| SAR Supplement: Digital Terrain Models from RADARSAT
Recent simulation work results demonstrate an exciting capability for DTM extraction and topographic mapping using RADARSAT data—an application previously considered to have low potential.
By M.D. Thompson and J.B. Mercer A telecommunications company in Europe wins a multi-million dollar contract in Indonesia to plan, design and install a cellular telephone tower network. Plans have to be in place within six months, so that the work on the route and initial tower siting can begin according to the stringent milestone requirements of the contract. In order for the engineers and planners to begin their work with the local utility company, information over a broad area related to land use, land cover and especially the terrain relief must be acquired for input to their geographic information system. The problem discovered by the chief engineer is that the information just does not exist. How to solve the problem within the tight timeframe and budget available? The digital terrain modelling capability of Canada's RADARSAT will come to the rescue! Demonstrating RADARSAT Mapping Potential
Recently completed development work by Intermap Technologies (a division of IITC Holdings) in Calgary, Alberta, Canada has demonstrated the potential for ortho-rectification and extraction of digital terrain models (DTMs) from stereo RADARSAT data. The RADARSAT synthetic aperture radar (SAR) satellite, successfully launched by the Canadian Space Agency on Nov. 4, 1995, is expected to provide high quality, multi-mode, C-band SAR data worldwide over the next five years, following a several-month commissioning period. Its ability to acquire image data in darkness and through cloud and haze will permit acquisition of mapping and monitoring information in such difficult areas as the tropical and arctic regions. For this reason, there is great interest within the geo-information and mapping community as to its application for digital terrain modelling and topographic mapping.
The objective of the work carried out to date was to evaluate the potential of RADARSAT for extraction of digital terrain information, using ERS-1 radar satellite data to simulate RADARSAT topographic mapping parameters and derived products. The results of the study exceeded expectations, and will therefore be followed by further development and commercial implementation in 1995-96 using RADARSAT data. Topographic Mapping from RADARSAT: Simulating the Real Thing with ERS-1
Starting with STARMAP
The work first involved modification of Intera's proprietary STARMAP process. This process, used to derive topographic mapping products from stereo STAR-1 airborne SAR data (STAR-1 is Intera's terrain mapping airborne SAR system), was extensively modified for use with ERS-1 radar satellite data. Then, as ERS-1 standard data swaths do not have adequate overlap for stereo mapping purposes due to orbital constraints, special data sets had to be acquired. Early in the ERS-1 mission, the satellite viewing geometry was changed briefly and data were acquired during this Roll/Tilt Mode (RTM) of operation. When combined with data from the normal viewing mode, good stereo models can be obtained. Two sets of RTM data acquired over Canada were used in the simulation study-one in an area of low-to-moderate relief near Ottawa, Ontario, and one in an area of low-to-high relief near Mission, British Columbia.
STARMAP has the capability to derive elevation information from the stereo data, with output products including DTMs, ortho-rectified image maps (ORIs), and contour maps derived from the DTMs. Typical hardcopy output scales are 1:50,000 to 1:100,000, and contour spacing is about 50m, with a vertical accuracy in the order of 15-30m RMS. To date, over 300,000 square km have been mapped in commercial programs using STARMAP, primarily in cloud-covered tropical regions. The STARMAP process has been described extensively elsewhere (e.g., Mercer & Griffiths 1993) and is briefly summarized here as background.
The STARMAP process consists of three major modules. The first, Stereo Extraction, uses the methods of photogrammetry. Coordinates of thousands of points used to describe the terrain elevation are acquired from overlapping stereo pairs of the digital radar imagery.
In the second, Radargrammetry, the radar geometry equations are solved for each of the acquired points and are now tied to a desired mapping frame of reference. These points are used for creation of the DTM and ortho-rectification of the images. The latter removes terrain and sensor-related geometric distortions from the image and allows it to be treated like an image map.
