For the latter approach we will rely on stitching through structure from motion (Sfm) and the generation of an orthomosaic or orthophoto. An ortho photo corrects remote sensing data so the data represents a perfectly downward looking image, free from any perspective distortion due to topographic relief, lens distortions and camera tilt, ensuring that the scale is uniform throughout (see image above). After these corrections aerial photographs accurately represent the Earth’s surface and measurements on such maps reflect true distances. Using the Sfm technique, relying on stereographic vision and photogrammetry, we can reconstruct a 3D surface and relative camera positions in order to correct arial photograhs. An oblique view of this 3D reconstructed surface of aerial images by a drone is provided below. Notice that at a very high resolution you pick up the relative height of the canopy as well.
A first quick pass using PhotoScan and a swath of survey images (path 9) shows promissing results. Although artifacts remain (as this was a test), the stitching of all images occured smoothly and provides a continuous view of the Yangambi region. At the confluence of the Congo river running from East to West and the Lomami (left side, left bank of the Congo river) is the the village of Isangi, while Yangambi and surrounding plantations is located in the middle of the image (right bank of the Congo river). As a reference, this view is approximately 70km long and 15km wide.
A high resolution version can be found here.
BLOG
aerial photography remote sensing measurements