Study On Methodology Of Near-infrared Diffuse Reflectance Spectroscopy For Sediment Components Analysis

The article summarized the present situation application of the near-infrared spectrumanalysis technology, introduced the pretreatment method of spectrum which widely used, alsomodel evaluation index. The key part is based on the work to set up the method of spectrumpretreatment and the optimization prediction model. Compared the result of principalcomponent regression model and the partial least-square regression model, the evaluationmodel of fitting prediction ability, we found suitable modeling method for all differentcomponents in sediment. Then we applied on the case in Qilihai area (Tianjin), use those modelto predict the components in sediment samples, and with the statistics processing of spectraldata, finally we get temperatures and precipitation factor consisted with local historymeteorological record. Specific content as follows:Use different spectral pretreatment method to evaluate to model, and finally determinedspecific pretreatment parameters apply to determine the content of components in the sedimentsamples:1, To balance the need of the accuracy of the efficiency of analysis we decided thatPCR model use the spectrum resolution of50cm-1, PCR model use the spectrum resolution of2cm-1.2, Use Savitzky Golay49point smoothing method to participate in all the spectrummodel preprocessing.3, All the spectrum in model use the multiple scattering spectra ofspectrum correction pretreatment.4, For baseline correction, PCR model use the spectrumOffset the treatment method, PLSR model use order and second order derivatives pretreatmentwould improve the fitting and forecasting ability of predict model.Select the optimal modeling method: according to the prediction result, differentcomponents should use different modeling methods, which final correlation coefficient issatisfactory and standard deviation were low. Among them, for Al, TFe, Mn, and other majorelements in sediment, those correlation coefficient are above0.8, Cr, Cd, Pb and other heavymetals are above0.85, and partial extraction of heavy metal elements, and rare earth elementsalso had good correlation.Take the model to apply in the case in Qilihai (Tianjin) the predict the content of sedimentsamples, the variation scope and changing trends of prediction are identical with the measuredvalues, and combined with the local history meteorological data, using statistical analysismethod extracts, we found the temperatures and precipitation factor consisted with local historymeteorological record.