A nonlinear correction method for the error caused by the change of the integration time of the fiber spectrometer

Spectrum acquisition is an important part of spectral analysis. During the collection process, it is a common method to obtain the spectrum with appropriate SNR by adjusting the integration time. However, the pixels of the spectrometer are different with each other, which means changing the integration time will change the sensitivity of the spectrum, but the errors introduced by this process are often unnoticed. Based on the “M plus N” theory, aiming at the systematic error with randomness caused by the change of integration time, a nonlinear correction method is proposed in this paper. In this method, the calibration equation obtained from the calibration of the spectrometer is used to correct the nonlinearity of the spectra collected under different integration time, in order to reduce the error caused by the change of integration time. A group of comparative experiments are designed: Collecting the spectra of different integration time, and then, the unmodified spectra and the modified spectra are respectively used for coverage modeling using PLS method. The result of the experiment shows that the result of modeling with modified spectra is better than that of modeling with unmodified spectra. Rc increased from 0.8933 to 0.9426, and the RMSEC decreased from 10.9075 to 8.102, decreased by 25.72%. This indicates that the correction method proposed in this paper can effectively improve the prediction accuracy of the model, fundamentally reduce the error caused by the change of integration time, and has practical significance for improving the accuracy of spectral analysis.