Automatic DEM Generation and 3D Change Detection from Satellite Imagery
Krauss, Thomas; d'Angelo, Pablo; Tian, Jiaojiao; Reinartz, Peter
DLR - German Aerospace Center, GERMANY

In this paper we present a new method for fully automatic generation of digital surface models (DSM) from very high resolution (VHR) satellite imagery and the consecutively automatic change detection from the derived 3D information.

With the launch of more and more high resolution satellites like the Sentinel series the available amount of earth observation data increases so rapidly that new fully automatic methods for information derivation from these data is increasingly essential. Very high resolution satellites like the commercial WorldView or GeoEye satellites and their predecessors QuickBird and Ikonos expanded the ground sampling distance for the first time below one metre. Such resolutions allow the automatic generation of high resolution DSMs even from urban areas. Through the newly available dense matching methodologies very high detailed DSMs can be derived for urban areas, especially if several viewing directions are acquired by the satellite within one orbit. Consecutive monitoring of areas and automatic derivation of the digital surface models make an automatic detection of changes more efficient than just using the spectral information. So e.g. the image based automatic change detection between vegetation and snow covered regions delivers too much false positive results where volume changes like new buildings, excavations or dumps mostly will not be detected correctly. In such cases the generation of DSMs give an additional information for detection of changes.

A newly developed method for automatic bundle adjustment of VHR stereo or multistereo images and the subsequent derivation of a DSM of high accuracy is presented, using the DLR developed semi global matching (SGM) method. By developing a methodology of fusing the information from imagery with the high resolution DSM from different epochs an automatic detection of three dimensional changes together with a detailled volume estimation can be performed.

In this paper we will present the involved methods for the generation of the high resolution surface models, the fusion and classification and finally the automatic 3D change detection. The methods are applied to some VHR stereo test data sets and the results are evaluated for quality and usefulness for automatic information derivation from large data sets.