Spectroscopic control of fish products: simultaneous recognition of species and thawed status in large-scale distribution.

In the context of globalized food production, traceability is a key requirement in the fisheries sector, particularly for prepared and packaged fish products, where processing may hinder visual inspection and increase the risk of mislabeling. Reliable analytical tools are essential, therefore, to discriminate product status (fresh or frozen-thawed) and to ensure accurate species identification, which is essential for food safety and consumer protection. This study investigates near-infrared spectroscopy as a rapid, non-destructive approach to support the traceability of packaged fish products in large-scale retail chains. An innovative in-field methodology was developed, enabling the acquisition of spectroscopic fingerprints directly at the point of sale using a handheld instrument operating in contact with the transparent PVC film. A total of 218 samples were analyzed across 40 different retail points. Spectral data were acquired both with and without packaging to generate two datasets. Chemometric and machine learning approaches were applied to process spectral data and develop classification models. The best models achieved satisfactory classification accuracies ranging from 0.92 to 0.69 and showed strong agreement between predictions obtained from spectra acquired with and without PVC film (Cohen's κ=0.83 for fresh vs. frozen-thawed classification and κ=0.77 for species identification). These results demonstrate the ability of portable near infrared spectroscopy to identify both fish species and physical status, supporting self-monitoring and official controls by reducing response times and product handling, with potential integration into blockchain-based traceability systems.