|
GEODEETTINEN LAITOS |
 |
09.06.2000 kirjasto@fgi.fi |
SUOMEN GEODEETTISEN LAITOKSEN TIEDONANTOJA |
|
|
| Interpolation and Accuracy of Contour-based Raster DEMs by J. Oksanen and O. Jaakkola Abstract This report describes three raster-based digital elevation model (DEM) interpolation algorithms, which use existing digital map data as an interpolation input. The specific purpose was to develop algorithms that generate geomorphologically realistic DEMs. The geometric quality of the DEMs was determined and the results were compared against four reference DEMs, two of which were interpolated using standard geographical information system (GIS) software tools. The first reference model, called RasterTIN, is based on Delaunay triangulation and uses linear interpolation to transform the triangulated irregular network (TIN) to regular raster data. The second reference model, called TOPOGRID, uses spline-based interpolation, which is optimized to produce a hydrologically sound DEM. The third and fourth reference models (called NLS10 and NLS25) are commercial DEMs produced by the National Land Survey (NLS) of Finland. All three developed interpolation methods are based on a deterministic linear interpolation and use raster operations of the Map algebra. The interpolated DEMs were generated using methods named LIDIC (Linear Inverse Distance Interpolation from Contours), IMAI (Iterative Medial Axis Interpolation) and SIFI (Simple Iterative Filtering Interpolation). LIDIC is based on a direct linear interpolation by cost distance transformation, IMAI and SIFI are based on an iterative contour handling by Euclidean allocation transformation and repeated contour data smoothing. The input data source for the interpolation algorithms was a topographic database produced and maintained by the NLS. The main data was contours and lake shorelines, but the usability of streamline and faultline data was also investigated. The accuracy assessment was performed using spot heights from two independent sources. The first set was field survey data of 318 points collected from the eastern part of Siuntio and the western part of Kirkkonummi. The data was gathered by the Finnish Geodetic Institute (FGI) using a geodetic GPS and tachymetry. The second set was photogrammetric data of 2818 points, gathered by Finnmap Oy for a large-scale town-planning mapping of Masala and Veikkola regions in Kirkkonummi township. The DEM's overall quality was also investigated using several visualization techniques. The differences in geometric quality of the investigated DEMs are small according to the spot height error analysis. The models' systematic errors (the mean of errors in all test points) range from -0.50 m to +0.10 m and random errors (the standard deviation of errors in all test points) 1.03-1.75 m. The analyses show that LIDIC was always ranked among the two of the best and IMAI among the three of the best models, when random errors were compared. TOPOGRID was ranked the best according to the photogrammetric data, but fourth according to the field survey data. NLS25 has the largest random errors with all test datasets. Despite the similarity of models in statistical error analysis, there are remarkable differences when the DEMs are visualized using appropriate techniques, such as hill-shading and modulo image techniques. The profiles created from the field survey data also highlight detailed differences in surfaces generated by the different interpolation methods. The visually most pleasing models appear to be LIDIC, IMAI, and TOPOGRID. All the TIN-based DEMs suffer from artifacts caused by the deficiencies of the Delaunay triangulation algorithm. The best DEM interpolation method appears to be impossible to declare unambiguously. All the investigated models have good and bad properties, and in the end the DEM user is responsible for correct use of the model in spatial data analysis. Therefore, the user should investigate DEM carefully and consider the actual accuracy needs thoroughly before deciding which DEM interpolation method fulfills the requirements in the best way. |
|
09.06.2000 kirjasto@fgi.fi |
|
|
|
| Geodeettinen laitos Geodeetinrinne 2, PL 15 02431 MASALA |
|
: (09) 2955 50