The Detection Method Of Mycotoxin In Food Based On Three-dimensional Fluorescence Spectrometry

Detection of mycotoxins in food become a focus of research in recent years because of strong toxicity and high frequency of pollution. The three-dimensional fluorescence spectroscopy can describe fluorescence intensity that responds with changes in excitation and emission wavelengths, which can completely describe the substance of the fluorescence characteristics of the spectrum and is a valuable fluorescence fingerprint technology. Therefore application of three-dimensional fluorescence spectroscopy to detect mycotoxins is feasible. However, the three-dimensional fluorescence spectra data preprocessing method is rarely. Meanwhile two-dimensional fluorescence spectra data preprocessing methods do not apply to the three-dimensional fluorescence spectroscopy, these methods including pro-smoothing processing, spectra normalized method of removing drift curve, and the selection method of the characteristic spectrum. Therefore, the paper studies some three-dimensional fluorescence spectrum methods used in detecting mycotoxins of wine, these methods including data pro-smoothing processing, differential spectrum, and the information region extraction. The studies are following:1. Based on the smoothing methods of Image Processing, the paper studies and compares the mean filtering,median filtering,polynomial filtering and Fourier analysis on the three-dimensional fluorescence spectra, which can effectively remove the high frequency interference noise sound of the three-dimensional fluorescence spectra.2. For the differential spectrum can eliminate baseline drift and strengthen the band characteristics this paper research the first order derivative spectra and second derivative spectra of the three-dimensional fluorescence spectra base on polynomial fitting.3. For three-dimensional fluorescence spectral data having high-dimensional data volume and large amount of data, the researchers cannot determine the useful information region based on prior knowledge and gross. So the method that is integration of convex domain and clustering analysis on the number of three-dimensional fluorescence is presented to reduce the data dimension and eliminate information irrelevant area.4. The proposed method is applied to the sample group of mycotoxins Bl experimental analysis in wine, and the result confirmed the validity of the study.