Synergistic Fusion of GPS and Photogrammetrically Generated Elevation Models

Digital elevation models (DEMs) produced from photogrammetric data sources have long relied on the use of ground control points to give them scale and orientation. However, in areas such as coastlines, landslides, or glaciers, where identification of suitable natural features and pre-marking is difficult, the use of conventional ground control may be unfeasible. This paper reports on research that uses independently collected DEMs derived from kinematic GPS to orient surfaces produced by aerial photogrammetric methods, using a least-squares surface matching algorithm. During algorithm development, three stages of testing were carried out, using increasingly more complex datasets. Initially, simulated surfaces were used to validate the matching theory and program. Then, a DEM derived from conventional aerial photography was matched with a GPS model, highlighting the effectiveness of surface matching to recover systematic errors in datasets. Finally, surfaces derived from small format digital imagery were successfully fused with wireframe GPS surfaces, the high redundancy and automation potential creating an elegant and cheaper alternative to photocontrol. Introduction Traditionally, photogrammetric processing has relied on a set of independently measured ground control points (GCPs) to scale and orientate stereomodels by relating them to an object-space reference coordinate system (e.g., Wolf and Dewitt, 2000). For monitoring purposes, it is often necessary to create high-resolution DEMs at different epochs, but in the same coordinate system, in order to allow accurate change detection (Cooper, 1998). Often, monitored areas are least suited to natural GCP identification, due to the dynamic processes that require measurement, such as landslides (Brunsden and Chandler, 1996), glacial movement (Baltsavias et al., 2001), and coastal erosion (Adams and Chandler, 2002). This absolute orientation stage has long been the most inefficient part of the photogrammetric flowline, as well as having the least potential for automation (Schenk, 1999). Modern automated aerial triangulation methods employed in digital photogrammetry have meant that the amount of required photocontrol, formerly a minimum of three height and two plan points per stereopair (Rosenholm and Torlegård, 1988), has been greatly reduced. However, in areas such as coastlines or landslides where few “hard” natural or manmade features exist, the identification and acquisition of ground control is made more difficult and time consuming (Warner et al., 1996). A common solution to this problem is the use of prefabricated control markers, positioned and coordinated before a photographic mission, but increased expense and unpredictable weather conditions still make for inadequacy (Baltsavias et al., 2001). The use of kinematic Global Positioning System (GPS) equipment and inertial systems to determine the exposure station coordinates has further reduced the need for photocontrol (Wolf and Dewitt, 2000), but increased expense and complication, as well as calibration difficulties (Cramer et al., 2000) make it inappropriate for low-cost surveys. A potential alternative solution is to use an existing DEM to orientate a photogrammetric elevation model produced after only the relative stage of orientation, in effect using a control surface to provide absolute orientation, rather than using discrete points (Schenk, 1999). The use of a terrain surface means that the DEM can be collected independently of the photography, and is not reliant on the presence and identification of visible ground features. The problem is instead to register the unorientated photogrammetric elevation model to the absolute coordinate system of the existing ground DEM. Research has been carried out previously in the area of surface matching, with methods ranging in complexity. Ebner and Strunz (1988) and Rosenholm and Torlegaård (1988) developed the absolute orientation of large blocks of aerial imagery using coarse national-level DEMs, by minimizing the vertical differences between surfaces in a least-squares based adjustment. Pilgrim (1991), Karras and Petsa (1993), and most recently Mitchell and Chadwick (1999) used surface matching to detect deformations between sets of ultra-small-scale surfaces at different intervals for medical applications, where the use of control markers is undesirable and unethical. Schenk (1999) and Habib et al. (2001) used a variation of this surface matching technique, by minimizing the distances between normals of the two surfaces, with reference to absolute orientation of imagery and change detection. Further application areas of surface matching have been seen in comparisons between photogrammetric and lidar derived surfaces (Habib et al., 2000), as well as recovering shifts between strips of lidar data (Maas, 2000). The Iterative Closest Point (ICP) algorithm (Besl and Mckay, 1992) has been developed in the field of computer vision to fully match threedimensional surfaces. However, as noted by Mitchell and Chadwick (1999), its relative complexity may be unnecessary for conventional 21⁄2 D topographic surfaces. Consequently, least-squares minimization of vertical differences forms the basis of this research, part of an ongoing project developing an optimum solution for monitoring coastal erosion, by removing PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING Apr i l 2003 341 Photogrammetric Engineering & Remote Sensing Vol. 69, No. 4, April 2003, pp. 341–349. 0099-1112/03/6904–341$3.00/0 © 2003 American Society for Photogrammetry and Remote Sensing J.P. Mills and S.J. Buckley are with the School of Civil Engineering and Geosciences, University of Newcastle upon Tyne, Tyne and Wear NE1 7RU, United Kingdom (j.p.mills@ncl.