Automatic generation of raster-based height data for the Netherlands based on the AHN2 data set

More Info


The recent emergence of Light Detection And Ranging (LiDAR) scanning technology has resulted in the availability of very large three dimensional point cloud data sets. These LiDAR data sets have become a main source for the modeling and reconstruction of both ground surface as well as above-ground objects such as buildings and vegetation. For the Netherlands, the second version of the Algemeen Hoogtebestand Nederland (AHN) is a country-wide high-resolution point cloud data set comprising the nation’s terrain by measuring heights, obtained by LiDAR, using Airborne Laser Scanning (ALS). None of the currently available raster-based products based on the AHN2 data set can be considered as being good: height maps contain holes, unintentional dynamic objects are present and more. Causes for errors can be found in deviations during the collection of the point cloud data as well as the applied methodology to process the data. This thesis will identify the possibilities to improve the quality of raster-based height map products based on the AHN2 data set with respect to currently available products. Quality will be determined with respect to the Geographic Information Quality principles standard, introduced by the International Organization for Standardization (ISO). A five-step methodology is proposed in order to generate raster-based height data from massive LiDAR point cloud samples from the AHN2 data set. In the first step, the massive LiDAR point cloud data will be split up in overlapping tiles in order to pipeline subsets of data in order to feed data sequentially to the computers’ main memory. In the second step filtering of different classes takes place in order to filter specific information within the point cloud data. In the third step, for each filtered class specific interpolation methods will be applied in order to achieve raster-based data from the point cloud data that fits best for a certain class of data. In the fourth step some post-processing steps will be applied in order to optimize the raster data and a composition of the tiles that were decomposed in the first step. In the fifth step raster visualization will be applied in order to support visual inspection of the data. Quality assessment shows that the methodology proposed within this thesis is capable to process data with a completeness rate of 100%. Positional accuracy is determined with respect to currently available raster-based height maps since no reference data is available. For Digital Elevation Model (DEM)s low positional errors are measured where for Digital Surface Model (DSM)s higher positional errors are measured due to erroneous data within current DSM products. Thematic accuracy is defined for a larger amount of classes in comparison with currently available raster-based height maps with a sensitivity rate between 56?100% ± 5%.