Identification of Hybrid Indica Paddy Rice Grain Varieties Based on Hyperspectral Imaging and Deep Learning.

Paddy rice grain variety classification is essential for quality control, as different rice varieties exhibit significant variations in quality attributes, affecting both food security and market value. The integration of hyperspectral imaging with machine learning presents a promising approach for precise classification, though challenges remain in managing the high dimensionality and variability of spectral data, along with the need for model interpretability. To address these challenges, this study employs a CNN-Transformer model that incorporates Standard Normal Variate (SNV) preprocessing, Competitive Adaptive Reweighted Sampling (CARS) for feature wavelength selection, and interpretability analysis to optimize the classification of hybrid indica paddy rice grain varieties. The results show that the CNN-Transformer model outperforms baseline models, achieving an accuracy of 95.33% and an F1-score of 95.40%. Interpretability analysis reveals that the model's ability to learn from key wavelength features is significantly stronger than that of the comparison models. The key spectral bands identified for hybrid indica paddy rice grain variety classification lie within the 400-440 nm, 580-700 nm, and 880-960 nm ranges. This study demonstrates the potential of hyperspectral imaging combined with machine learning to advance rice variety classification, providing a powerful and interpretable tool for automated rice quality control in agricultural practices.