With the continuous development of camouflage technology, the spectral similarity between the camouflage target and the background is getting higher and higher, which brings challenges to the recognition task. Most of the existing band selection methods focus on the information content of the band or the overall separability of each band image, so it is difficult to select the feature band combination that can distinguish the similar spectrum effectively. Therefore, a hyperspectral image band selection method for camouflage target recognition was proposed. The spectral difference index model was constructed to quantitatively describe the spectral difference between camouflage target and similar background in each band, and then guided the selection of feature bands. Firstly, the spectral gradient angle was introduced to explore the local morphological features of the spectrum. Then, the amplitude differences between spectra were measured by Fréchet distance and normalized to eliminate the effect of scale changes. Finally, Pearson correlation coefficient was used to strengthen spectral difference and band independence. The experimental results on the camouflage vehicle data set and the San Diego airport public data set show that the proposed method is better than the comparison method in the camouflage target recognition task.