Abstract
Modern high-resolution satellite systems now provide users with optical or synthetic aperture radar (SAR) data with sub-meter accuracy, making it feasible to perform real-time damage estimation of individual infrastructure in a region devastated by an earthquake or other natural hazard. This paper presents a hierarchical framework for object-based segmentation using satellite images. Satellite images, pre-earthquake event, were obtained, and using a fine-level segmentation, an estimated ground truth (EGT) depicting the number and geometric boundaries of individual houses in an urban region was determined. To address the issue of optimal parameter settings in obtaining the EGT, a statistical approach based on receiver characteristic curve (ROC) analysis was implemented. Examples are illustrated for satellite images from the Bam, Iran, region. Damage maps based on different damage indices were generated with individual houses rendered with colors representing the severity of damage.
Get full access to this article
View all access options for this article.