Prediction and visualization of fat content in polythene-packed meat using near-infrared hyperspectral imaging and chemometrics

Abstract

There is a growing demand for non-destructive techniques in the meat industry. This study aimed to use hyperspectral imaging to predict intramuscular fat (IMF) in pork vacuum-packed with polyethylene (WPE) film. Data pre-processing was applied to the WPE data by using second-order derivative (SD) followed by correlated optimized warping (COW) in order to align and remove the contributions of the PE film from the spectral data, respectively. The IMF values were correlated with the spectral data using support vector machines regression (SVMR) or partial least squares regression (PLSR) to develop the calibration models. Results based on SVMR applied to the WPE showed high accuracy (R2 = 0.89, RMSEP = 0.37%, RPD = 3.38) in the prediction. The results indicated that the SD and COW were effective in pre-processing hyperspectral images of PE-packed pork for IMF prediction. Overall, the study reveals that hyperspectral imaging could be used to accurately evaluate IMF content in vacuum-packed pork.