Prediction of ice fraction and fat content in super-chilled salmon by non-contact interactance near infrared imaging

It has been shown that super-chilling can increase the shelf life of fish and meat products without the reduction in quality encountered with normal freezing. The shelf Life and quality achieved depend on the fraction of water frozen and how it is preserved in the product. In this paper, we report on the possibility of using non-contact near infrared (NIR) interactance spectroscopy in combination with multi-spectral imaging as a rapid, non-destructive way to predict the average ice fraction as well as the spatial distribution of ice in super-chilled salmon fillets. For average ice fraction calibrations, samples in the range of 0% to 30% ice were investigated. Partial Least square regression (PLSR) models with high explained variance (R-2 approximate to 0.98) and root mean square error of cross-validation between 1.6% and 2.4% were obtained. The time of NIR measurement after super-chilling (Oh, 1/2h, 2h and 24h) did not seem to affect the prediction ability significantly. The spatial ice distribution, as predicted by PLSR modelling, was affected considerably by the fat distribution. This was avoided by the use of multivariate curve resolution at pixel level and calibration against the amount of ice divided by the total amount of water in both Liquid and solid state. Excellent images of ice distribution were achieved and allowed observation of changes during storage and thawing. The ability to predict fat content in super-chilled salmon fillets was also investigated. Fat content predictions were affected by ice fraction, but were observed to be independent of time of measurement after freezing.