Abstract: Unmanned aerial vehicle has many characteristics, such as the relatively stable aerial navigation, flight speed, flight height. These characteristics decide that there are relatively fixed positions between the neighbor images. Based on these characteristics of the unmanned aerial vehicle (UAV) images, the constraint condition of the mosaic method on the UAV images is proposed, and then a fast Automatic Mosaic Method of the UAV images is proposed. Firstly, the proposed method use the AR model which use the center position of the history overlap areas as the input data to forecast the overlap areas of the two images where the SIFT features are extracted, and then mosaic the two image using the extracted SIFT features. The results of experiments on the UAV images show that the method has good performance on the mosaic image, and the time cost reduced largely.