Multiocular stereo image point matching and three-dimensional surface reconstruction in digital photogrammetry

The recent evolution of digital photogrammetry from classical photogrammetry has created a great deal of interest in applying numerous computerized techniques that were believed inapplicable in classical photogrammetry. This document presents research in which a different approach to performing digital terrain modeling and three-dimensional surface reconstruction has been demonstrated. The new approach is based on a multi-frame (multiocular stereo) technique rather than the commonly used binocular stereo techniques. This method has been investigated by researchers in the computer vision discipline since the mid 1980s but for different types of applications. In this thesis it has been investigated for its applicability in the discipline of digital photogrammetry. Four different paradigms have been suggested and applied on two sets of aerial photography. The accuracy of the resulting digital elevation model (DEM) data has been assessed by comparing against reference DEMs that were manually collected via analytical stereoplotters and against some individually surveyed ground control points. Various levels of accuracy were achieved by the different paradigms.