Registering randomly oriented scenes through moving object attributesRegistering randomly oriented scenes through moving object attributes Peer-Reviewed Journal Publication Δημοσίευση σε Περιοδικό με Κριτές 2015-10-132011enWe present a procedure to register motion imagery scenes captured from different viewpoints and under various orientations. Spatiotemporal (ST) trajectories of moving or changing objects and phenomena comprising a scene provide a series of interrelation attributes which are used to define the scene topology. Spatial proximity, temporal algebra, arbitrary directional measures, and shape complexity descriptors construct an abstract orientation-invariant space where conjugate pairs of trajectories between scenes are matched. This topology-centric matching procedure compensates for varying scale, occlusion, random error, and data uncertainty. Scene registration is performed through least square projective transformation. The procedure is quite precise, since the matched ST trajectories consist of populated sets of points following a common geometry compared with classic single point registration prone to errors or misinterpretation. Thus, while the matching procedure is abstract in nature, registration itself uses rigid geometry to provide accurate scene correspondence. Experimentation is quite encouraging even under extreme orientation, error, and occlusion conditions. http://creativecommons.org/licenses/by/4.0/International Journal of Geographical Information Science25111793-1807 Partsinevelos Panagiotis Παρτσινεβελος Παναγιωτης Impact Journals