Non-destructive prediction and visualization of chemical composition in lamb meat using NIR hyperspectral imaging and multivariate regression

The main goal of this study was to investigate the potential of hyperspectral imaging in the near-infrared (NIR) range of 900-1700 nm for non-destructive prediction of chemical composition in lamb meat. Hyperspectral images were acquired for lamb samples originated from different breeds and different muscles. The mean spectra of the samples were extracted from the hyperspectral images and multivariate calibration models were built by using partial least squares (PLS) regression for predicting water, fat and protein contents. The models had good prediction abilities for these chemical constituents with determination coefficient (R²_p) of 0.88, 0.88 and 0.63 with standard error of prediction (SEP) of 0.51%, 0.40% and 0.34%, respectively. The feature wavelengths were identified using regression coefficients resulting from the PLSR analyses. New PLSR models were again created using the feature wavelengths and finally chemical images were derived by applying the respective regression equations on the spectral image in a pixel-wise manner. The resulting prediction maps provided detailed information on compositional gradient in the tested muscles. The results obtained from this study clearly revealed that NIR hyperspectral imaging in tandem with PLSR modeling can be used for the non-destructive prediction of chemical compositions in lamb meat. Industrial relevance: The results obtained from this study clearly revealed that NIR hyperspectral imaging in tandem with PLSR modeling can be used for the non-destructive prediction of chemical compositions in lamb meat for the meat industry.