Petroleum hydrocarbons such as machine oil and diesel are important components of soil pollution, and are of great significance for rapid and accurate detection of organic pollutants such as machine oil and diesel in soil. Laser-induced fluorescence (LIF) technology has the advantages of fast detection speed, high sensitivity and on-site detection. However, when detecting organic pollutants in soil, it faces serious problems such as overlapping fluorescence spectra. In order to study the overlapping characteristic of the fluorescence signals of the machine oil and diesel mixture in the soil, 10 soil samples containing different concentrations of machine oil and diesel mixture were prepared. By establishing the LIF experimental system, the fluorescence signals of different mixing concentrations of machine oil and diesel were obtained, and the inversion relationship between the mixed spectra of machine oil and diesel was established. The iterative approximation algorithm was used to calculate the fluorescence contribution rate of diesel and machine oil samples in soil fluorescence spectra. In the process of calculating the fluorescence contribution rate, the two methods of full spectrum and intercepted characteristic spectrum were compared. When linearly fitting with the machine oil sample concentration, the fitting coefficient R of the intercepted characteristic spectrum method was 0.989, and the average relative error was 3.38%, which was better than the full spectrum of 0.923, 8.79%. At the time of verification, the average relative error of multiple linear regressions was 10.11% compared with the multiple linear regression method, which prove that the intercepted characteristic spectroscopy method is still excellent. There was a good linear relationship between the fluorescence contribution rate of machine oil and diesel in soil and its own concentration, indicating that there is no chemical reaction after mixing machine oil and diesel in soil, and the overlapping characteristic of fluorescence signals in soil are linearly superimposed. The method is equally applicable to the separation of fluorescence spectra of other petroleum hydrocarbon mixtures in the soil. Through the research in this paper, the accuracy of qualitative and quantitative detection of petroleum hydrocarbon pollutants in soil by LIF technology was improved. It provided method support for rapid detection of petroleum hydrocarbons in the soil.