The cloud detection technology is one of the key steps to implement the application of the assimilation of the Atmospheric Infrared Sounder (AIRS). Mitch Goldberg’s cloud detection is a very fast model based on an empirical combination of 3 tests applied to AIRS channel and co registered AMSUA channels. The cloud detection does not need to apply a channel bias correction. Also, it is relatively independent of atmospheric prior information, except for sea surface temperature. According to the Mitch Goldberg’s thought of cloud detection, the original 3 tests of NESDIS Goldberg cloud detection scheme have been improved to 7， 4 in sea surface and 3 in land surface. The improved cloud detection is conducted to AIRS FOV of the land and sea respectively with the use of GRAPES 3DVAR (Global/ Regional Assimilation and Prediction System). There is a cloud, if every single FOV passed the cloud detection scheme. The NESDIS Goldberg cloud detection scheme and GRAPES Goldberg cloud detection scheme are used to detect the 6 scenes of AIRS on 1 July 2006 respectively. The analysis result shows that, the GRAPES Goldberg cloud detection scheme can effectively find out the FOV contaminated by cloud, and the clear FOV accounts for 10.1% of the total FOV.