Tree supported road extraction from arial images using global and local context knowledge

Authors:
Matthias Butenuth;Bernd-M. Straub;Christian Heipke;Felicitas Willrich
Affiliations:
Institute of Photogrammetry and GeoInformation, University of Hannover, Hannover;Institute of Photogrammetry and GeoInformation, University of Hannover, Hannover;Institute of Photogrammetry and GeoInformation, University of Hannover, Hannover;Institute of Photogrammetry and GeoInformation, University of Hannover, Hannover
Venue:
ICVS'03 Proceedings of the 3rd international conference on Computer vision systems
Year:
2003

Citing 1
Cited 1

An Unbiased Detector of Curvilinear Structures

IEEE Transactions on Pattern Analysis and Machine Intelligence

Network snakes-supported extraction of field boundaries from imagery

PR'05 Proceedings of the 27th DAGM conference on Pattern Recognition

Quantified Score

Hi-index	0.00

Visualization

Abstract

The quality control and update of geo-data, in this case especially of road-data, is the primary aim of the system, which is presented in the paper. One important task of the system is the automatic extraction of roads from aerial images. Structural knowledge about the scene, provided by existing information from a GIS database, is subdivided into global and local context knowledge. The "classical" global context approach was enhanced in such a way that additional context regions and relations were defined, mainly based on the different appearance of roads in these regions. Additionally, trees were added to the context model on the local level. After the extraction of rows of trees the road network is generated using this information as candidates for road segments. The rows of trees obtain evidence from the functional part of the road network model. Both extensions make the approach for road extraction more robust and more general, as is shown in various examples using 1: 12500 panchromatic orthoimages.