A Novel Smart Vision Hybrid Classifier for Mango Leaf Disease Precision Agriculture Revolution

The prevention of economic loss and assurance of crop quality in agriculture heavily relies on faster, more timely, and accurate identification of leaf diseases that affect mango trees. In this study, a new method of automatic detection and classification of mango tree diseases that combines image processing methods with a sophisticated classification approach is proposed. The proposed method comprises several image preprocessing methods that include adaptive contrast enhancement, modified color space image, and multi-level noise elimination, to extract discriminable features from the leaf images. The feature extraction methodology utilized a hybrid texture–color descriptor from a deep learning approach. An entirely new lightweight but fast and powerful classification method–Hybrid Attention-Optimized Deep Ensemble Classifier (HAODEC)–that is made up of convolutional neural networks (CNNs) is proposed. The model was trained and validated on a customized dataset of infected mango leaves and healthy mango leaf images and achieved a promising classification success of $98.96 \%$, which was far better than previously used comparative models. The research habits show great promise for real-time identification of mango diseases, with a strong promise to develop automated implementations for plant health monitoring strategies and smart farming actions in real-time.