<p>Maintaining food security and increasing crop yields depend on the accurate diagnosis of rice illnesses. Traditional disease diagnosis frequently uses manual inspection, however this method is typically laborious and prone to errors. The rapid advancement of machine learning and image processing technology offers a potential chance to automate and improve the precision of rice plant disease identification. This proposed work introduces an advanced Rice Disease Detection model incorporating a multi-step approach of image augmentation, preprocessing, feature extraction, and hybrid detection. The model enhances dataset diversity using image transformations such as rotation, scaling, cropping, and flipping. An Improved Wiener Filter is used for preprocessing in order to lower noise while maintaining the edges and details of the image. Feature extraction combines texture, color, shape, and an enhanced entropy-based method to capture a comprehensive range of image attributes. Detection is achieved through a hybrid model that integrates YOLO V7 for object detection with an Improved Long Short-Term Memory (ILSTM) network, which is fine-tuned using a novel hybrid optimization algorithm that merges Zebra and Pufferfish optimization techniques. This approach aims to enhance accuracy in rice disease detection by optimizing feature extraction and classification processes, addressing both computational efficiency and detection precision. Using the proposed strategy, the highest values recorded were accuracy (0.956), precision (0.957), and F-measure (0.955).</p>