Abstract: Recently,image registration technology is one of the rapid development field in image processing area.In remote sensing field,it is a significant step for image fusion,moving detection,image correction,image mosaic and so on.Although there are many methods for image registration in the world,different methods apply to different kinds of images,most time the methods are selected by human intervening but not automatically.It becomes a key issue that how to combine the advantages of different methods to achieve automatic image registration,especially for UAV images.In this paper,a pre-classifying method based on geometric moment is proposed after the comparison of image registration methods based on SIFT and SURF feature extraction,in order to decide which image registration method is the best one,thereby,achieve the whole automatic process.The experiments show that this automatic image registration method makes sure a good matching effect and at the same time it broaden the types of applicative images.

Key words: UAV (unmanned aerial vehicle) image, automatic registration, SIFT (scale invariant feature transform), SURF (speeded-up robust features), geometric moment