摘要： 针对目前模型传递方法研究大多在不同仪器之间且均采用近红外光谱建立模型，采用高光谱技术建立猪肉pH值定量检测模型，并针对不同品种间的模型传递提出了一种光谱和预测值同步校正(sync correction of spectrum and prediction value, CSPV)的传递算法，并与模型更新方法进行比较。当模型满足预测相关系数(correlation coefficient of prediction, rp)rp≥0.837，且剩余预测偏差(residual prediction deviation, RPD)RPD≥1.9时，表明预测结果可靠。以杜长大、茂佳山黑猪和零号土猪3个品种猪肉样品为研究对象，以杜长大作为主品种，茂佳山黑猪和零号土猪作为从品种，采用偏最小二乘(partial least squares regression, PLS)法建立主品种猪肉pH值定量检测模型，模型校正相关系数(correlation coefficient of cross-validation rc)和预测相关系数rp分别达到0.922和0.904，交互验证均方根误差(root mean squared error of cross validation, RMSECV) 和预测均方根误差(root mean squared error of prediction, RMSEP)分别为0.045和0.046，RPD为2.380。用主模型分别对茂佳山黑猪和零号土猪pH值进行预测，rp仅达到0.770和0.731，RMSEP分别为0.111和0.209，RPD分别为1.533和1.234，预测精度较差。分别采用CSPV传递算法和模型更新方法对主模型进行传递和修正，比较并验证了两种方法的模型传递和修正结果。采用CSPV算法对模型传递后，当标样个数分别为9个和10个时，rp可提高到0.889和0.900，RPD提高到2.071和2.213，均满足rp≥0.837，且RPD≥1.9;而采用模型更新方法对模型修正后，当添加的代表性样品分别为11个和9个时，rp分别达到0.869和0.845，但RPD仅达到1.934和1.804，不满足RPD≥1.9的条件。结果表明，CSPV传递算法能实现主模型对茂佳山黑猪和零号土猪样品的预测，而模型更新方法只能实现对茂佳山黑猪品种的预测，不能实现对零号土猪样品的预测，且CSPV传递算法预测结果优于模型更新方法。
关键词：猪肉;pH值;高光谱;模型传递;模型更新
Abstract：Calibration transfer is an effective approach to solve model invalidation problems caused by the change of instruments or the prediction samples. However, most studies on calibration transfer were based on different instruments, and models were established by Near Infrared Spectroscopy. In this study, hyperspectral detecting model of pork pH value was established, and in order to enhance the applicability of model to different breeds of pork samples, a new transfer algorithm based on spectra Mahalanobis distance, sync correction of spectrum and prediction value (CSPV), has been proposed, and was compared with model updating method. Equations with correlation coefficient of prediction (rp)≥0.837 and residual prediction deviation (RPD)≥1.9 were considered as applicable to predict pork quality. In this paper, three breeds, duchangda, maojia and linghao pork were researched, and a pH detecting model of duchangda (the primary breed) was established using partial least squares (PLS) regression method with rc of 0.922, rp of 0.904, root mean squared error of cross validation (RMSECV) of 0.045, root mean squared error of prediction (RMSEP) of 0.046 and RPD of 2.380. However, the prediction of the model to samples from maojia and linghao breeds (the secondary breeds) was very poor with rp of 0.770 and 0.731 respectively, RMSEP of 0.111 and 0.209, RPD reached only 1.533 and 1.234 separately. Obviously, the PLS model of duchangda was unable to achieve the prediction to maojia and linghao samples. With the transformation of CSPV algorithm to duchangda model, only 9 and 10 standard samples from maojia and linghao breeds were used respectively, the prediction ability was improved with rp of 0.889 and 0.900, RPD grew to 2.071 and 2.231, which met the requirement of rp≥0.837 and RPD≥1.9. While with model updating method, when 11 and 9 representative samples from maojia and linghao breeds were added to calibration set of duchangda model, rp increased to 0.869 and 0.845, but RPD only raised to 1.934 and 1.804 exclusively, even though tally rp≥0.837, it didn’t meet that RPD≥1.9. The results demonstrate that CSPV transfer algorithm could realize the pH value prediction of duchangda model to maojia and linghao samples, while model updating method was only applicable for maojia samples instead of linghao samples, and the performance of CSPV transfer algorithm was better than model updating.
刘 娇，李小昱*，金 瑞，徐森淼，库 静 . 不同品种冷鲜猪肉pH值高光谱检测模型的传递方法研究 [J]. 光谱学与光谱分析, 2015, 35(07): 1973-1979.
LIU Jiao, LI Xiao-yu*, JIN Rui, XU Sen-miao, KU Jing . Extending Hyperspectral Detecting Model of pH in Fresh Pork to New Breeds . SPECTROSCOPY AND SPECTRAL ANALYSIS, 2015, 35(07): 1973-1979.
[1] Kamruzzaman M, ElMasry G, Sun D W, et al. Innovative Food Science & Emerging Technologies, 2012, 16: 218. [2] Cen H, Lu R, Ariana D P, et al. Food and Bioprocess Technology, 2014, 7(6): 1689. [3] ElMasry G, Sun D W, Allen P. Food Research International, 2011, 44(9): 2624. [4] ZHANG Hui, SONG Yan, LENG Jing, et al(张 卉, 宋 妍, 冷 静, 等). Chinese Journal of Spectroscopy Laboratory(光谱实验室), 2007, 24(3): 388. [5] LU Wan-zhen(陆婉珍). Modern Near Infrared Spectroscopy(现代近红外光谱分析技术). Beijing: China Petrochemical Press(北京：中国石化出版社), 2007. [6] LIU Xian, HAN Lu-jia, YANG Zeng-ling, et al(刘 贤, 韩鲁佳, 杨增玲, 等). Analytical Chemistry(分析化学), 2012, 40(4): 596. [7] Sulub Y, LoBrutto R, Vivilecchia R, et al. Vibrational Spectroscopy, 2008, 46(2): 128. [8] Du W, Chen Z P, Zhong L J, et al. Analytica Chimica Acta, 2011, 690(1): 64. [9] TAN Chao, QIN Xin, WU Tong(谭 超, 覃 鑫, 吴 同). Computers and Applied Chemistry(计算机与应用化学), 2009, 26(5): 645. [10] GAO Yun, GAO Jian-qin, WANG Xiao-tian(高 云, 高建芹, 王小天). JiangSu Agricultural Sciences(江苏农业科学), 2012, 40(3): 292. [11] Fátima B L L, De Vasconcelos F V C, Pereira C F, et al. Fuel, 2010, 89(2): 405. [12] Salguero C L, Palagos B, Pea R F, et al. Computers and Electronics in Agriculture, 2013, 96: 202. [13] XIE Li-juan, YING Yi-bin(谢丽娟, 应义斌). Journal of Jiangsu University(江苏大学学报), 2012, 33(5): 538. [14] Farrell J A, Higgins K, Kalivas J H. Journal of Pharmaceutical and Biomedical Analysis, 2012, 61: 114. [15] Kapper C, Klont R E, Verdonk J M A J, Urlings H A P. Meat Science, 2012, 91(3): 294.
