Aiming at the UAV air-ground channel, based on the Space-Alternating Generalized Expectation-maximization(SAGE), the channel characteristic parameters are estimated with high precision, and the small-scale fading characteristics of the air-ground channel such as multipath delay and multipath power are extracted. Then, using uniform quantization and non-uniform quantization methods, the wireless channel key quantization is carried out on the main path power and the main path-multipath power difference of the measured data. The quantification efficiency, randomness, and algorithm running time of the key are analyzed for takeoff and cruise scenarios respectively, and compared with the key quantification results based on large-scale features. The key quantization efficiency results show that the non-uniform guard quantization is superior to the uniform guard quantization; the quantization method based on high-precision estimation of channel characteristics is better than traditional quantization based on large-scale features; and the quantization in the take-off scenario is better than that in the cruise scenario. The results of the performance test show that the keys obtained in these quantizations have good randomness; the running time difference of different quantization methods is small, therefore, the two quantization methods based on high-precision extraction bear low complicity.