Creation of high resolution, precise digital elevation models

• Main Authors: Young-ran Lee, Tony Schenk, William Krabill, Jan Mcgarry
• Other Authors: The Pennsylvania State University CiteSeerX Archives
• Format: Text
• Language: English
• Published: 2001
• Subjects: Pacific; Southern Ocean; Airborne Topographic Mapper
• Online Access: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.624.6131; http://www.isprs.org/proceedings/xxxiv/3-w4/pdf/csatho.pdf

KEY WORDS: DEM generation, outlier detection, validation site Researchers of NASA’s GSFC are currently developing a scanning airborne photon-counting laser altimeter. This paper summarizes the creation of high resolution, precise DEMs from repeat passes of airborne laser scanning surveys to validate this new system. To study coastal erosion several sections of the Atlantic and Pacific coast have been mapped by NASA’s Airborne Topographic Mapper (ATM) conical scanning altimeter. We selected two 5 km long and about 1 km wide stretches of the coast in Maryland to create precise Digital Elevation Models. One site is urban area (southern Ocean City) and the other is characterized by coastal marshes and beaches (northern Assateague Island). To obtain a dense data set, we combine laser altimetry data from 21 swaths acquired in 4 different missions. Laser points over planar surfaces, such as flat roofs and parking lots were analyzed to check the the repeatibility of the measurements. The absolute accuracy of the laser scanning was assessed by comparing surfaces extracted from the laser point cloud with GPS and photogrammetry results. These studies confirm the 0.1-0.2 m vertical and submeter horizontal accuracy of the ATM system. At the heart of the interpolation procedure is a bilinear interpolation that determines the surface elevation at the grid posts from planes fitted through the points located within 2 m by 2 m grid cells. Outlier observations are detected by using a robust estimator. The residual of the plane fitting and the uniformity of the distribution of the observations within the grid cells are used to assess the accuracy of the DEM. These parameters suggest an accuracy of 0.3 m or better for 90.4 % of the DEM points on the urban area, and 48.5 % of the points have an accuracy of 0.1 m or better. Similar accuracy is achieved on the site covered by natural vegetation on Assateague Island. 1