Abstract: The quality change of pork during storage was evaluated by measuring the changes in total volatile basic nitrogen (TVB-N) content and total viable count (TVC). The results showed that TVB-N content of pork refrigerated for up to 15 days at 4 ℃ exhibited a J-shaped curve, while TVC exhibited an S-shaped curve. On day 15, TVB-N content reached the maximum limit for freshness stipulated in the national standard (15 mg/100 g), while TVC reached 7.92 (lg(CFU/g)), far higher than the maximum limit stipulated in the national standard, 6 (lg(CFU/g)). While TVC reached 6 (lg(CFU/g)) on day 5.5, TVB-N content was only 10.65 mg/100 g. Pork meat judged to be spoiled meat based on TVC was found to be still fresh according to TVB-N content. Furthermore, visible short-wave near-infrared hyperspectral reflection technology was used to collect hyperspectral data for pork samples, and partial least squares regression (PLSR) models for the prediction of TVB-N content and TVC were established using different spectral pretreatments. The results showed that the model established using multiplicative scatter correction (MSC) preprocessing displayed the best prediction performance for TVB-N content, while the model established using first derivative preprocessing exhibited the best prediction performance for TVC. The correlation coefficients of prediction set (Rp) were 0.957 2 and 0.968 2, the standard errors of prediction (SEP) were 2.802 5 mg/100 g and 0.332 7 (lg(CFU/g)), the relative percent deviation (RPD) was 3.093 7 and 3.434 1, the correlation coefficients of external validation set were 0.928 3 and 0.930 5, and the standard errors were 3.556 2 mg/100 g and 0.515 7 (lg(CFU/g)) for the TVB-N and TVC prediction models, respectively. In conclusion, this study can provide a theoretical basis for better application of hyperspectral technology in pork quality detection.

Key words: freshness; degree of spoilage; hyperspectral imaging; total volatile basic nitrogen; total viable count; partial least squares regression