A multi-spectral remote sensing image mosaic technique with scale invariant feature transform (SIFT) feature matching is proposed to deal with images obtained from an unmanned airship. The acquired pictures usually do not meet the specifications of aerial photography because the airship is unstable in wind. We propose to use SIFT feature vectors with spectral information to improve robustness of the mosaicing algorithm. The BBF(best-bin-first) algorithm and RANSAC（random sample consensus）methods are used for coarse and fine matching processing, and error removal. The optimal transformation matrix from SIFT feature matching calculation is used to achieve image mosaicing. Experimental results show that the algorithm can produce a large number of matching feature points to obtain a stable transformation matrix for further image mosaicing, with accuracy that meets the needs of image interpretation.