Rapid, non-destructive, and precise salmon species identification is a pressing necessity in the food safety domain. This paper proposes an innovative method for salmon identification grounded in lipid characteristics. To monitor the disparities in lipid traits across various salmon species, Raman spectroscopy was employed for swift and non-invasive measurement of fatty acid composition. The findings revealed significant differences in the unsaturation levels and content of certain long-chain fatty acids among salmon varieties. C18:3n3, C16:0, C14:0, C16:1n7, C18:1n9c, and C18:2n6c were identified as the most crucial fatty acids for salmon differentiation. Raman spectroscopy effectively captured variations in fatty acid profiles, accurately predicting the content of C14:0, C16:1n7, and C18:2n6c, with determination coefficients (R2) of 0.96, 0.90, and 0.87, respectively. To predict salmon species, a novel ensemble model, PLSR_SVM_AdaBoost-LinearRegression (PSA-LR), was proposed. The synergy of Raman spectroscopy and PSA-LR facilitated precise identification of different salmon species, with a determination coefficient (R2) of 0.86, Root Mean Square Error of Prediction (RMSEP) of 0.37, and an accuracy of 100%. The results demonstrate that lipid characteristics serve as effective indicators for salmon species identification, and the combination of Raman spectroscopy with an ensemble model enables rapid, non-destructive, and accurate salmon species discrimination.
Read full abstract