Detection of expired vacuum-packed smoked salmon based on PLS-DA method using hyperspectral images

Consumers want fresh food with a long shelf-life, which in 2010, resulted in an important increase in packaged and processed food. This is especially important for fishery products due to their highly perishable nature. One problem is that it is not possible to measure freshness in packaged food only using the visible spectrum. Moreover, the detection of freshness is a complex problem as fish has different tissues with different biodegradation processes. Therefore, it would be especially interesting to have a non-destructive method to evaluate the shelf-life of packed processed fish. This paper proposes a method for detecting expired packaged salmon. Firstly, this method uses hyperspectral imaging spectroscopy (HIS) using visible and SW-NIR wavelengths. Secondly, a classification of different salmon tissues is carried out by image segmentation. Finally, classifications of expired or non expired salmon are performed with the PLS-DA statistical multivariate method due to the large amount of captured data. In a similar way, spectral data and the physicochemical, biochemical and microbiological properties of salmon are correlated using partial least squares (PLSs). The result obtained has a classification success rate of 82.7% in cross-validation from real commercial samples of salmon. Therefore, this is a promising technique for the non-destructive detection of expired packaged smoked salmon.