Extraction and Calibration Methods for Fluorescence Spectrum in Mixed Signals

Biological aerosols pose significant threats to human health and the environment. UV LIF lidar serves as an effective detection tool, with the ability to obtain mixed fluorescence spectra that reveal the composition and concentration of bioaerosols in the detection area. This study focuses on enhancing the pre-processing of echo signals from UV LIF lidar for improved bioaerosol analysis. The approach involves utilizing Fast Independent Component Analysis (Fast ICA) to decompose the mixed fluorescence spectrum, Raman signal, and noise. Genetic Algorithm (GA) is then applied to calibrate the extracted mixed fluorescence signal, followed by improved Savitzky-Golay (SG) filtering to eliminate high-frequency noise. The fluorescence spectra are simulated at an excitation wavelength of 266 nm, with certain noise components added to simulate the echo signals of a UV LIF lidar. After amplitude calibration and high-frequency filtering, the mixed fluorescence spectrum extracted by Fast ICA closely match the original spectra, thereby validating the effectiveness of the proposed pre-processing method. The improved SG filter with multiple windows ensures greater stability in both peak offset and intensity error. With appropriately selected parameters, it achieves a minimal peak shift of 1 nm and maintains a low average relative error in peak intensity of 0.8%. To investigate the anti-interference capability and stability of this method, the effects under various conditions, including increased amplitude of uniformly distributed random noise, varying Poisson noise, different levels of Gaussian noise, and diverse mixed fluorescent components were systematically analysed. As long as the fluorescence signal is not overwhelmed by noise, it can be effectively extracted. Furthermore, the use of different mixed components can also achieve effective extraction. This lays a solid foundation for further quantitative and qualitative analysis of biological aerosols, and helps to improve the accuracy and reliability of measurements.