Frontiers in Sustainable Food Systems, Год журнала: 2024, Номер 8

Опубликована: Сен. 12, 2024

Millet is one of the major coarse grain crops in China. Its geographical origin and Fusarium fungal contamination with ergosterol deoxynivalenol have a direct impact on food quality, so rapid prediction origins toxin essential for protecting market fairness consumer rights. In this study, 600 millet samples were collected from twelve production areas China, traditional algorithms such as random forest (RF) support vector machine (SVM) selected to compare deep learning models content. This paper firstly develops model (wavelet transformation-attention mechanism long short-term memory, WT-ALSTM) by combining hyperspectral imaging achieve best effect, wavelet transformation algorithm effectively eliminates noise spectral data, while attention module improves interpretability selecting feature bands. The integrated (WT-ALSTM) based bands achieves optimal origin, its accuracy exceeding 99% both training datasets. Meanwhile, it content, coefficient determination values 0.95 residual predictive deviation reaching 3.58 3.38 respectively, demonstrating excellent performance. above results suggest that combination has great potential quality assessment millet. study provides new technical references developing portable inspection technology on-site agricultural product future.

Язык: Английский