The matching effect of conventional methods is poor, sometimes even not matching due to the instability of Unmanned Aerial Vehicle (UAV) air gesture and the defects such as the existence of its tilted capture images, uneven exposure. SIFT algorithm has a good adaptability for image geometric distortion, resolution differences and rotation, etc. The SIFT algorithm is introduced to extract the feature points of UAV images, and the least-squares algorithm is used to match the images precisely. Moreover, the block splicing method is proposed to reduce the error spread in the process of stitching. The experimental results show that SIFT image matching algorithm has a good adaptability in matching UAV images. The matching error can be eliminated effectively to introduce the least-squares algorithm in the matching process. What is more, the time and space complexity can be obviously reduced.