This work explores the use of hyperspectral imaging (HSI) technology for non-destructive monitoring of fish quality and its combination with mathematical models to predict its evolution. As a proof of concept, we focus on farmed turbot (whole) and wild-caught hake (filleted). Quality is evaluated in terms of four different indicators: bacterial growth, total volatile basic nitrogen (TVB-N), nucleotide catabolism, and sensory assessment. Data obtained from several experiments, carried out at different storage temperatures, were used to: (i) find models that correlate such data with HSI information in order to derive a methodology for a fast, real-time, and non-invasive evaluation of fish quality; (ii) calibrate the mechanistic model describing the evolution of bacterial growth, TVB-N formation, and nucleotide degradation, and (iii) train the machine learning models that will describe sensory evolution.
Muñoz et al. (Wed,) studied this question.