Method comparison between real-time spectral and laboratory based measurements of moisture content and CIELAB color pattern during dehydration of beef slices

In this study, partial least square (PLS) regression models were developed to predict moisture content (MC) (model 1), CIELAB color (model 2) or all four parameters (model 3) of beef slices during drying. Model development was based on data from two measurement campaigns of MC (%), CIELAB L*, a* and b* values and hyperspectral data in the range of 500-1009 nm. To increase the robustness of the models, the beef samples varied dependent on cattle breed, cut and pre-treatment. With low-cost, non-invasive continuous monitoring systems in mind, the models were simplified by wavelengths selection. The Deming and Passing-Bablok regression and the Bland-Altman plot revealed high model performances. Mean differences (full/reduced model) of -0.64/-0.64 for MC, -0.14/-0.15 for CIELAB L*, 0.05/0.04 for a* and 0.08/0.06 for b* values were achieved for model 3, which shows the high potential for simple real-time monitoring applications combining all investigated factors and parameters.