ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Publications Copernicus
Articles | Volume II-7
ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., II-7, 15–21, 2014
ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., II-7, 15–21, 2014

  19 Sep 2014

19 Sep 2014

Generation of 2D Land Cover Maps for Urban Areas Using Decision Tree Classification

J. Höhle J. Höhle
  • Dept. of Planning, Aalborg University, Skibbrogade 3, 9000 Aalborg, Denmark

Keywords: Mapping, Land Cover, Urban, Classification, Decision Support, Point Cloud, Sampling, Cartography

Abstract. A 2D land cover map can automatically and efficiently be generated from high-resolution multispectral aerial images. First, a digital surface model is produced and each cell of the elevation model is then supplemented with attributes. A decision tree classification is applied to extract map objects like buildings, roads, grassland, trees, hedges, and walls from such an "intelligent" point cloud. The decision tree is derived from training areas which borders are digitized on top of a false-colour orthoimage. The produced 2D land cover map with six classes is then subsequently refined by using image analysis techniques. The proposed methodology is described step by step. The classification, assessment, and refinement is carried out by the open source software "R"; the generation of the dense and accurate digital surface model by the "Match-T DSM" program of the Trimble Company. A practical example of a 2D land cover map generation is carried out. Images of a multispectral medium-format aerial camera covering an urban area in Switzerland are used. The assessment of the produced land cover map is based on class-wise stratified sampling where reference values of samples are determined by means of stereo-observations of false-colour stereopairs. The stratified statistical assessment of the produced land cover map with six classes and based on 91 points per class reveals a high thematic accuracy for classes "building" (99 %, 95 % CI: 95 %-100 %) and "road and parking lot" (90 %, 95 % CI: 83 %-95 %). Some other accuracy measures (overall accuracy, kappa value) and their 95 % confidence intervals are derived as well. The proposed methodology has a high potential for automation and fast processing and may be applied to other scenes and sensors.