Non-destructive freshness assessment of mackerel (Scomber japonicus) using colorimetric analysis and machine learning-based prediction models.

Freshness is a critical attribute of seafood quality. However, conventional assessment methods are time-consuming and destructive. This study investigated a non-destructive approach using colorimetric analysis of the eye, belly, and dorsal regions of mackerel (Scomber japonicus), and correlated these changes with microbiological and physicochemical freshness indicators. RGB, HSV, and Lab color parameters showed progressive darkening during storage at 4 °C and 10 °C, corresponding with increases in microbial load, pH, and total volatile basic nitrogen. Multivariate linear regression (MLR), partial least squares regression, and support vector regression (SVR) models were developed to predict freshness based on color data. While MLR performed well for linear indicators including viable cell count and quality index method, SVR provided superior prediction for non-linear indicators including pH and total coliforms. These findings demonstrate the potential of integrating color analysis with machine learning to enable real-time, non-destructive seafood freshness evaluation, supporting its applicability in industrial quality control systems.