基于平行因子算法的三维荧光光谱检测真菌毒素

The three-dimensional fluorescence spectrometric detection of mycotoxin based on parallel factor analysis

真菌毒素具有荧光特性,因此可以用三维荧光光谱分析方法来检测真菌毒素浓度。在检测过程中,通过去除瑞利散射、平滑等光谱预处理后,采用平行因子算法(PARAFAC)建立回归模型,并由该模型依据待测样本的三维荧光光谱数据计算出真菌毒素的浓度。对一组含有不同浓度真菌毒素B1的白酒样本进行实验,结果表明:针对三维荧光光谱数据,由于掺杂的干扰比较多,进行光谱预处理可以提高建模效果;采用三维的PARAFAC建模方法效果要比常规的二维建模方法好;由于PARAFAC对组分数很敏感,所以建模过程中选择适当的组分数对模型的效果影响比较大。

The three-dimensional fluorescence spectrometry could be used to measure the concentration of mycotoxin owing to the fluorescence properties of the mycotoxin. In the experiments, parallel factor analysis （PARAFAC） was used to develop a regression model after the spectrum preprocessing such as removing Rayleigh scattering and three polynomial smoothing. And then the mycotoxin concentrations of unknown samples could be measured via the detected three-dimensional fluorescence spectrometry with the regression model. In this paper, a group of white spirit samples which contain different concentrations of the mycotoxin were detected to carry on experiment research. The results show that the spectral pretreatment can improve modeling effect as the three- dimensional fluorescence spectrometry has many interference factors; it is better to do the modeling using the PARAFAC than the conventional two-dimensional method. To chose the appropriate group scores is very important in the modeling process because the PARAFAC is sensitive to the group scores.