Digital Orthophoto Production In the Desktop Environment 1 By Dr. Roy A. Welch and Thomas R. Jordan Digital orthophotos are proving suitable for a variety of mapping, GIS and environmental monitoring tasks. Increased demand is sparked by the availability of hardware and software suitable for use in a desktop mapping environment. The transition to softcopy photogrammetry has resulted in photogrammetric operations becoming an integral part of geographic information system (GIS) database construction tasks. In particular, digital orthophotos are now recognized as valuable base map coverages that can be used to identify ground control, update existing coverages for land use/cover and transportation routes and assess changes in the landscape caused by construction or environmental phenomena (Figure 1). Figure 1. The digital orthophoto base layer is ideal for updating vector data sets in a GIS. It is particularly significant that the evolution of computer technology has now reached the point where the demanding computational requirements associated with differential rectification and orthophoto generation can be met in a desktop environment using 486, Pentium, and Pentium Pro computers and softcopy photogrammetry computer programs. These possibilities are further enhanced by the availability of commercial scanning services, eliminating the need for many small photogrammetric businesses to invest in a costly scanner. In this article, some considerations for digital 1 Reprinted from Welch, R., and T.R. Jordan, 1996. Digital Orthophoto Production in a Desktop Environment, GIM Magazine, Vol. 10, No. 7 (July), pp. 26-27. 1
orthophoto production are briefly reviewed, and examples provided of digital orthophoto production with desktop computer systems in support of mapping and GIS applications. Digital Orthophoto Considerations The procedures for producing digital orthophotos are well established and were derived from techniques developed in the 1950 s and 1960 s for generating orthophotos. Even in today s digital age, orthophotos in hardcopy format are standard products produced from digital files. Required inputs to the digital process include aerial photographs, ground control points and a digital elevation model (DEM). In a softcopy environment, the DEM can be imported from government or commercial sources, or generated from the aerial photographs by interactive heighting or automated stereocorrelation techniques. Photographs to be used for orthophoto applications should be scanned at a resolution that takes into account the quality and scale of the original images, and the desired scale and information content of the orthophoto output product. For example, as the unaided human eye will resolve approximately 3 to 5 lpr/mm for low-contrast detail, the features on an output orthophoto will appear sharp and distinct if they are represented by 5 to 10 pixels/mm. Consequently, to ensure optimum visual quality the digital orthophoto should be developed so that the output pixel size is between 100 µm and 200 µm (or 250 to 125 dpi). Given an output pixel dimension requirement of approximately 100 to 200 µm, the appropriate scanning resolution can be determined. A good approximation of this value is obtained by dividing the output pixel dimension value by the enlargement factor required to bring the original photo to the output orthophoto scale. For example, if the original photo scale is 1:10,000 and the desired orthophoto scale is 1:2,500, a 4x enlargement is necessary. This, in turn, implies that a scanning resolution of approximately 25 to 50 µm will be required. It is also pertinent to consider the resolution of the original photograph. Most modern aerial cameras when used with black-and-white mapping films will produce aerial photographs with low-contrast resolution values of 30 to 50 lpr/mm. Such photographs will easily withstand enlargements of 6 to 10x. However, a scale differential of 10x between the original photo and the output orthophoto will dictate a scanning resolution of between 10 and 20 µm. At these scanning resolutions data files of 529 to 132 Mbytes will be created. Despite problems associated with file size and computational overhead, satisfactory orthophoto products can be generated on personal computers. Two brief examples are presented in the following section. Digital Orthophoto Production on Desktop Computers A test was conducted with aerial photographs used by the U.S. Geological Survey (USGS) to produce the new Digital Orthophoto Quarter Quads (DOQQs) that meet 1:12,000 scale map accuracy standards. The aerial photographs were scanned at a resolution of 50 µm, resulting in a ground pixel dimension of 2 m. According to the previously discussed requirement for an output pixel dimension of 200 µm or less, a scan 2
