Extension of an Automatic Building Extraction Technique to Airborne Laser Scanner Data Containing Damaged Buildings

Airborne laser scanning systems generate 3-dimensional point clouds of high density and irregular spacing. These data consist of multiple returns coming from terrain, buildings, and vegetation. The major difficulty is the extraction of object categories, usually buildings. In the field of disaster management, the detection of building damages plays an important role. Therefore, the question arises, if damaged buildings can also be detected by a method developed for the automatic extraction of buildings. Another purpose of this study is to extend and test an automatic building detection method developed initially for first echo laser scanner data on data captured in first and last echo. In order to answer these two questions, two institutes share their data and knowledge: the Institute of Photogrammetry and Remote Sensing (IPF, Universität Karlsruhe (TH), Germany) and the MAP-PAGE team (INSA de Strasbourg, France). The used 3D LIDAR data was captured over an area containing undamaged and damaged buildings. The results achieved for every single processing step by applying the original and the extended algorithm to the data are presented, analysed and compared. It is pointed out which buildings can be extracted by which algorithm and why some buildings remain undetected.