Neural Network-based Study for Rice Leaf Disease Recognition and Classification: A Comparative Analysis Between Feature-based Model and Direct Imaging Model
Rice leaf diseases cause substantial yield losses, necessitating early and accurate identification. This study systematically compares two paradigms: Feature Analysis–based Detection Models (FADM) and Direct Image-Centered Detection Models (DICDM). FADM integrates multi-scale feature extraction, dimensionality reduction (e.g., PCA), feature selection (e.g., mRMR), and Extreme Learning Machine (ELM) classification, evaluated via 10-fold cross-validation; DICDM adopts an end-to-end image-input approach. Experimental results demonstrate that FADM achieves significantly higher classification accuracy and computational efficiency than DICDM across multiple rice disease classes. These findings validate the effectiveness of the “feature-driven” paradigm in resource-constrained agricultural settings. The work contributes a lightweight, interpretable, and empirically grounded framework for intelligent crop disease diagnosis, offering both methodological innovation and practical guidance for deploying AI in low-infrastructure farming environments.