resolution of 50 µm should prove satisfactory for generating hardcopy orthophoto products at scales of 1:10,000 to 1:20,000. Ground control points obtained by aerotriangulation and Global Positioning System (GPS) survey procedures were then used in conjunction with the R-WEL, Inc. Desktop Mapping System (DMS) software package to derive orientation parameters for the photographs and to provide a basis for registering existing USGS DEMs to the photograph for the purpose of removing displacements due to terrain relief. The output orthophoto files (2710 x 3529 pixels) each covered 5.4 x 7.1 km and occupied approximately 10 Mbytes of disk space. Total processing time on a 90 MHz Pentium processor computer was less than 16 minutes, or about 1.7 minutes/mbyte. The resulting digital orthophotos demonstrated the feasibility of generating quality products from existing DEMs and aerial photographs in a desktop environment. In rural southeastern United States, recent floods have stimulated government agencies to request digital orthophoto products for use in estimating potential damage caused by earthen dam failure and flooding. In one such project, 1:10,000 scale photos were acquired along a 15 km stretch of river in northeast Georgia. Project specifications included the production of digital orthophotos and associated hardcopy output products at 1:2,400 scale. In order to meet this objective, the original film transparencies were submitted to an aerotriangulation process to densify control, and then scanned at 25 µm resolution. Stereo pairs were created in a softcopy photogrammetry environment and DEMs generated on a 90 MHz Pentium processor computer by automatic stereocorrelation procedures using the DMS software package (Figure 2). Elevation posts were calculated at a 6 m (20 ft) interval. On average, a total time of 26 minutes was required to generate a DEM by means of stereocorrelation, including the formation of the stereomodel. Production of orthophotos corresponding to the stereomodels (3948 x 6130 pixels) required an additional 41 minutes (or 1.7 minute/mbyte). Assessments of the end products produced RMSE values of + 0.6 m, consistent with accuracy specifications for the project (Figure 3). Costs of Digital Orthophotos The costs associated with producing digital orthophotos were recently evaluated by Survey Resources International of Houston, Texas who are routinely generating high resolution orthophotos on Pentium-processor personal computers using the DMS software package. They concluded that where digital orthophotos are produced with the intent of generating hardcopy output products at 3 to 4 times the original photo scale, the total cost, including flying aerial photos, surveying ground control, computing a DEM and generating the digital orthophoto can exceed US$1000 per frame. By comparison, if high resolution photos recorded with modern mapping cameras are scanned at resolutions of 15 to 20 µm, it is possible to generate digital orthophotos suitable for the production of hardcopy products at scales 8 to 10x those of the original photos. Consequently, fewer photographs are required and the cost of the aerial photo mission, ground control survey, aerotriangulation and orthophoto generation can be substantially reduced. 3
Figure 2. A DEM is required to perform the differential rectification and produce a digital orthophoto. This DEM was produced by automatic stereocorrelation. Figure 3. Digital orthophoto of a potential flood area with contours, planimetry, crosssection locations and coordinate grid superimposed. 4
Conclusion The rapid growth in the use of digital orthophotos for mapping and GIS applications is due to improvements in scanners, emergence of commercial scanning services, and the availability of reasonably priced software products that permit the use of raster images and vector data in a desktop mapping environment. When linked with DEMs and map files, digital orthophotos allow on-screen map compilation/update, as well as terrain visualization, environmental monitoring, and change detection. Overall, the utility of a traditional vector-based GIS can be greatly enhanced through the inclusion of digital orthophotos properly registered to the map coordinate system. Biographies of the Authors Dr. Roy A. Welch is Director of the Center for Remote Sensing and Mapping Science (CRMS) at The University of Georgia. He is past-president of the American Society for Photogrammetry and Remote Sensing (ASPRS) and is currently President of Commission IV, Mapping and Geographic Information Systems, of the International Society for Photogrammetry and Remote Sensing (ISPRS). Dr. Welch is also President of R-WEL, Inc., a company that markets the Desktop Mapping System (DMS) software package. Mr. Thomas R. Jordan is Associate Director for Geoinformatics at the Center for Remote Sensing and Mapping Science (CRMS). He is responsible for softcopy photogrammetry and digital image processing operations. Mr. Jordan is also Vicepresident of R-WEL, Inc. Center for Remote Sensing and Mapping Science The University of Georgia Athens, Georgia 30602-2503, USA 5