Automatic relative orientation of large scale imagery over urban areas using Modified Iterated Hough Transform

The automation of relative orientation (RO) has been the major focus of the photogrammetric research community in the last decade. Despite the reporte d progress, there is no reliable (robust) approach that can perform automat ic relative orientation (ARO) using large-scale imagery over urban areas. A reliable and general method for solving matching problems in various photo grammetric activities has been developed at The Ohio State University, This approach has been used to solve single photo resection using free-form lin ear features, surface matching and relative orientation. The approach estim ates the parameters of a mathematical model relating the entities of two da tasets when the correspondence of the involved entities is unknown, When ap plied to relative orientation, the coplanarity model is used to relate extr acted edge pixels and/or feature points from a stereo-pair. In its executio n, the relative orientation parameters are solved sequentially, using the c oplanarity model to evaluate all possible pairings of the input primitives and choosing the most probable solution. As a result of this technique, the matched entities that correspond to the parameter solution are implicitly determined. Experiments using real data conclude that this is a robust meth od for relative orientation for both urban and rural scenes. (C) 2001 Elsev ier Science B.V. All rights reserved.