Rick Hanson

The Truth About Airborne GPS

Though airborne GPS has been widely used since1993, there is still a wide diversity of opinions as to how much ground control is required and where the control should be placed. In the spring of 1999, Merrick & Company entered into a contract with the County of Boulder, Colorado to provide aerial photography and digital orthophotography of the eastern two-thirds of the county. The project would be controlled by airborne assisted analytical aerotriangulation with approximately 65 widely dispersed, targeted control points. This paper examines and compares the results of five sets of aerial triangulation data generated from various control configurations.

INTRODUCTION


Background


Because of the expense of field control for a photogrammetric mapping project, photogrammetrists have always been interested in ways to reduce the amount of field work involved with producing a reliable product. Conventional control methods required no less than 4 field points for each stereo model. When analytical aerial triangulation (AT) came into wide use in the late �60�s the number of required field points was reduced to approximately 1 point for every 3 stereo models. With recent advancements in GPS technology an entire project can be controlled by as few as three field points. Some even say that zero control points are required, but a photogrammetrist will always want some way of performing a qualitative check of his or her work in order to guarantee the accuracy of the final product. This means that sooner or later some field measurements will be required so they might as well be done in advance of the aerial triangulation. The uncertainty comes with deciding how many control points to use, and where they should be placed.

Overview


Since the Boulder County scope called for 65 control points, approximately 1 point for every 2 stereo models, the photogrammetrists at Merrick took advantage of the situation by performing a number of adjustments using various control configurations holding only five points for the AT adjustment and using the remaining 60 points for checks.

Aerial Photography


The project located near the eastern foothills of the Rocky Mountains covers an area of 400 square miles. Black and white aerial photography was flown at a height of 12,000 feet above mean terrain with 60% endlap and 15% sidelap. A GPS receiver in the aircraft collected satellite data at 1 second intervals along with an event mark generated by electronic impulse each time the camera exposed a photograph. A GPS ground station was positioned at Jefferson County Airport, ten to fifteen miles from the project area. Seven north-south flight lines covered the project area with a total of 160 exposures. All of the photography was flown on the same day under optimum conditions using a Wild RC-30 precision mapping camera with forward motion compensation (FMC). Field Control Merrick measured 65 pre-targeted points by post-process kinematic method using two base stations and collecting satellite data at 3 second intervals for a minimum of 15 seconds so that no less than 3 epochs were recorded for each station. The expected accuracy for this exercise was +/- 0.1� horizontally and +/- 0.5� vertically, accuracy�s more than adequate for the production of 1�=400� digital orthophotos to an accuracy of +/-3.4 feet for 90% of the points measured. Aerotriangulation Aerial photographs were pugged with 3 pass points in each photo and flight lines were joined with 1 tie point for each adjacent photograph. Then control targets and analytical points were measured on Zeiss P1 analytical stereo plotters. Measurements were processed using Eriotech�s ALBANY software. With due consideration for aircraft GPS antenna offset and exposure impulse time delay, post processed airborne GPS data was combined with field control, flight and camera calibration data to produce final AT coordinates.
The first adjustment was performed holding all 65 field points as control. Results of the first adjustment are shown in figure 1.

Figure 1

A second adjustment was performed by holding 5 field points, 1 near each corner of the project and 1 near the center of the project area. The remaining 60 points were used as check points only. This is seems to be the preferred configuration of control for an airborne GPS project. The RMSE�s of the check points in the second adjustments are shown in figure 2. Figure 2

At first glance, the control placement for the next three scenarios would appear to be insufficient to support accurate aerial triangulation but the results indicate otherwise. The third AT adjustment shows the RMSE�s of the 60 check points as controlled by 5 points closely grouped near the center of the project area. Result of the third adjustment are shown in figure 3. Figure 3
The fourth adjustment shows the RMSE�s of the 60 check points controlled by 5 points aligned in a north-south direction ranging the length of the project area. Results of the fourth adjustment are shown in figure 4. Figure 4
The fifth AT adjustment, shown in figure 5, shows the RMSE�s of the 60 check points as controlled by 5 points arranged in an east-west direction ranging the width of the project area.
Figure 5

Summary of results


 
Xft Yft Z ft.
Adjustment 1 RMSE: .495 .541 .432
Adjustment 2 RMSE: .838 .678 .629
Adjustment 3 RMSE: .899 .701 .596
Adjustment 4 RMSE: .880 .696 .617
Adjustment 5 RMSE: .857 .683 .596

CONCLUSION

Though the first aerial triangulation adjustment using all of the points as control yields the best results, it is not practical to develop 1 field control point for every two stereo models as the abundance of control obviates the need for airborne GPS. The RMSE�s of scenarios 2,3,4 and 5 show more than sufficient accuracy�s on the check points to support 1�=200� mapping with a 5� contour interval, more than twice as accurate as Boulder County�s project scope called for. If we accept that only five field points are required to control a given block of photography, and this does appear to be the case, the question of control placement seems to be answered by the negligible differences in the results of adjustments 2,3,4 and 5. When significant figures are taken into account, the placement of the field control matters very little in a practical production environment.
As to the question of the number of control points that should be used in a given project, every project is different and often the number of points is controlled by what the budget will bear. Additionally, if we think of the check points as confidence points the number may vary according to the confidence level of the photogrammetrist to produce accurate data over a given area at a given altitude. When all else fails, the author suggests not less than 10% of the number of exposures it takes to cover the project area at the prescribed photo scale, using as many of the points a possible for check points.

Acknowledgments


I must thank Paul Tessar, Boulder Counties GIS Director, for permitting me to use his project as a topic for my paper. Aero-graphics, Inc. of Salt Lake City did a very professional job of the aerial photography and airborne GPS data collection and post processing. Brian Raber of Merrick & Company in Aurora, Colorado for encouraging me to write this paper.

Rick Hanson
Project Manager and Certified Photogrammetrist
Merrick & Company, Aurora, Colorado