Cloud camera is designed to perform cloud recognition and cloud coverage discrimination of the target area, and it adopts the threshold method for cloud detection. To satisfy the accuracy of cloud detection, it is important to determine the relationship between the response value of the detector and the radiance. We take the Bayer array based plane array color CCD camera adopted by the cloud camera as the research object, analyze its imaging process, and find that the radiation response spectra of R, G, and B channels overlap. Based on this phenomenon, a multi-channel CCD detector radiation response calibration model is established. For the R and B channels used by the cloud camera, the 747 nm LED and 443 nm LED are used as the light sources of the integrating sphere, and the minimum, typical, and maximum values of spectral radiance are set. The absolute radiation response calibration experiment of the cloud camera is carried out under the single light source condition to determine the corresponding absolute radiation response coefficients, and the correctness of the proposed radiation response calibration model is verified under the dual light source condition. This method eliminates the calibration uncertainty caused by the detector response aliasing and effectively improves the accuracy of absolute radiation calibration.