With the data in Urad Front Banner, Inner Mongolia on November 14th, 2010, hyper-spectral camera on UAV was calibrated adopting reflectance-based method. During the in-flight absolute radiometric calibration, 6 hyper-spectral radiometric gray-scale targets were arranged in the validation field. These targets' reflectances are 4.5%, 20%, 30%, 40%, 50% and 60% separately. To validate the calibration result, four extra hyper-spectral targets with sharp-edge spectrum were arranged to simulate the reflection and absorption peaks in natural objectives. With these peaks, the apparent radiance calculated by radiation transfer model and that calculated through calibration coefficients are much different. The result shows that in the first 15 bands (blue bands), errors are somewhat huge due to the noises of equipment. In the rest bands with quite even spectrum, the errors are small, most of which are less than 10%. For those bands with sharp changes in spectral curves, the errors are quite considerable, varying from 10% to 25%.