In the third module, QA-Edit, the derived contours are overlaid on the ortho-rectified image and the contours and DTM edited for internal consistency. Ground control data may be used at this stage. Substantial software changes were required to enable STARMAP to utilize the SAR satellite stereo data, and further modifications will be needed for the RADARSAT application. Adding the Data
ERS-1 data were acquired through the Announcement of Opportunity Program of the European Space Agency and through RADARSAT International (RSI) of Richmond, B.C., the distributor of RADARSAT data. Table 1 shows the data used for the two Canadian study areas. The Mission area ranges from low to high relief, rising from the level agricultural land in the Fraser River valley to the forested Pacific Coast mountains in the north, with total scene relief of 1850m. The Ottawa area has low relief in the Ottawa River valley, and moderate relief in the Gatineau Hills to the north. The land cover varies from mixed agriculture, urban and wooded areas in the valley, to treed hills in the north. Analyzing the Data
Using these data and the modified STARMAP software, DTMs and ortho-rectified images were created for both areas. Ground Control Points (GCPs) to assess the horizontal error were acquired by digitizing recognizable features at a range of elevations from 1:50,000 scale National Topographic Survey maps. In order to assess the quality of the ERS-1 DTMs, government DTMs were acquired for the two areas to be used as "truth." B.C. government Terrain Resource Information Management (TRIM) DTMs (horizontal uncertainty of 10m, vertical uncertainty of 5m, both at 90 percent confidence interval) were used for Mission; Canadian government Digital Terrain Elevation Data (DTED) Level 1 data were used for Ottawa. Figure 3 shows the unrectified and ortho-rectified Roll/Tilt Mode ERS-1 images for a portion (about 10 percent, or 15 x 15 km) of the Mission area. The results of the ortho-rectification process, particularly in the location of the features of more significant elevation, is easily noted. The corresponding TRIM DTM and the ERS-1 DTM for the same area were displayed as UTM coordinates with a 50 x 50m grid, with elevations shown between 0 and 1850m.
A range of other products were generated as well, including contour plots (Figure 2) and color-coded perspective views (Figure 1) for each area. The contour plot in Figure 2 was generated using the ERS-1 DTM, and shows 50m contours ranging in elevation from 50 to 1250m. Figure 1, a three-dimensional perspective view of Mission was derived from the ERS-1 DTM and the ortho-rectified image, the image was draped over the DTM, and colour-coded with respect to elevation. Surpassing Expectations with Simulation Results
The results of the analysis were better than originally anticipated. Visual comparison of the government and ERS-1 DTMs, and calculation of difference surfaces for both areas (not shown) showed very good correlation between the two. Table 2 shows the summarized results from both areas, separated into relief types.
For Ottawa, the ERS-1 DTM corresponded very well to the government DTED, likely due to the low relief in this area. For Mission, the qualitative correlation was also very strong, but detail was missing in the steep terrain due to the extreme viewing geometry of ERS-1. This causes excessive terrain displacement, leading to layover and obscuration of the valley floors. While this is a significant limitation on the use of ERS-1 data for this application, RADARSAT will be a better system for DTM extraction. The larger incidence angles available from RADARSAT will help to minimize layover and thus the problems experienced by ERS-1 in steep terrain. Thus the RADARSAT beam diversity will markedly improve the results, but so also will the use of both ascending and descending orbits to achieve (approximately) opposite viewing conditions to fill in the data gaps.
Evaluation of the horizontal uncertainty for the ortho-rectified images created from ERS-1 showed that 30-40m RMS uncertainties are achievable, at least in the low to moderate terrain areas. It should be possible to achieve this same level with RADARSAT, and likely better levels with the fine resolution mode. RADARSAT will not have the same orbital precision as ERS-1, and thus it will be necessary to correct the ephemeris information using ground control.
Overall, however, the results from RADARSAT are expected to exceed DTED1 Level 1 standards (18m RMS vertical accuracy at 3 second postings) in areas of low to moderate relief, and to nearly meet those standards in areas of high relief. Given the difference in RADARSAT parameters from ERS-1 parameters described above, it remains to be determined how much improvement will be realized in steep terrain. Who Needs DTMs? Almost Everybody!
