• Citer

Change detection for update of topographic databases through multi-level region-based classification of VHR optical and SAR data

par Hanson, Emilie ;Wolff, Eléonore
Référence The international archives of the photogrammetry, remote sensing and spatial information sciences, 38-4
Publication Publié, 2010

Article révisé par les pairs

• Détails
• Contenu

Résumé :

The geographic database producers need improved and faster updating methods for their topographic databases to fulfil the user's demand. In this context, change detection methods using remotely sensed data are interesting tools applicable on wide areas and are less demanding for human work. Indeed, database update is currently mainly realized through a tedious and laborious photogrammetric restitution. The objective of this study is to compare built-up and roads extracted with a hierarchical region-based classification method (Definiens) from VHR optical and SAR data to the ones of an old geographic database in order to pinpoint changes. Simulation of PLEIADES data on Toulouse at a resolution of 0.7m in the panchromatic band and 2.8m in the multispectral one are used together with a DSM derived from SAR Cosmo-Skymet data. The NGI-France Topo-Pays database is generalised to map 5 classes: building, road, vegetation, water and other. Their outline is used as a first level during the segmentation process; the objects of the database are then subdivided according to their spectral homogeneity. This procedure allows the use of the database as prior knowledge and avoids objects matching between detected and database objects to map changes. Data are classified with a hierarchical scheme into 5 classes using spectral thresholds and the nearest neighbour algorithm. The results show that the use of the DSM in the classification process allows a better accuracy in the change detection process.