Mapping of TBARS distribution in frozen–thawed pork using NIR hyperspectral imaging

• HSI was applied to predict TBARS value in frozen–thawed pork.
• 13 effective wavelengths were selected by SPA algorithm.
• SPA–PLSR model was built and achieved a good result.
• Distribution of TBARS was visualized among different frozen–thawed cycles.

In this study, NIR hyperspectral imaging technology was applied to determine the distribution of TBARS in frozen–thawed pork. A total of 240 fresh pork samples were assigned to 4 treatment groups (0, 1, 3, 5 frozen–thawed cycles). For each sample, a hyperspectral image (874–1734 nm) was collected, followed by chemical TBARS analysis. Successive projection algorithm (SPA) was applied to choose effective wavelengths (EWs). The selected 13 EWs of the calibration set and relevant TBARS value were used as inputs of partial least squares regression (PLSR) model, yielding correlation coefficient of prediction of 0.81 and root mean square error of prediction of 0.33. The developed PLSR model were applied pixel-wise to produce chemical maps of TBARS for 24 selected samples in the prediction set. The results indicated that NIR hyperspectral imaging combined with image processing has the potential to visualize TBARS distribution in frozen–thawed pork. This technique could be useful in real-time quality monitoring in meat industry.