近红外技术在南极磷虾油关键品质指标快速检测中的应用研究

苗钧魁; 张雅婷; 金永霈; 刘小芳; 于源; 冷凯良; 杨增光; 蒋永毅

doi:10.13386/j.issn1002-0306.2021120098

近红外技术在南极磷虾油关键品质指标快速检测中的应用研究

Application Research of NIR Technology on the Fast Quantification of the Key Quality Indicators of Antarctic Krill Oil

摘要

摘要: 采用偏最小二乘法（Partial least squares，PLS）作为建模方法，对磷虾油近红外光谱的一阶微分（First-order difference，FD）、FD+SG（Savitzky-Golay，SG）滤波、FD+N（Norris，N）滤波、二阶微分（Second-order difference，SD）、SD+SG和SD+N等6种单一或复合方法进行处理，通过对不同方式处理后预测模型的交互验证均方根误差（Root mean square error of cross validation，RMSECV）、外部验证残差均方根（Root mean square error of external prediction，RMSEP）和外部验证用样品真实值的标准差（SD）与RMSEP的比值（The ratio of the RMSEP to standard deviation of reference data in the prediction，RPDEV）、建模相关系数（Correlation coefficient in calibration，RC）、交互验证相关系数（Correlation coefficient in cross validation，RCV）和外部验证相关系数（Correlation coefficient in external validation，REV）等参数比较，确定了磷虾油磷脂、EPA和DHA的近红外预测模型最佳处理方式为FD、FD和SD+N，酸价指标模型不需处理。在最优条件下，四种成分近红外预测模型的RC、REV和RCV，除了酸价的RCV略小（0.917）其余均达到0.95以上，同时，四种成分的RPDEV和RPDCV值，除酸价的RPDCV为2.365，略小于2.5，其余均符合大于2.5的要求，说明磷虾油磷脂、EPA和DHA的近红外预测模型预测准确度良好；RMSEC和RMSECV相差不大，说明模型稳定性较好。由于含量低、组成复杂等原因，磷虾油虾青素近红外检测模型的RC、RCV和REV均在0.60以下，说明近红外检测不适用于磷虾油中虾青素成分的快速检测。本文证实了近红外光谱技术可作为磷虾油中磷脂、EPA、DHA和酸价等主要指标的快速检测方法，是传统化学检测方法的有效替代和补充。

Abstract: In this paper, near-infrared spectroscopy technology was used to establish a rapid determination of the content of phospholipids, eicosapentaenoic acid (EPA), docosahexaenoic acid (DHA), astaxanthin and acid value in Antarctic krill oil. Partial least squares (PLS) was used as modeling method. The NIR spectra of krill oil were treated by first-order difference (FD), FD+Savitzky-Golay (SG), FD+Norris (N), second-order difference (SD), SD+SG and SD+N. And root mean square error of cross validation (RMSECV), root mean square error of external prediction (RMSEP), the ratio of the RMSEP to standard deviation of reference data in the prediction (RPDEV), correlation coefficient in calibration (RC), correlation coefficient in cross validation (RCV) and correlation coefficient in external validation (REV) of the prediction models were compared. It was determined that the best treatment methods for phospholipid, EPA and DHA in krill oil were FD, FD and SD+N, and the acid value model did not need to be treated. Under the optimal conditions, RC, RCV and REV of the four components in the NIR prediction model were all above 0.95, except that RCV of acid value was slightly lower (0.917). Meanwhile, the RPDCV and RPDEV of the four components, except for the acid value, which was 2.365, slightly less than 2.5, the rest meet the requirements of greater than 2.5. It showed that prediction model of phospholipid, EPA and DHA of krill oil using near infrared spectroscopy had a good prediction accuracy. The difference between RMSEC and RMSECV was not significant, indicating that the model had good stability. Due to the low content and complex composition of astaxanthin in krill oil, the RC, RCV and REV of the NIR quantification model were all under 0.60, near infrared spectroscopy (NIR) was not suitable for the rapid quantification of astaxanthin in krill oil. In this study, it was confirmed that NIR could be used for the rapid quantification of phospholipids, EPA, DHA and acid value in krill oil and was able to be used as an effective substitute and supplement for traditional chemical detection methods.