Detection and monitoring for enhanced prevention of grain plant disease using classification-based deep ensemble neural networks in smart agriculture

Plant diseases lead to productivity loss, although they may be controlled with ongoing observation. Examiners have struggled long with the challenge of accurately detecting plant diseases in grain crops, such as wheat, rice, maize, millet, and ragi. Traditional manual monitoring methods can be time-consuming and prone to errors. To address this issue, this study explores the application of deep learning techniques for automated and accurate disease identification. The proposed method begins with image pre-processing using a Hybrid Gaussian-Wiener filter to reduce noise in leaf images. Deep learning models are then implemented the Moore-Penrose pseudo-inverse Weighted Deep Ensemble Neural Networks (DENN) for common bacterial and fungal diseases, and the Squeeze-and-Excitation Vision Transformer (SEViT) for diseases caused by specific fungi. To improve efficiency and accuracy, transfer learning is employed. The resulting system achieves an impressive classification accuracy of 99.56%. This research demonstrates the potential of deep learning for plant disease detection, which could significantly benefit farmers and plant pathologists by enabling early disease identification and intervention.