Building features extraction based on facade regularities using TerraSAR-X ST data

Abstract

The grammar of facade structures is often related to regularly distributed signature patterns in high-resolution synthetic aperture radar (SAR) images. Given the structural details in the SAR image, algorithms for building extraction/monitoring should exploit this contextual information for an object-based analysis. In this paper we propose a framework to extract detailed high rise building features including building masks, orientations and stories. We developed a top-down approach which consists of processing at three levels, namely image level, region-of-interest (ROI) level and storey level. The analysis follows assumptions on the intensity distribution, the linear arrangement, and the regularity of point-like signatures. The proposed approach is validated using a high resolution staring spotlight TerraSAR-X image. A preliminary experiment for change detection using the extracted building features is also studied in this paper.