Earth Imaging Journal

Home

Market Guide 2010

Free Subscription

Industry Resources

Industry Directory

Missions/Launches


By Qassim Abduallah, chief scientist, Fugro EarthData (www.fugroearthdata.com), Frederick, Md.
	Geospatial customers are embracing airborne light detection and ranging (LiDAR) mapping technology like never before. Recent data acquisition improvements and new processing techniques are expanding the use of LiDAR beyond traditional applications such as base orthoimagery, floodplain mapping and urban modeling applications. Today, users are requesting LiDAR to support activities requiring much tighter data specifications, most commonly 1-foot contour models. While some service providers are delivering data to meet these specifications, others question whether this type of deliverable is pushing LiDAR technology beyond its practical limits. With little available research published on the subject, Fugro EarthData recently conducted its own study to evaluate the ability of medium-density LiDAR data to support 1-foot contour modeling. In addition to validating contour model accuracy, the study also examined the role of breaklines in the modeling process, assessing the impact of several variables on production throughput and contour model quality. Designing a Medium-Density Test Model The beauty of LiDAR is in some ways its downfall: the capacity for massive amounts of high-accuracy data. Although accurate 1-foot contour modeling is possible using high-density LiDAR datasets, acquiring sub-meter post spacing LiDAR data can be costly, even for the most advanced LiDAR mapping systems. To collect such high-density data, an aircraft must fly low and slow, which in turn narrows the field of view, limits access by efficient/fast aircraft, and increases the number of flight lines required to cover the project area. Additionally, the large volumes of data resulting from such dense collections challenge current data management and processing systems. All of these factors lead to increased project costs. For these reasons, most LiDAR mapping projects are collected at medium point densities. Ranging between 1- and 2-meter post spacing, these datasets still are considered relatively dense compared with traditional photogrammetric modeling.

	To meet the American Society for Photogrammetry and Remote Sensing (ASPRS) Class I requirement for 1-foot contours, a project's vertical root mean square error (RMSE) must fall under 10 centimeters. For medium-density LiDAR data, that specification requires careful attention to the LiDAR system's error budget, which is controlled largely by position and orientation subsystems. As a result, access to reliable airborne Global Positioning System (GPS) and drift-free inertial data are critical. Strategies for lowering the error budget include limiting the distances from base stations to the aircraft, flying during times of strong satellite constellation and incorporating relatively shorter flight lines. For the contour modeling study, the Fugro EarthData project team did just that. Using an ALS50-II sensor from Leica Geosystems (www.leica-geosystems.com) with multiple pulse in-the-air technology, the company acquired LiDAR data from an altitude of 2,200 meters for 1.4-meter post spacing. In addition to airborne LiDAR acquisition, the study also involved collecting high-resolution digital imagery to create baseline 1-foot contours. Airborne imagery was acquired at a ground sample distance of 7.5 meters (3 inches) using a Leica ADS40-SH52 imaging sensor. From the source imagery, technicians performed aerial triangulation using a robust network of 33 ground control points. After generating stereo pairs, breaklines and mass points were collected and translated into a triangulated irregular network (TIN), which was used to extract 1-foot contours that meet ASPRS Class I accuracy standards.


