PUBLIC WORKS
Behind the Scenes
Public works GIS data technologies lend crucial support to local governments
By Peggy Ammerman

Digital ortho images derived from aerial photography are becoming the preferred land-base medium today for many AM/FM/GIS applications. Unlike labor-intensive traditional methods of land-base production, which typically rely on manual conversion of geographic features and topology, digital ortho imagery-base maps are cost-effective and highly accurate. In addition to base map functions, digital ortho imagery and the digital terrain model (DTM), which is used in orthorectification, can also be called on to verify the completeness and correctness of vector features in a GIS and perform other analytical applications. Today, the usefulness of digital ortho imagery in various GIS functions and its value as a primary component of a GIS only continues to increase as technology advances.
    Even with substantial advancements though, certain limitations have followed digital ortho imagery up the chain of technological evolution. While orthorectification adjusts scanned aerial photography and removes displacement caused by camera tilt and topographic relief, the limitations of conventional digital ortho imagery are most evident in urban areas with structures such as bridges and tall buildings. "Conventional orthorectification simply does not treat displacement shown in bridges and tall buildings," says Anthony Thorpe, vice president of research and development for ASI Technologies based in Colorado Springs, Colorado. ASIT is a customer service center of Analytical Surveys, Inc. (ASI); a leading provider of a full range of computerized mapping services.
    Conventional digital ortho imagery shows a distorted scene with buildings that lean and bridges that are warped in appearance. In addition, when conventional digital ortho imagery is overlaid with vector data layers such as building footprints, edge-of-pavement and other features, the registration is mismatched, offering those viewing the imagery a skewed picture of the scene. In effect, limitations such as these can sharply undermine the usefulness and value of digital ortho imagery.
    To solve the problem of relief displacement, ASIT under the direction of Thorpe developed a special, fully automated process called METRO that rectifies building lean and similar situations through the production of "true ortho" imagery. According to the modeling solution employed, the true ortho process increases accuracy and reduces or eliminates building lean and other distortions. As a result, bridges, buildings, and other objects are moved back to their true locations and then imagery is used to fill in "blind spots" that are left behind.
    In explaining the objective behind the development of METRO, which stands for Method for the Elimination of Tilt and Relief displacement in Orthophotography, Thorpe says ASIT's true ortho solution is aimed at "maintaining the value of digital ortho imagery in a downtown area." ASIT has been working on the METRO true ortho solution since October 1997, and recently put METRO to the test in producing true ortho imagery for a small portion of New York City that is densely populated with closely spaced tall buildings. ASIT's METRO combines about 24 software programs and a 5-step process. What sets METRO apart from other true ortho solutions, says Thorpe, is a fully automated process. The majority of true ortho solutions introduced in recent years have been based on manual processes.
    In understanding the distinction between conventional and true digital ortho imagery, it's important to first understand the fundamentals of digital ortho imagery and then the distinctions between conventional and true ortho production methods. The making of a conventional digital ortho image requires three elements: a scanned image of an aerial photograph, data on the position and orientation of the aerial camera at the time of exposure, and a DTM that shapes the corresponding terrain. Usually, the camera's orientation is obtained from aerial triangulation which, along with the DTM, gives the geometrical equation required to orthorectify the image. A pixel's color is determined by interpolating an elevation for the pixel and then projecting it onto a scanned image. Although the production of true ortho imagery uses many of the elements of the conventional methodology, there are two key distinctions of the true ortho process: the DTM must model buildings and bridges accurately, and the final ortho image is pieced together from hundreds of irregularly shaped image patches to produce a seamless pattern.
    The first key that distinguishes the METRO process according to Thorpe is "modeling buildings and bridges in the DTM." Thorpe notes that, "Objects like bridges and buildings must be captured as part of the DTM and must be accurately modeled as well." An inherent characteristic of aerial photography is relief displacement caused by elevation differences in the terrain being photographed. The process of orthorectification corrects these distortions using a DTM. Thorpe emphasizes that orthorectification can only "correct relief displacement in buildings and bridges when these features are modeled accurately in the DTM."
