基于多源光谱分析技术的鱼油品牌判别方法研究

多源光谱分析技术被用于鱼油品牌快速无损鉴别。采用可见光谱分析技术、短波近红外光谱分析技术、长波近红外光谱分析技术、中红外光谱分析技术和核磁共振光谱分析技术采集了7种不同品牌的鱼油的光谱特征,并应用偏最小二乘判别分析法(partial least squares discrimination analysis,PLS-DA)和最小二乘支持向量机(least-squares support vector machine,LS-SVM)建立判别模型并比较判别结果。基于长波近红外光谱的PLS-DA模型和LS-SVM模型取得了最高识别正确率,建模集和预测集识别正确率均达到100%。采用中红外光谱和核磁共振谱分别建立的LS-SVM模型,也可以获得100%的判别正确率。而可见光谱和短波近红外光谱则判别准确率较差。且LS-SVM算法较PLS-DA更加适合用于建立光谱数据和鱼油品牌之间的判别模型。研究结果表面长波近红外光谱技术能够有效判别不同鱼油的品牌,为将来鱼油品质鉴定便携式仪器的开发提供了技术支持和理论依据。

Study on Brand Discrimination of Fish Oil Based on Multiple Spectroscopy Techniques

In this paper, multiple spectroscopy techniques were used to distinguish different brands of fish oil in a rapid and non-invasive manner. Spectral characteristics of seven brands of fish oil, collected by visible spectroscopy, short wave near infrared spectroscopy (SNIR), long-wave near infrared spectroscopy (LNIR), mid-infrared spectroscopy (MIR), and nuclear magnetic resonance (NMR) spectroscopy, were set as inputs in partial least squares discrimination analysis (PLS-DA) and a least-squares support vector machine (LS-SVM) to establish the discrimination models. The discrimination results of the PLS-DA and LS-SVM models were subsequently compared. The results showed that LNIR achieved the highest discriminant accuracy, and the accuracies of modeling set and prediction set were up to 100%. The LS-SVM model using MIR and NMR spectroscopy also yielded a discriminant accuracy of 100%. On the other hand, the discriminant accuracies of those based on visible spectroscopy and SNIR were poor. In addition, LS-SVM was more suitable than PLS-DA to build identification models for fish oil brands using spectroscopic data. The results indicated that LNIR spectroscopy technique could effectively distinguished fish oil brands, providing the technical support and theoretical basis for developing portable instruments for the analysis of fish oil quality in the future.