Application Of Orthogonal Method In Near-infrared Spectral Analysis

Near-infrared spectral analysis relies on mathematical models to predict the samples information, but near-infrared spectra derives in multiplied and combined frequency absorption of molecular vibration, and its background is complex, therefore, it is particularly important to adopt scientific methods to pre-process spectral data. Direct orthogonal method is a method that the spectral matrix and concentration matrix directly orthogonal to filter out irrelevant signals, there is importantly academic value and practical significance to apply this method in near infrared spectral data pre-processing.The knowledge of multi-disciplinary has been integrated in this paper. On the basis of studying the analytic methods and the basic principles of near-infrared spectral samples, near-infrared spectral pre-processment methods have been researched, and the basic principles, the algorithm design of orthogonal method and the application in near-infrared spectral pre-processment have been explored. Around the orthogonal spectral data pre-processing, the related methods—samples selection, singular points detection, wavelength optimization, which involved in near-infrared spectral analysis have been introduced. On this basis, the calibration model of near-infrared spectra of the typical samples has been built by partial least-squares (PLS) method. The typical samples are the home-made dual-component solution (albumin and glucose) and the milk whose spectra have been pre-processed by orthogonal method, the predicted ability of the model has been evaluated. And the feasibility of direct orthogonal (DO) method applied in spectral pre-processing has been verified.The algorithm design and program implementation of direct orthogonal(DO) method and the PLS model method in the Matlab environment have been completed in this paper, and been successfully verified by the tests. Direct orthogonal(DO) method is to the spectral matrix and concentration matrix orthogonal deal before setting up at quantitative calibration model, By this way, the information which was unrelated the composition and concentration of samples can been filtered out from the spectral information, and the purpose of purification spectra has been achieved. This method not only effectively solves the spectral data pre-processing problem in the near-infrared spectral analysis, but also can give the similar analysis referenced ways and means.