Spectral calibration is the premise to realize quantitative of hyperspectral remote sensing. A spectral calibration method based on atmospheric absorption features is presented. The spectral calibration techinque first separate the reflectance caused by atmosphere and surrounding, which is indepedent of the character of the remote sensor, and then is executed by spectral to match between the equivalent spectral reflectance and the one after low-pass filter. Using the spectral calibration method, a domestic airborne hyperspectral sensor is calibrated in flight. The result shows that both the central wavelength and the bandwidth of the sensor change. The accuracy of the spectral calibration can be verified by checking the sharp protruding in the inversion spectral reflectance or by comparing measured and radiative transfer calculated spectral radiance in the oxygen 760 nm absorption peak profiles. Analysis shows that, the main influence factors of the spectral calibration uncertainty are data signal-to-noise and sensor radiometric calibration.