Abstract:When the freshprepared steaks exceed the shelf life, its putrid odour is easily masked by the seasoning odour, making it difficult for consumers to distinguish. Total volatile basic nitrogen (TVB-N) is an effective indicator of meat freshness. Due to the chemical method of TVB-N content determination was cumbersome and time-consuming, TVB-N content in freshlyprepared steaks were predicted by hyperspectral imaging technique in this study. Furthermore, the influence of different wavelengths selection algorithms on the prediction effect of the model was discussed. Fresh prepared steaks were taken out on day 0, 2, 4, 6, 8 respectively for hyperspectral data collection and TVB-N content determination. The spectral data were pretreated by six methods, first derivative (1st Der), second derivative (2nd Der), mean centering (MC), multiplicative scatter correction (MSC), Savitzky-Golay (SG), standard normal variate transformation (SNVT), which established the Partial least squares model of TVB-N content and obtained the optimum pretreatment method by comparison. After the optimum pre-processing method, the characteristic wavelengths were selected by competitive adaptive reweighted sampling (CARS), variables combination population analysis(VCPA),interval random frog (iRF), iRF-CARS, iRF-VCPA, the prediction effects of models were evaluated on different selection wavelengths methods. CARS and VCPA were run 50 times repeatedly to assess their stability, and higher frequency wavelengths were selected to establish the PLS model to compare with a single run. The results suggested that 1st Der was the best spectral pretreatment method among the six methods. 21 and 11 wavelengths were selected by CARS and VCPA single run respectively, and the prediction model by VCPA wavelengths selection method had better effect, with the RC and RP at 0.944 and 0.931 respectively, while RMSECV and RMSEP at 1.12 and 1.28 mg·(100 g)-1 respectively. The selected frequency for each wavelength when CARS and VCPA were run 50 times repeatedly indicated that VCPA had better stability because its binary matrix sampling method provided the same sampling opportunity for each variable. At the same time, it was found that the two methods have important common wavelengths: 694.9, 696.6, 761.8, 763.5, 811.5 and 813.3 nm, etc. The performance of the model established byhigher frequency wavelengths was poor.IRF was combined with CARS and VCPA respectively for wavelengths selection, and iRF-CARS showed good predictive performance. The prediction model by iRF-CARS was established using 24 wavelengths, with the RC and RP at 0.966 and 0.938 respectively, while RMSECV and RMSEP at 0.91 and 1.22 mg·(100 g)-1 respectively. The results suggested that the combination of wavelengths interval selection and wavelengths point selection couldrealize their complementary advantages. The research suggested that hyperspectral imaging technique combined with wavelengths selection method can predict TVB-N content in prepared steaks well,which provides a theoretical reference for hybrid variable selection strategy and rapid detection of the freshness of prepared steaks.
Key words:Hyperspectral imaging; Freshprepared steaks; Total volatile basic nitrogen; Competitive adaptive reweighted sampling; Variables combination population analysis
[1] Velásquez L, Cruz-Tirado J P, Siche R, et al. Meat Science, 2017, 133: 43.
[2] Zheng X C, Li Y Y, Wei W S, et al. Meat Science, 2019, 149: 55.
[3] XIE An-guo, KANG Huai-bin, WANG Fei-xiang, et al(谢安国,康怀彬,王飞翔,等). Food & Machinery(食品与机械), 2018, 34(11): 20.
[4] Yun Y H, Li H D, Deng B C, et al. TrAC Trends in Analytical Chemistry, 2019, 113: 102.
[5] Guo Z M, Wang M M, Wu J Z, et al. Food Chemistry, 2019, 286: 282.
[6] Yu H D, Yun Y H, Zhang W M, et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2020, 224: 117376.
[7] Jiang H, Xu W D, Ding Y H, et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2020, 228: 117781.
[8] Guo Z M, Wang M M, Akwasi A A, et al. Journal of Food Engineering, 2020, 279: 109955.
[9] Ouyang Q, Yang Y C, Wu J Z, et al. LWT, 2020, 118: 108768.
[10] Yun Y H, Li H D, Wood L R E, et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2013, 111: 31.
[11] PAN Xiao-qian, ZHAO Yan, ZHANG Shun-liang, et al(潘晓倩,赵 燕,张顺亮,等). Meat Research(肉类研究), 2016, 30(3): 15.
[12] Van Beers R, Kokawa M, Aernouts B, et al. Meat Science, 2018, 136: 50.
[13] HUANG Chang-ping, ZHU Xin-ran, ZHANG Chen-lu, et al(黄长平,朱欣然,张辰璐,等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2018, 38(2): 552.
[14] Talens P, Mora L, Morsy N, et al. Journal of Food Engineering, 2013, 117(3): 272.
[15] Dai Q, Cheng J H, Sun D W, et al. Food Chemistry, 2016, 197: 257.
[16] YANG Bin, CAO Yin-juan, YU Qun-li, et al(杨 斌,曹银娟,余群力,等). Food Science(食品科学), 2019, 40(23): 199.