A Multitemporal UAV Images Registration Approach Using Phase Congruency

If there are great illumination and contrast changes for multitemporal unmanned aerial vehicle (UAV) images, common image matching and registration methods may not work well on these images. To improve matching effects, this paper conduct the image registration using the structure consistency between the images to be matched, first, use phase congruency(PC) to describe images' structure characteristics and generate the PC images for both images to be matched; then, conduct the image matching by extracting the features from accelerated segment test (FAST) in the PC images; finally, use the method of randomized sample consensus (RANSAC) to eliminate the mismatches and obtain the affine transformation matrix between the two images. We use this method to pick out known control points on new UAV images on the basis of old image database of ground control points (GCP), the test proves that the method is independent of the radiation information and has a good effect on multitemporal UAV images registration.