2D building change detection from high resolution satelliteimagery: A two-step hierarchical method based on 3D invariant primitives

The analysis of remotely sensed data for object extraction is a key step in an increasing number of GIS (Geographic Information Science) applications, in particular for mapping, updating and change detection purposes. The main goal of this paper is to present an automatic method for detecting changes in a 2D building database, starting from recent satellite images. The workflow of our method is divided into two steps. 3D primitives, extracted from multiple images or from a correlation Digital Surface Model (DSM), are firstly collected for each building and matched with primitives derived from the existing database in order to achieve a final decision about acceptance or rejection. A specific algorithm, based on the DSM and a computed Digital Terrain Model (DTM), is subsequently used to extract new buildings. The method is here introduced and tested in two test areas, very different regarding the land use and topography. The outcomes of the method are assessed and show the good performance of our system, especially in terms of completeness, robustness and transferability.