以多环芳烃中的芴和苊为研究对象, 提出一种将三维荧光光谱技术与Krawtchouk图像矩、 广义回归神经网络相结合的定量分析的方法。 利用FS920荧光光谱仪获取样品的三维荧光光谱数据, 得到对应的三维光谱灰度图。 直接计算三维光谱灰度图的Krawtchouk矩, 将得到的Krawtchouk矩经平均影响值筛选后作为广义回归神经网络的输入, 建立多环芳烃(PAHs)的定量模型。 预测8组混合溶液的测试样本, 芴和苊的平均相对误差分别为0.98%和2.15%。 研究结果表明, Krawtchouk矩经过筛选后预测结果更为准确, 该方法能够有效提取光谱的特征信息, 简单、 准确的预测PAHs的浓度。

The study objects of this paper were PAHs fluorene and acenaphthene. A method combining three-dimensional (3D) fluorescence spectroscopy with Krawtchouk moment and generalized regression neural network was proposed for quantitative analysis of PAHs. By using the 3D fluorescence spectra data of samples measured directly, the corresponding grayscale images of 3D spectra could be obtained. The Krawtchouk moments were directly calculated based on the grayscale images of 3D spectra, and the quantitative models for the PAHs were established on the mean impact value and the generalized regression neural network. The average relative errors of the 8 groups mixed samples of fluorene and acenaphthene were predicted to be 0.98% and 2.15%, respectively. The results showed that the proposed method can extract the characteristic information of the spectra effectively and predict the concentration of PAHs simply and accurately.