In recent years, the rapid development of global industry and the advancement of urbanization have triggered a series of environmental problems, in which soil heavy metal lead (Pb) pollution has caused widespread concern among researchers. X-ray fluorescence (XRF) spectroscopy has the advantages of low cost, fast analysis speed and suitable for large-area monitoring. It has been widely used in many fields, such as soil pollution detection and ecological environment protection, and it has broad prospects for development. Therefore, it is of great practical significance to fully excavate the XRF spectral information of soil, which can provide solutions for the efficient detection and prevention of soil pollution, the inversion of soil environmental and ecological parameters, and the early warning of heavy metal pollution in mining areas. At present, most of the research on XRF spectroscopy focuses on the accurate evaluation of soil heavy metal concentration measurement and environmental quality assessment, while there are few in-depth studies on the variation of soil XRF spectral characteristics. Time-frequency analysis method can transform complex signals in the time domain into frequency domain space, and detect abnormal information in spectral signals from the angle of the frequency domain. It is an effective method for detecting the change of spectral difference features. Among them, harmonic analysis (HA) could be used for noise removal of electromagnetic signals; smoothed pseudo Wigner-Ville distribution (SPWVD) selects the appropriate basis function in advance, which can highlight the time-frequency local details of the original signal. This paper firstly used HA method to explore the denoising effect of soil XRF spectra with different Pb concentrations, and then used SPWVD of the spectrum to study the local law of denoising XRF spectra of soil samples sampled in the field. The results showed that when the number of harmonic decomposition was 400, soil XRF spectrum de-drying effect was better and saved time, and retained the characteristics of the spectrum. The Pb concentration of soil samples and the distribution of frequency peaks of SPWVD in the XRF spectrum on the 400, 600～700 band sequences had certain regularity. According to this regularity, the Pb concentration of soil in this area could be identified as exceeding the standard, in all in-situ sampled soil, 75% of samples with excessive Pb concentrations could be identified, there was a higher frequency peak near the band sequence of 400 (frequency less than 400 Hz) or 2 very strong frequency peaks (frequency greater than 400 Hz); in all in-situ sampled soil, 79.17% of samples with Pb concentration not exceeded the standard could be identified, there was a strong frequency peak (frequency greater than 400 Hz) near the band sequence of 400, and there were 3 distinct frequency peak distributions between the 600 and 700 band sequences. Accordingly, it was inferred that the XRF spectral characteristic band interval of soil Pb concentration exceeding the standard in this area was 6.42 and 9.42~10.92 keV. Therefore, by introducing the time-frequency analysis method, soil XRF spectral frequency domain analysis and visualization are realized, which provides a new idea for deep mining the characteristics of Pb pollution spectrum and abnormal information.