Study on the Influence of Resolution on Near Infrared Spectra and Quantitative Analysis
Li-juan Xie,Dong-hong Liu,Yu-huan Zhang,Huirong Xu,Xing-qian Ye,Yi-bin Ying
DOI: https://doi.org/10.3964/j.issn.1000-0593.2007.08.009
2007-01-01
Abstract:Near-infrared (NIR) spectroscopy technique is one of the qualitative and quantitative analysis techniques that were developed quickly in recent years. NIR spectroscopic analyses are non-destructive, simple and fast, and require no sample pretreatment, which makes this technology ideally suited for on line process monitoring and quality control. As a powerful analytical tool in product quality determination, this technology is based on the measurement of the frequencies of vibrations of chemical bonds in functional group such as C-C, C-H, O-H, CO and N-H upon absorption of radiation. A good and robust model is very important for NIR spectroscopy analysis. However, it was found that the status of NIR spectrometer and the set of parameters when scanning, such as accuracy of wavelength, resolution of apparatus, noise, scan time and uniformity of sample size, had effect on the spectrum quality. In the present paper, NIR transmittance spectra of bayberry juice were obtained and quantitative analysis was carried out. The influence of resolution on the NIR spectra and quantitative models was studied. Forty three bayberry samples with different varieties and ages were used in this experiment. Cross-validation was used to develop models and evaluate and compare these models. The results show that resolution does have effect on the NIR spectra and quantitative models. The spectrum with high resolution was rougher than that with low resolution. Moreover, the scan speed was lower and more data size was required when high resolution was used. The difference in root mean square noise and mean absorbance at different resolutions was distinct (alpha=0.05). The best acid model was developed with 4 cm(-1) spectral resolution, giving the relative high correlation coefficient of 0.994 04, and a root mean square error of calibration (RMSEC) and root mean square error of cross validation (RMSECV) of 0.023 3 and 0.153, respectively. However, if no information was lost, a resolution of 8 cm(-1) was suggested considering the scanning time and data size.