Experimental Study of Spectral Processing Technology for Improving Detection Ability of Low-Concentration NaCl Aerosol by Filament-Induced Fluorescence Spectrum

Air pollution greatly impacts the productivity and life of people. Aerosol is an important pollutant and cannot he ignored. It is significant to improve the detection accuracy of atmospheric aerosol concentrations, especially low-concentration aerosols. In this study, filament-induced fluorescence spectrum (RFS) of NaCl aerosol is preprocessed and combined with partial least squares (PLSR) to establish a prediction model and explore the impact of different preprocessing methods on the detection accuracy of the model. To choose the preprocessing method, this study divides the preprocessing methods into three aspects according to their effects: scattering correction, smoothing and denoising , and baseline correction, and the significance of peak algorithm is proposed. The optimal preprocessing method is selected after comparing no preprocessing, single preprocessing, and combined preprocessing and analyzing the influence of different preprocessing methods on the accuracy of PIPS spectral prediction model. The experimental results show that the combined preprocessing of multiple methods reduces the root mean square error to 0.03 compared with no preprocessing, and the relative prediction error is reduced by 60%. Compared with the direct observation of spectral signal selection preprocessing method , the best preprocessing method can be selected more accurately according to the improvement of spectral signal-to-noise ratio and the modeling effect of predicted components. The present study provides a reference for the analysis and research of low-concentration air pollutants.