Key wavelength selection using CARS method in near infrared spectra

In near infrared spectroscopy, there exist a lot of wavelength points, and bands are difficult to be attributed. In addition, there exit large measurement error because of serious spectral overlap. So, it is difficult to build multivariate calibration model based on all the spectra. Moreover, the overlap of the strong absorption background wave section of water and the other components in the near infrared spectra will reduce the model robustness and metastatic. Therefore, the CARS method was adopted for the wavelength variable selection in this paper. The selected key points are used to build model for total sugar, total nitrogen and total nicotine index. Compared with full spectrum, the performance of quantitative model has been greatly improved. ©, 2014, Binary Information Press.