三维荧光光谱结合Zernike图像矩快速鉴别掺伪芝麻油

为了实现对掺伪芝麻油的快速鉴别，应用FS920荧光光谱仪测定样品的三维荧光光谱数据。将三维荧光光谱图视为灰度图，在没有任何预处理的前提下，直接应用Zernike图像矩提取三维光谱灰度图的特征信息，然后采用类平均法对特征信息进行聚类分析，从定性角度实现掺伪芝麻油的鉴别，并解析其组成成分。最后应用广义回归神经网络(GRNN)对掺伪样本的成分进行定量分析。聚类分析能够以很高的辨识率来识别掺伪芝麻油，并能够正确解析其组成成分。定量模型预测了2组掺伪样本中各成分的相对体积，其平均相对误差分别为2.23%，8.00%，9.70%和9.70%。分析结果表明，Zernike矩能够有效提取光谱的特征信息，光谱数据的Zernike矩特征结合聚类分析以及GRNN模型能够获得良好的定性和定量分析结果，为掺伪芝麻油的鉴别提供了一种新的方法。

3D Fluorescence Spectra Combined with Zernike Image Moments for Rapid Identification of Doping Sesame Oil

In order to realize the rapid identification of dopingsesame oil, the three-dimensional fluorescence spectra of the samples were measured by FS920 fluorescence spectrometer.The three-dimensional fluorescence spectrum was regarded as the gray scale graph,and the characteristic information of three-dimensional spectral grayscale was extracted directly by Zernike image moment without any pretreatment.Then, the characteristic information was clustered and analyzed by using the class mean method to identify the doping sesame oil and its constituent components. Finally, the generalized regression neural network (GRNN) was used to quantitatively analyze the components of the dopingsesame oil. Clustering analysis can identify adulterated sesame oil and its composition. The average relative error of the two groups was 2.23%, 8.00%, 9.70% and 9.70%, respectively. The results showed that the Zernike moments can effectively extract the characteristic information of the spectra. The proposed method of Zernike moments combined with clustering analysis and GRNN model can obtain satisfactory qualitative and quantitative analysis results, which will provide a new method for the identification of doping sesame oil.