VERIFICATION OF A BUILDING DAMAGE ANALYSIS AND EXTENSION TO SURROUNDINGS OF REFERENCE BUILDINGS

摘要

After a disaster not necessarily all buildings in affected areas are damaged. Therefore, a building damage analysis does not only have to distinguish different damage types. In fact, unchanged buildings have to be identified correctly with a high reliability, too. For this purpose, a building damage detection and classification method based on airborne LIDAR data is applied to data of two Swiss villages which did not change between the acquisition dates of pre- and post-event data. The aim is to verify the correct classification of unchanged buildings. The achieved results which are very satisfying are presented in this paper. In its current state, this building damage detection and classification method analyses only changes inside reference building contours. In order to integrate also changes outside the reference building contours into the approach, different preprocessing steps are necessary. One of them - the elimination of vegetation in the post-event LIDAR data - is explained in this paper. It is necessary because remaining vegetation in the surroundings of a building could erroneously be interpreted as debris and might, therefore, cause misclassifications. Different procedures for eliminating vegetation are described. Investigations showed that their suitability depends on the vegetation's foliation state. Thus, in summer when trees are leafy it is very helpful to have additional multispectral data. In this case, the Normalised Difference Vegetation Index (NDVI) in combination with the near infrared can be used. In winter, when trees are bare and, therefore, at least partially penetrable for the laser beam differences of first and last echo LIDAR data proved to be suitable.