2004 UC Proceedings Abstract

back
   Back



A Complete Building Extraction System From Elevation Data
Track: Remote Sensing Imagery
Author(s): Chris Finlayson, David Opitz

This paper describes a complete semi-automated system for extracting building footprints from remotely sensed images. Buildings have been noted as the most time-consuming feature for maintaining a GIS database, and attempts at automating this process has traditionally been met with limited success. Our proposed system extracts footprints automatically and provides tools to solve the commonly encountered post-extraction manual clean up scenarios. The building footprints can be analyzed in either 2 or 3 dimensions using ArcMap or the 3D Analyst extension to ArcGIS. The core of the system is Feature Analyst, an extension to ArcGIS. This work extended Feature Analyst to incorporate new and existing building extraction techniques while maintaining an intuitive workflow. The results of tests on several datasets demonstrate great promise for the system in its accuracy, simplicity, overall low cost of labor, and convenience because of its integration within a well known environment: ArcGIS.

Chris Finlayson
University of Montana
Computer Science
Department of Computer Science
Social Science Building
Missoula , MT 59812
US
Phone: 406-241-2831
E-mail: opitz@cs.umt.edu

David Opitz
University of Montana
Computer Science
2822 Carla Jo Road
Missoula , MT 59803
US
Phone: 406-251-2690
Fax: 406-829-3593
E-mail: opitz@vls-inc.com