Our initial project covered a small area near Kansas City: The second, a much larger project near Minneapolis, Minn. A major lesson learned during the second project was the difficulties involved with doing water body edits on terrain matrices in areas of 10,000 lakes. This same project forced us to address the scheduling and coordination associated with photogrammetically compiling DEMs in parallel with producing orthophotos.
      Our third project was a very noteworthy one. When it was announced that the project encompassed Glacier National Park, everyone exclaimed what a scenic place that was and if we needed help on-site, they were willing to go. This was a foreshadowing of things to come. Yes, it is a very scenic area and the terrain is extremely rough with very few manmade features. The Grand Canyon may be the only area in the Continental United States over which it is more difficult to produce orthophotos.
      USGS supplied the photography, approximately 60 percent of the required control, and 73 percent of the DEMs. The total number of DOQQ to produce was 611.
      An immediate concern was acquiring the required ground photo identifiable GPS points in a timely manner at a reasonable cost. Our window of opportunity was from the time the delivery order was awarded June 20, 1994 until the time the area became snow bound at some time in early September. To complicate matters, the park prohibited the use of helicopters and restricted the use of vehicles to hard surface roads, of which there were very few. We decided to contract with a small, relatively local firm, Equinox Inc. of Ketchum, Idaho, which specializes in GPS surveys and was familiar with the park.
      The total project covered an area approximately equal in size to that covered by the state of Massachusetts, with Glacier National Park in the center of the project covering the majority of the area. GPS baselines varied in length from 10 to 100 kilometers making direct communications between crew members difficult. Occasionally, crew members were able to communicate via cellular phone, but in most instances this was impossible. Normally, the survey crews relied on strict planning schedules which were developed on a nightly basis. The required GPS points were collected often via horseback, hiking long distances, by boat in some cases, and others via wheeled vehicles. Figure 1 illustrates, from a ground perspective, the difficult terrain. With all of this, the points were collected just before the snow came.
      Now, all we had to worry about was that all of the points were good and that we could make all of the USGS supplied control points work. This was a major concern since the majority of the USGS points were described as lone trees and identified on 1960 vintage photographs. We did, however, have the advantage of having a few surplus USGS points, so if some were lost, we would still have adequate control. The process of viewing the relatively poor quality paper control photos stereoscopically and transferring the "lone tree" into the NAPP imagery print proved to be tedious work.
      Once the control was collected and identified, we began the work of measuring and computing for the analytical triangulation (AT). Figure 2 illustrates the control points used in the AT. The triangles are horizontal control points, those with dots in the center are new GPS points while those with triangles in the center are USGS points. The small green dots are vertical points which were extracted from the 1:24,000 USGS quad sheets. Flightlines are indicated within the large block circles. The triangulation processing was accomplished by a Melbourne, Fla. firm, JFK Inc., which is well known for its expertise in this technology.
      Measuring for the AT was difficult due to the very steep terrain and the numerous changes in camera systems and dates of photography. Even very short periods of seasonal change (less than one month) created major differences in the appearance of adjacent images. We expected major challenges in the computational phase of the AT and anticipated loosing at least 30 percent of the USGS supplied control points. With this in mind, we included all the available points even though we had classified some of them as very poor quality. We had to rework several of the points by returning to the original control photos and reworking the transfers. The original transfers from the very old control photos to the NAPP imagery, were accomplished using a B&L differential zoom stereoscope with backlighting on the paper print. During the rework process, we also had the NAPP stereomodel on the analytical plotter available which gave us better resolution in the receiving material.
      Due to JFK's and SAIC's persistent work, the results were much better than we had anticipated. If we ignore the points that we, along with USGS, initially rated as poor, only about 15 percent of the points were lost. The remaining points were adequately distributed to solidly control the project. The majority of the perimeter points were new points which had been established as part of the GPS survey.
      Once the AT was completed, work began on collecting the missing DEMs where they were missing and producing orthophotos in areas where DEMs were available. DEM production was accomplished via a manual approach in areas of very steep terrain and via automatic correlation techniques in some of the areas of more moderate terrain relief.
      Figure 2 is a perspective view derived from the DEM data compiled near Ohern Pass in the north central area of Glacier Park.
      As orthophoto production was initiated, it became apparent that due to the very steep terrain, image "smears" would occur on many of the images. To identify where these "smears" were going to occur, SAIC developed on algorithm and subsequent software which computed and plotted the areas where "smears" would occur. These plots were provided to USGS with each delivered orthophoto. After their reviews, USGS determined that some 22 percent of the orthophotos had "smears" which should be supplemented with alternate imagery. SAIC established a procedure which would (1) determine the ideal photograph to replace each smear area, (2) perform the orthorectification, (3) modify the image intensity, if required, and (4) mosaic the required imagery into the original orthophoto. The actual mosaicking of the two images was accomplished using interactive techniques also developed by SAIC. These techniques use the mathematical model for each of the photographs being processed thus guiding the operator along the cut lines. Surprisingly, we found that no metric adjustments were required between the images as they were being mosaicked.
      Figure 3 is a single quarter quad, the northern portion of which shows the Canadian border. A ground level equivalent of the cleared border area is shown in Figure 1.
      A second, very different, project was over Lawrence County, S.D. This relatively small project involved the production of 96 DOQQs over a rugged area for which no existing USGS control was available. It was determined that the most cost-effective, shortest schedule approach was to acquire the imagery with airborne GPS control. Proceeding with this approach, we selected Horizons Inc., of Rapid City, S.D. to acquire the photography and associated airborne GPS data. After several weather delays, Horizons Inc. was able to acquire the photography and the associated airborne GPS data.
      To validate the airborne GPS data and as a reference to remove any unresolved biases, four targeted ground control points (one in each corner of the project) were included in the GPS acquisition phase. Analytical triangulation was performed by JFK Inc. using the airborne GPS exposure stations as constraints. Our evaluation of the results indicated that the airborne GPS control was good to approximately .3m at the one sigma level.
      A major advantage of this approach was that no extra effort was required to identify and transfer ground control into the photography. Each photograph was controlled by its associated GPS data. Other advances were that the imagery was acquired specifically for the project area, thus eliminating seasonal changes or the numerous breaks in flightlines typically experienced when pre-existing NAPP photography is used. In addition, the acquired photography is very current as it is acquired and immediately processed. This entire project was completed in two months after the imagery was acquired.

About the Author:
Jerry Maupin is currently SAIC's manager for the United States Geological Survey (USGS) National Digital Orthophoto (NDOP) Program. He may be reached at 407-676-3102.

Back