Classification of urban surface materials with airborne hyperspectral images

With evolving technology of hyperspectral remote sensors, land surface analyses have been successfully applied to complex urban environment. Also, numerous methods to classify urban areas have been developed in the past years. However, methodological advances are rarely im-plemented in software systems and thus end users that cannot program their own algorithms are not able to make use of them.This study tested a method to classify urban surface materials by applying a library with robust spectral signatures of impervious materials gathered under laboratory conditions (SLUM) on VIS-NIR hyperspectral aerial images (APEX).This was conducted by only using commonly available GIS software like ENVI with its Spec-tral Angle Mapper classification tool. The method was tested with data from the cities of Basel and Baden (Switzerland).The results show that a surface material classification for complex urban areas based on hyper-spectral data and a spectral library is possible with common GIS software. Therefore, it can be conducted by GIS practitioners without the need of programming individual algorithms.The results also show some limitations of the method. Especially large shadows make correct material detection difficult. The accuracy of the classification depends mainly on the quality of the hyperspectral image and should therefore be improved with the latest atmospheric and BRDF correction algorithms. Further development of the method is needed to address problems such as mixed pixels and to improve classification of vegetation and asphalt.