A COMPARISON BETWEEN "OLD AND NEW" FEATURE EXTRACTION AND MATCHING TECHNIQUES IN PHOTOGRAMMETRY
Andrea Maria Lingua, Davide Marenchino, Francesco Nex · PORTO Publications Open Repository TOrino (Politecnico di Torino) · 2009
The development of new photogrammetric systems has changed the user demand. At present, the images acquired by Unmanned Aerial Vehicles (UAV) and Mobile Mapping Technologies (MMT) are far from the normal condition and they also still need to reach reliable results for bad-textured images. The interest point operators and image matching techniques that have traditionally used in Photogrammetry are unable to give good results for these applications. The algorithms that are used in Computer Vision (CV) community could instead assure good results, in terms of number of matched points. For this reason, a comparison analysis between the SIFT operator [1] and traditional photogrammetric feature extraction and matching techniques has been carried out. Many experimental tests on UAV aerial images and MMS terrestrial acquisitions with high geometric distortions (rotation, 3D viewpoint, scale) have been performed, in order to evaluate the reliability of SIFT for automatic homologous point extraction.