to its productivity and quality, which in turnemphasizes the need for an effective disease prediction technique. This research focuses on prompt detection of pathogenic threats to paddy leaves through image analysis. The inputs are pre-processed using an Adaptive Gabor Filter (AGF), for improving the quality and reducing the noise from the image. Subsequently, using a Histogram of Oriented Gradient (HOG), the extraction of relevant disease-related information is assured. For the classification stage, an OptimizedCapsule Networks (CapsNet) is employed for identifying diseases affecting paddy foliage with high accuracy. The proposed CapsNet achieves improved generalization and robustness by integrating the Glow Worm Swarm Algorithm (GWSA) for parameter optimization. In addition to disease prediction, the proposed system integrates a Fertilizer-Based Disease Management (FBDM) approach, which suggests precise fertilization strategies based on the identified disease. By analysing disease symptoms, the system recommends optimal nutrient applicationssuited to enhance plant resistance and recovery. This integration of disease prediction with precision agriculture techniques enables site-specific fertilizer application, preventing overuse and reducing environmental impact.