Extracting Artificial Surface Objects from Airborne Laser Scanner Data

Abstract

This paper describes an approach to automatically identify and extract artificial surface objects from the data of imaging laser altimeters. It makes use of both the 3D geometry and surface reflectance data which advanced imaging laser altimeters are able to provide. In a first step, surface objects are detected by applying a morphology-based filter to the elevation data. These surface objects are then separated into artificial objects (buildings) and natural objects (vegetation) in the second step, using surface reflectance data, and/or elevation ‘texture’ and surface orientation. To compare the effectiveness of these identification criteria for building detection they were are applied to a test data set collected with the ScaLARS laser altimeter. A classification error analysis shows that artificial surface objects are detected best using reflectance data. A combination of reflectance and surface orientation can be used to improves the reliability further.KeywordsNatural ObjectInertial Navigation SystemSurface ObjectSurface GradientArtificial ObjectThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.