Rapid Walnut Impurity Detection Using Near-Infrared Hyperspectral Imaging

Wang, D., Zheng, J., Li, L., Liang, Z., Zheng, Y., Huang, S., Zheng, X., Zhou, Z., & Dai, D. (2024). Rapid detection of endogenous impurities in walnuts using near-infrared hyperspectral imaging technology. Journal of Food Composition and Analysis, 132, Article 106290. https://doi.org/10.1016/j.jfca.2024.106290

Ensuring walnut purity is crucial for food safety and market quality standards. Traditional impurity detection methods are slow and inefficient, highlighting the need for rapid and reliable alternatives. This study introduces near-infrared hyperspectral imaging (NIR-HSI) combined with advanced machine learning techniques for the rapid detection of internal walnut impurities.

Hyperspectral imaging was employed in the near-infrared spectral range to collect data from walnut samples containing endogenous impurities. Preprocessing and wavelength selection techniques were applied to enhance the quality of spectral data. Advanced algorithms such as Support Vector Machines optimized by Butterfly Optimization Algorithm (BOA-SVM) and a novel deep learning network, WT-NIRSNet, incorporating wavelet transform and attention mechanisms, were developed to classify impurities.

The developed models demonstrated remarkable accuracy, achieving a maximum classification accuracy of 99.03%. The WT-NIRSNet method notably outperformed traditional models, proving hyperspectral imaging’s significant potential as an accurate, fast, and non-destructive method for walnut impurity detection.