Enhanced Rice Variety Identification Using Hyperspectral Imaging and Deep Learning with Multi-Feature Fusion

Original Title: Hyperspectral imaging for accurate determination of rice variety using a deep learning network with multi-feature fusion

Weng et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy
Volume 234, 15 June 2020, 118237 https://doi.org/10.1016/j.saa.2020.118237

Rice adulteration is a significant issue affecting producers, traders, and consumers. To address this, hyperspectral imaging (HSI) and deep learning networks were used to accurately identify rice varieties.

HSI images of 10 rice varieties in China were analyzed, extracting spectroscopic, morphological, and textural features. Classification was performed using the PCANet network, compared with methods such as K-nearest neighbour (KNN) and random forest (RF).

Advanced data processing techniques, including normalization or principal component analysis (PCA), were applied to reduce interference and enhance accuracy. PCANet achieved the best results, with classification rates of 98.66% in training and 98.57% in prediction.

This method enables precise rice identification and can be extended to other agricultural products.

Would you like to know more? Check it out Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, Elsevier