The fluorescence detection method is based on the linear relationship between fluorescence intensity emitted by the material and the concentration of material to make a quantitative analysis. When using the fluorescence detection of atmospheric sulfur dioxide and other harmful gases, photodetectors and other optoelectronic components without fluorescence will continue to produce the dark current noise, and the background signal has a direct impact on the measurement results. On the base of analysis Boxcar filtering algorithm, the research used three algorithms of wavelet filtering, EMD filter and Boxcar filter to extract and recover the fluorescence signal drowned in the noise floor. In comparison with the previous two filtering methods, Boxcar filter had a better effect on the suppression of the background noise. It also verified that the number of sampling affects the fluorescence signal to noise ratio improvement.