Building and Evaluation of a Mosaic of Images Using Aerial Photographs

This paper addresses the image mosaicing problem using images from a UAV. First the Harris-Laplace method is used to find scale-invariant keypoints. Using the same method as in SIFT, it is possible to compute a descriptor for each keypoint. Ransac, together with the DLT algorithm, is then used to robustly estimate the homography between two images. In order to assess the robustness of the method, a simulator was developed to take aerial photographs from an input image representing the Earth surface. Results show the importance of the minimization of the tilt angles of the camera, as well as the necessity for a significative overlapping between images. Results also show that the images should be taken at high altitudes to reduce parallax errors.