Topographic information in digital form is increasingly required by a wide range of organizations worldwide. Stereo digital imagery acquired from a stable satellite platform such as RADARSAT should permit lowered production costs, since the time required to extract DTM information and ortho-rectify a base image will be significantly reduced. This digital terrain information is required today for ortho-rectification of image data, for use as data layers in geographic information systems (GIS), for derivation of terrain-related information (slope, aspect, relief), and as the basis of terrain simulations (viewsheds, tower siting by line-of-sight, flight simulations). Uses of this digital information span a broad range of requirements; users include topographic mapping organizations, urban planners, thematic mappers, telecommunications tower siting planners, forest logging operators, environmental managers, watershed planners, flight simulation trainers, corridor route planners, engineers, land developers, and many others. Recent references in EOM to the real-world uses of digital terrain models include their use for environmental impact studies. One application involved evaluation of impact downstream from a dam site on the Colorado River (August 1995, pp. 18-21); another was their use in conjunction with SPOT satellite data to determine transport direction of contaminants in surface and groundwater in Russia (August 1995, pp. 26-28). The use of digital elevation data in Desert Storm in 1990, for flight simulation training and for battlefield planning, has been reported in many publications and interviews (November 1994, pp. 22-24). Organizations who are immediately interested in the use of DTMs include government mapping agencies, the telecommunications industry, military organizations, resource management agencies, the resource industry and environmental groups.
In the future, demand will be even greater as "non-traditional" users realize the potential of satellite DTMs for their businesses, and when positioning and terrain-related information become a part of almost daily life-for vehicle routing and mileage optimization for the trucking industry, for example; for drug interdiction, for tourism, for military intelligence, for education and for many other activities. One example is provided in a recent EOM article (July 1995, pp. 34-37) which graphically described one way in which DTMs are becoming part of our everyday lives, in this case through television weather reporting. DTMs and remote sensing images are used to provide weather graphics, animated to display not only weather but natural disasters, wars, and even sporting events. And while the obvious interest is in daily cloud cover, they also need to create base scenes with "zero cloud cover, a gremlin with which all remote sensing professionals are well acquainted," and today do so with time-consuming and artificial cosmetic "painting" processes. Enter RADARSAT!-providing not only the terrain modelling information for the 3-D perspectives, but the ortho-rectified image bases-anywhere in the world! And, as the article pointed out about the future, "three-dimensional dynamic perspectives will be used to enhance and illustrate news other than weather, and will be used more and more as part of the entertainment industry, providing backgrounds and special effects for television programs and motion pictures." Similar expansion of interest is anticipated in many applications. Why Acquire DTMs from RADARSAT?
The DTMs and ortho-rectified images to be acquired from RADARSAT stereo data have numerous advantages for users:
• wide area coverage
• low cost per square kilometer (compared to air borne sources)
• reliable coverage from cloud- and haze-covered regions
• adequate detail and accuracy for many applications
• rapid turnaround and DTM product generation
• ortho-rectified image base provided for other mapping and GIS uses Conclusion
The results of the RADARSAT simulation work demonstrate an exciting capability for DTM extraction and topographic mapping using RADARSAT-an application which was previously considered to have low potential. Thus, mappers, planners and resource managers with digital terrain information requirements can look forward to the availability of reliable high quality, low cost products and/or product generation software over broad areas and areas previously difficult to access due to cloud or haze. Both DTM services and software products will be available from Intermap Technologies in 1996, following further development using RADARSAT data. Acknowledgments
This work was partially funded by the Radar Data Development Program (RDDP) of the Canadian government (Canada Centre for Remote Sensing), with data contributions by RADARSAT International (RSI), the British Columbia government, the Canada Centre for Remote Sensing (Dr. B. Guindon), and the European Space Agency.
1DTED is the standard of the U.S. military mapping organization, Defense Mapping Agency (DMA), and as such is a de facto standard worldwide. About the Authors:
Diane Thompson is the manager of business development for Intermap Technologies (a division of IITC Holdings Ltd., formerly Intera Mapping Services in Calgary, Alberta. Bryan Mercer is the manager of technology development for the same organization. Both may be reached at 403-266-0900 in Calgary. Back |
