Pushbroom Aerial Imagery Scanner Mapping Accuracy Put To The Test

Pushbroom Aerial Imagery Scanner Mapping Accuracy Put To The Test
February 28, 2011 Mike Tully

Can the Leica ADS82 digital camera system meet the rigid horizontal and vertical positional accuracies required for 50-scale mapping? This was the question Aerial Services attempted to answer.  Because this “pushbroom” camera system acquires “strips” of imagery instead of “frames” of imagery* and is a superior camera system in many ways to others, clients sometimes question whether it can be used for mapping projects and meet the rigid positional accuracies required. This was tested using a 50-scale mapping project in Burlington, Iowa. The accuracy testing was conducted by Chris Berggren, one of several certified photogrammetrists and Geographic Information System Professionals (GISP) at Aerial Services.

Tested Project Specifications

BurlingtonIowaLayout.jpg

Color aerial photography was acquired for the City of Burlington, Iowa at 2000′ above ground level (agl). The control points used were both photo-identifiable control and targeted GPS monuments.

Three base stations (part of the Iowa Department of Transportation [IDOT] statewide real-time GPS network) within or near the project area were established. A total of 20 monumented and photo-identifiable ground control established by the City of Burlington was used as checkpoints to test the positional accuracies of the orthophotography and topographic mapping.<

The mapping products were assigned the Iowa State Plane, Iowa South projection and the NAD83 (1996) horizontal datum and the NAVD88 (Geoid 03, and U.S. Survey Feet) vertical datum. All AT solutions were performed using Leica XPro v4.3 and the same 20 unmeasured check points. All NSSDA statistics used here are reported for well-defined features at the 95% confidence interval.

Three-band (RGB) orthophotography (0.5′ GSD) was generated using an existing DEM acquired in 2006 at 1″=50′ scale with 25′ point spacing and supplemental breaklines. The DEM was tested to meet or exceed NSSDS standards of 1.67′ at a 95% confidence level.

How Accuracy Was Assessed

The objective of this study was to determine the achievable positional accuracies using the Leica ADS80-SH82 for a typical 50-scale mapping project using three different configurations of ground control: none, minimal, and vertical. In each test the same 20 photo identifiable check points were measured to test the positional accuracy. In all the tests, horizontal measurements of the orthophotography were performed using ArcMap 9.3. Vertical measurements were made from the stereomodels using Cardinal Systems VR v5.0.

The first run of aerotriangulation consisted of processing the project area with only the ABGPS/IMU data and no control. The second test consisted of using 5 full (XYZ) control points located in the four corners and the center of the mapping area. The third test consisted of using no horizontal control and only 36 vertical control points distributed throughout the project area.

“No Control” Test Results

Orthophotography for this test was produced using only the ABGPS/IMU data with no ground control. The points were measured and had an RMSEXY of 0.38′ at 95% confidence level (NSSDA = 0.68 feet). An NSSDA statistic of 1.90′ at 95% confidence level is the standard. This result clearly exceeds this and indicates that horizontal accuracies needed for 50-scale orthophotography are easily achieved using this camera system and no ground control.

The horizontal RMSEXY measured in the stereomodels was 0.39 feet (NSSDA = 0.68 feet) which compared closely to the horizontal accuracy measured in the orthophotography and clearly exceeds required ASPRS Class I horizontal accuracies of 0.5 feet. The resultant vertical RMSEZ was 0.50 feet (NSSDA = 0.97 feet). This does not meet the ASPRS (Class I) vertical accuracies of 0.33 feet.

It was not a surprise that vertical accuracies could not be achieved using no ground control. Using traditional aerial photography, as few as 160 horizontal and vertical control points would be needed to ensure 50-scale vertical accuracies were achieved. Further tests were designed to determine the minimal ground control needed to meet or exceed these positional accuracies.

“Minimal” & “Vertical” Control Test Results

A “minimal” test was designed using only five control (XYZ) points: one in each corner and one in the center of the mapping area. The horizontal RMSEXY was 0.29 feet. This still easily exceeds the ASPRS Class I accuracies and demonstrates that including just 5 ground control points in the aerotriangulation improves the horizontal accuracy by about 25%. The vertical RMSEZ was 0.51 feet (NSSDA vertical accuracy of 0.99 feet) but still did not meet the ASPRS Class I accuracies for 50-scale mapping (0.33 feet).

The “vertical” test used 38 vertical control points in the aerotriangulation. The horizontal RMSEXY was 0.39 feet and the vertical RMSEZ improved to 0.21 feet (NSSDA vertical accuracy of 0.41 feet). The addition of 38 vertical control points throughout the project area succeeded in improving vertical accuracies that meet or exceed ASPRS Class I vertical specifications (RMSEZ = 0.33 feet).

Summary

In conclusion, horizontal accuracies required to produce highly accurate orthophotos were easily achieved using only ABGPS/IMU data and no ground control. In practice, some ground control is always used to ensure that the ABGPS/IMU components are performing within specifications. But these tests confirm that projects requiring only orthophotography the geometry of the Leica ADS80-SH82 camera system is capable of creating orthophotography without requiring ground control that meets ASPRS Class I accuracies.

Vertical accuracies are generally much harder to meet. Typically, a minimum of 160 ground control points are needed to ensure the mapping products can meet the more difficult ASPRS class I vertical accuracy requirements. Surveying photo-identifiable control points may cost upwards of $100 per point. These results indicate a minimal configuration of vertical control in the corners and center of the block are insufficient to meet vertical accuracy specifications. However, using 38 vertical control points distributed throughout the project area were able to meet accuracy specifications.  This supports the practice of using significantly fewer vertical control points for mapping when using the ADS80-SH82.  In this test, using only 38 points enabled a cost savings to the client of approximately $12,000.

The Leica ADS80-SH82 and XPro are excellent tools for producing orthophotography and topographic and planimetric mapping. The cost advantages of using none to minimal ground control and still easily achieve horizontal map accuracies are important. However, when this camera system is used for mapping projects that require high vertical accuracies additional ground control must be planned, but the project will require far less than using traditional film cameras and aerotriangulation.

*The ADS80-SH82 acquires “strips” but “frames” are produced and delivered to clients as needed.