Abstract:
It is the key step of the Near Infrared Spectral analysis that the chemometrics is applied to the near infrared spectral modeling. With the deep-going research, much more researchers pay attention to the improving of existing algorithms and the application of new methods to the NIR analysis. The data preprocessing, qualitative and quantitative research and modeling optimization methods have been analyzed. In De-noising methods, Wavelet Transform (WT) is the most commonly used and effective tool. Now, wavelength selection method is one of the key study, the main including Genetic Algorithm (GA),Uninformative Variable Elimination (UVE), Successive Projections Algorithm (SPA) and so on. In terms of modeling, a variety of algorithms has been improved based on Partial Least Squares (PLS), Artificial Neural Networks (ANN) and Support Vector Machine (SVM), which can be optimized the predictive models effectively. However, each algorithm have the advantages at same time with some limitations, also the optimization methods of modeling are different for different types of analytes. So it will be the future research trends that a variety of mathematical methods have been combined and mutual complemented in the practical application.