plSSN : 0374-8111 | elSSN : 2287-8815
plSSN : 0374-8111 | elSSN : 2287-8815
plSSN : 0374-8111elSSN : 2287-8815
박슬기1·조두민2·심연주3·김영목3,4·박기재1,5·이규석1*·조정석1,5*
1한국식품연구원 스마트제조사업단, 2국립해양생물자원관, 3부경대학교 식품공학과, 4부경대학교 식품과학부, 5한국식품연구원 안전유통연구단
We aimed to validate the efficacy of a freshness classification model developed through continuous acquisition of hyperspectral data from 500 fresh mackerel Scomber japonicus, speciments. Samples were procured immediately following an auction from the Busan Cooperative Fish Market and subjected to refrigerated storage for 72 h. A custom- designed continuous spectral data acquisition device collected 256 spectral bands within the wavelength range of 900-1,700 nm. Spectral noise reduction was achieved using standard normal variate (SNV) and Savitzky-Golay (SG) filters. Progressive spoilage was confirmed through total volatile basic-nitrogen (TVB-N) analysis, with values increasing from 14.2 to 35.8 mg/100 g over ther 72 h storage duration. Principal components analysis (PCA) revealed spectral pattern variations throughout the storage period, accounting for 96.68% of the total expressed variance. The hyperspectral mean reflectance spectra exhibited primary absorption bands at 1,100, 1,200, and 1,300 nm. A freshness classification model employing partial least squares discriminant analysis (PLS-DA) exhibited robust predictive performance, attaining a maximum accuracy of 93.01%. The devised system demonstrated efficacy in continuous spectral data acquisition for real-time mackerel freshness classification. This device provides foundational data for the future advancement of nondestructive analysis techniques through the refinement of classification models via machine learning and algorithm development.
Mackerel, Hyperspectral imaging, Chemometrics, Predictive modeling