In Fourier transform infrared spectroscopy (FTIR), the original interferometric image needs to be pre-processed by apodization, trend term removal and phase correction before the gas irradiance signal can be obtained by Fourier transform, of which trend term removal is the most important. The common method is least squares (LS), which requires high initial values and is susceptible to noise interference. In this paper, M-estimated sample consistency (MSAC) and genetic algorithm (GA) are used to remove the trend term from methane FTIR simulated interference data and compare them with the least squares method. The results show that: compared with the least squares method, the MSAC algorithm can improve the trend term fit by about 20%, but the trend term pattern needs to be known in advance; compared with the MSAC algorithm, the GA algorithm has a slightly lower fit effect of about 5%, but requires lower initial values, is more robust and is suitable for situations where the trend term pattern is unknown; combining the two, the GA-MSAC algorithm proposed in this paper, which both reduces the initial value requirement and greatly improves the accuracy of the trend term removal, is of great importance to Fourier transform infrared spectroscopy.