Abstract
This study highlights the potential of Fourier-transform near-infrared (FT-NIR) spectroscopy for the on-site, nondestructive detection of spoilage caused by bacterial action in raw salmon (Salmo salar) fillets. A stepwise multiple linear regression model with first-derivative spectrum transformation was combined with the standard normal variate and detrend preprocessing techniques. The model achieved correlation values of 0.97 in both the calibration and validation sample sets, with root mean square error values of 0.18 and 0.20 log CFU/mL, respectively. These accurate results reveal the precision of FT-NIR spectroscopy for assessing the spoilage caused by bacteria. The most informative wavelengths (885.27 nm, 1026.27 nm, 1039.93 nm, 1068.38 nm, 1257.55 nm, 1267.75 nm, and 1453.49 nm) related to the total bacterial count's identification were obtained. The innovative, cost-effective, and feasible approach outlined in this article is a promising methodology for enhancing the safety and quality standards of various fishery products.