基于近红外光谱法快速测定许氏平鲉脂肪和 水分含量的方法研究

Research on rapid determination of fat and moisture content of Sebastes schlegeli by near infrared spectroscopy

中文关键词:  近红外光谱法  脂肪  水分  偏最小二乘法  许氏平鲉鱼

英文关键词:near infrared spectroscopy  fat  moisture  partial least square  Sebastes schlegeli

中文摘要:

      目的 建立近红外光谱法结合偏最小二乘法测定许氏平鲉鱼肉中的脂肪和水分含量, 以期简便、快速地对许氏平鲉进行品质分析与评价。方法 采用常规分析手段测定70个样品的脂肪和水分含量, 同时采集其近红外光谱数据, 结合偏最小二乘法(partial least square, PLS)建立许氏平鲉鱼肉中脂肪和水分的定量预测模型, 并对比不同光谱预处理方法、光谱范围和因子数对定量预测模型的影响。结果 光谱经Savitzky-Golay(S-G)和标准正态变量变换(standardized normal variate, SNV)预处理后, 在5341.85~4007.36 cm-1、6556.79~5345.71 cm-1和8651.10~7162.33 cm-1光谱范围内, 选取主因子数10, 建立脂肪的校正模型性能最优; 光谱经过SNV预处理后, 在8886.38~4061.35cm-1光谱范围内, 分别选取主因子数为9时, 建立的水分的校正模型性能最优。脂肪和水分含量相对最优PLS模型的校正集相关系数分别为0.9918和0.9912, 校正标准偏差分别为0.2680和0.3300, 交叉验证相关系数分别为0.9820和0.9810, 交叉验证均方差分别为0.3980和0.4850, 验证集相关系数分别为0.9804和0.9798, 验证集均方差分别为0.3260和0.3070。结论 本方法可较为准确地预测许氏平鲉鱼肉中的脂肪和水分含量, 能够满足快速分析评价许氏平鲉品质的要求。

英文摘要:

      Objective To establish a method for determination of fat and moisture content in Sebastes schlegeli by near infrared spectroscopy (NIRS) combined with partial least square (PLS), so as to evaluate the quality of Sebastes schlegeli simply and quickly. Method Fat and moisture results of 70 samples were obtained by ordinary analytical methods. Meanwhile, NIRS data of these samples were investigated in order to establish quantitative prediction model for Sebastes schlegeli nutrients combined with PLS. The influences of different spectra pretreatment methods, different spectra regions and the number of factors were compared. Results The performance of fat content model was established in 5341.85~4007.36 cm-1, 6556.79~5345.71 cm-1 and 8651.10~7162.33 cm-1 after Savitzky-Golay (S-G) and standard normal variate (SNV) pretreatment, and the optimal main factor number of 10 was selected. The performance of moisture content model was established in 8886.38~4061.35 cm-1 with SNV pretreatment, and nine factors were optimal. The correlation coefficients of calibration (Rc) of fat and moisture were 0.9918 and 0.9912, and the root mean square errors of calibration (RMSEC) were 0.2680 and 0.3300, respectively. The correlation coefficients of cross validation (Rcv) were 0.9820 and 0.9810, and the root mean square errors of cross validation (RMSECV) were 0.9804 and 0.9798. The correlation coefficients of prediction (Rp) were 0.9804 and 0.9798, and the root mean square errors of prediction (RMSEP) were 0.3260 and 0.3070. Conclusion The method has acceptable accuracy and prediction capability, which is suitable for rapid quality analysis and evaluation of Sebastes schlegeli.