EXPERIMENTS FOR UNMANNED AERIAL VEHICLE MULTISPECTRAL IMAGE REGISTRATION

Although the task of image registration has already been studied for a long time by the remote sensing (RS) community, it is a fact that multispectral data from Unmanned Aerial Vehicles (UAV) have only recently been employed with great interest and success in several application domains. The studies for this type of data are reduced, and a multi-view evaluation considering not only different algorithms but also distinct types of LULC images (Water Body, Vegetation, Built-up Area), band combination and periods (Morning, Midday, Late Afternoon) is still missing. Our experiments considered images from three different flights in Pará State, three classical methods: Enhanced Correlation Coefficient (ECC) with Homography/Affine motion models, and SIFT. The results show that 3D effects present in the Homography model overcome other perspectives and methods. This conclusion is important to maintain attention on the classical approaches, launching the challenge of improving the computational effort to this new data.