The analysis of the missile plume shows that the main factors influencing its spectral change include the engine, fuel, etc., so the purpose of missile model recognition can be achieved according to the tail flame spectrum. In order to ensure the efficiency of recognition, the characteristic spectrum was used to represent the plume characteristics, which can greatly reduce the data volume. Firstly, the radiation difference of each wavelength was calculated according to the spectral difference model. By setting the threshold value, the radiation difference segment was obtained, and the band of higher than the threshold was the selected characteristic band. Changing the number of steps and thresholds can obtain several sets of different data. The Spectral Angle Matching algorithm (SAM) and fuzzy algorithm were used to deal with the data of different precision and different feature bands. The accuracy of the recognition results and the similarity distance between the samples and the different spectra were measured, and the fuzzy algorithm can identify with SAM, but it was better than SAM in algorithm complexity.