    Capturing buildings and bridges to the level of detail required for true ortho imagery is a very complex, time-intensive task. To illustrate, Thorpe points to a test project using digital ortho imagery for a .5 square mile area of Manhattan Island in New York City. The area contains many tall buildings with elaborate architecture of various vintages. These buildings characteristically display and require modeling of multi-layered effects and a variety of complex surface details such as flat and sloped roofs, terraces, exterior housings for elevator shafts and HVAC units, domes, pyramids, and spires, among others. In addition, stereo coverage of street level features in this area was marginal due to the extremely tall buildings and narrow streets. Thorpe says despite these complexities, ASIT's stereo operators were able to capture most of these features as part of the true ortho DTM.
    Considering the precision required and time involved in capturing such details, "The cost of capturing a detailed DTM to this extent is probably the biggest drawback of true ortho imagery," according to Thorpe. However, he notes, "Downtown areas containing tall buildings are usually a relatively small portion of the entire city," therefore expenditures associated with true ortho production constitute just a small portion of the overall costs. In support of the cost justification for true ortho imagery, Thorpe adds that "these areas tend to be well established and the DTM will probably undergo little change over time and retain its value and usefulness over the long term."
    Capturing buildings and bridges as part of the DTM is only part of the true ortho solution. These features must be accurately modeled as well, Thorpe emphasizes. Conventional orthorectification typically uses a DTM with regularly spaced points in the X and Y coordinate directions. He explains, "This regular grid DTM methodology is insufficient for producing true ortho imagery because the point spacing is too large to model small features."
    To compensate for the limitations of the regular grid method, METRO uses a triangular irregular network. The TIN allows modeling of the ground surface, buildings, and bridges with a series of flat, triangular surfaces. In cases of significant height variations in the model, the triangles are small, and vice versa. "In other words," Thorpe says, "the TIN is adaptive and can model buildings and bridges to the precision required for producing true orthos."
    The second key to ASIT's METRO process is sophisticated mosaicking that combines imagery from many perspectives. Thorpe says mosaicking is also "the most technically challenging part of the process." Rectification moves buildings and bridges into their proper locations. As a result, holes or blind spots are left behind. Thorpe points out that in the case of the Manhattan test project, there was considerable overlap between photographs. He says, "Imagery hidden in one aerial photograph was often visible in another." This imagery was used to fill in blind spots and create an image with the most complete coverage possible.
    The mosaicking used in ASIT's METRO process is fully automated. Thorpe says, "The best imagery is chosen for every final orthophoto pixel." Thorpe defines "best" as a combination of two factors. The first is proximity to the nadir point in order to minimize distortion of trees and other such features that have not been modeled. The second factor is the relative orientation of the ground surface to each image. In selecting pixels, Thorpe says "pixels are chosen from imagery that has the best view of each triangular facet in the TIN." This helps avoid "smears" or areas of imagery where pixels are replicated and appear stretched or smeared due to severe oversampling, Thorpe adds.
    The METRO process is also beneficial for digital ortho imagery of steep terrain. METRO selects imagery from the aerial photograph with the best view of the terrain. For example, Thorpe says, "When sloping ground faces toward image A but away from image B, imagery from image A is selected for the ortho imagery of that area." This helps eliminate smears says Thorpe.
    ASIT's "true ortho" DTM can also be combined with imagery to create a virtual reality model that ASIT calls a METROScape. The 3D Analyst extension to ArcView can be used to drape the "true ortho" imagery on the DTM in order to produce a perspective view. Thorpe says, "On a high-end PC with a suitable graphics card, the model can be rotated and viewed from any angle, practically in real time."

About the Author:
Peggy Ammerman is a writer with the Indianapolis, Indiana office of Analytical Surveys, Inc. (ASI) and specializes in GIS and geospatial information technology issues. She can be reached at 317-634-1000, ext. 285 or by e-mail at [email protected].

Back