	Modeling the Data, Analyzing the Results For LiDAR contour modeling, the project team employed lidargrammetry techniques, creating LiDAR stereo pairs from bare-earth, reflective-surface and intensity data. Once LiDAR intensity stereo pairs were established, breaklines were compiled at similar locations to those typically collected in a standard photogrammetric project, including road edges and centerlines, ditches, retaining walls, culverts, railroads, etc. From this information, Fugro EarthData generated three different sets of contour models: 1. 1-foot contours from bare-earth LiDAR point data only, 2. 1-foot contours from bare-earth LiDAR supplemented with LiDAR-derived breaklines and 3. 1-foot contours from bare-earth LiDAR supplemented with photogrammetric breaklines. Initially, it appeared that producing contours solely from bare-earth LiDAR point data was less than successful. Lacking aesthetics and cartographic quality, unwanted details made the dataset look noisy. To address this issue, the team used smart algorithms to filter the LiDAR data based on key points where elevation was necessary to accurately represent the terrain; the algorithms then applied the elevation of the desired contour as another filter criterion. Once complete, the modeled surface provided an acceptable visual and cartographic contour quality (Figure 1), although it was clear the data would benefit from carefully placed breaklines in areas of significant elevation change. The contours developed with the aid of LiDAR-derived breaklines revealed two interesting issues. First, LiDAR-derived breaklines may not be accurate enough on flat terrain. Second, the density of contours compiled from LiDAR-derived breaklines is much less than in the photogrammetric baseline model (Figure 2). Fugro EarthData determined that a lack of sharp definitions and clarity in the black-and-white LiDAR-intensity imagery—but found within digital imagery—caused technicians to compile breaklines with abnormal vertical and horizontal measurement errors. To improve the results of this process in areas of flat terrain, operators should consider superimposing LiDAR points over the stereo pairs for better height reference opportunities. The issue also may be resolved by removing breakline elevations over these areas and then reassigning new elevation for all the vertices of the breaklines using the bare-earth TIN surface. As for the lack of breakline density, the discrepancy may be attributed to the relative coarseness of the 1.4-meter post spacing LiDAR data compared to the 7.5-centimeter resolution digital imagery.


	The quality of contours produced from bare-earth LiDAR supplemented with photogrammetric breaklines was optimal; the data aligned well. The TIN surfaces (Figure 3) show elevation differences between the two datasets that are within the ± 5- to 7-centimeter range in open ground. On the downside, combining LiDAR point data with photogrammetric breaklines is an expensive process. To mirror the results performed in this study, a project would need two airborne missions—one to acquire LiDAR and one to acquire imagery due to the different flying altitudes and speeds required for each type of acquisition. It is possible to acquire LiDAR and imagery simultaneously, but most dual-capability systems provide fairly low-end imaging quality designed to aid LiDAR data interpretation, not breakline collection. Systems that incorporate metric aerial cameras improve imaging quality, but would require an extensive amount of ground control to obtain the 10 centimeter or better accuracy needed here, making it inefficient for producing 1-foot contours. Digging Deeper into the Findings The notion that LiDAR-derived breaklines may not be accurate in areas of flat terrain prompted additional research. Because the breakline issues in the low-lying areas would likely disturb the quality of generated contours, Fugro EarthData's research and development team questioned whether the indiscriminate collection of breaklines, as is practiced in photogrammetry, is necessary for lidargrammetry. LiDAR data are, after all, much denser than their photogrammetric counterparts.


	To test this hypothesis, the project team created a slope map of the project site, flagging only areas of abrupt change for compilation (Figure 4), which reduced the number of breaklines significantly (Figure 5). Comparing the new LiDAR-derived contours to the baseline model revealed much improved accuracy, supporting the idea that full photogrammetric-level breakline generation in lidargrammetry is unnecessary and possibly damaging, especially in areas with little relief. In addition to improving overall accuracy, the slope-based approach increases production efficiency by limiting the need for time-intensive breakline compilation. A Good First Step As a first step in documenting the effectiveness of 1-foot contour modeling from medium-density LiDAR data, this study offers several important conclusions. First and foremost, 1-foot contours can be reliably generated to meet all mapping accuracy standards. Further, LiDAR-derived contour modeling is a cost-effective method, provided missions are planned with accurate airborne GPS and drift-free inertial data; smart data algorithms are employed for high-quality contours; and breakline collection is limited to true breaks in the smoothness of the terrain for better production throughput and overall data accuracy. In addition, LiDAR-derived breaklines should be used with caution, especially in flat regions and low definition areas—otherwise service providers risk introducing error into the final contour models. Further studies are needed to determine the effect of denser LiDAR data in the range of .50- to 1-meter post spacing, the quality of the contours and the degree of reliance on breaklines. Author's note: Thanks to several of my Fugro EarthData colleagues for their contributions to this article: Tian Wang, Dave Chavez, Radha Kandukuri, Nora May and Debbie Simerlink.
See more Featured Articles See Featured Images Subscribe to Earth Imaging Journal Go to Home Page

[none]

Copyright ©2003-2010 Earthwide Communications LLC - Powered by eNetwork Marketing