Share Email Print

Proceedings Paper

Sensor data fusion for textured reconstruction and virtual representation of alpine scenes
Author(s): Gisela Häufel; Dimitri Bulatov; Peter Solbrig
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

The concept of remote sensing is to provide information about a wide-range area without making physical contact with this area. If, additionally to satellite imagery, images and videos taken by drones provide a more up-to-date data at a higher resolution, or accurate vector data is downloadable from the Internet, one speaks of sensor data fusion. The concept of sensor data fusion is relevant for many applications, such as virtual tourism, automatic navigation, hazard assessment, etc. In this work, we describe sensor data fusion aiming to create a semantic 3D model of an extremely interesting yet challenging dataset: An alpine region in Southern Germany. A particular challenge of this work is that rock faces including overhangs are present in the input airborne laser point cloud. The proposed procedure for identification and reconstruction of overhangs from point clouds comprises four steps: Point cloud preparation, filtering out vegetation, mesh generation and texturing. Further object types are extracted in several interesting subsections of the dataset: Building models with textures from UAV (Unmanned Aerial Vehicle) videos, hills reconstructed as generic surfaces and textured by the orthophoto, individual trees detected by the watershed algorithm, as well as the vector data for roads retrieved from openly available shapefiles and GPS-device tracks. We pursue geo-specific reconstruction by assigning texture and width to roads of several pre-determined types and modeling isolated trees and rocks using commercial software. For visualization and simulation of the area, we have chosen the simulation system Virtual Battlespace 3 (VBS3). It becomes clear that the proposed concept of sensor data fusion allows a coarse reconstruction of a large scene and, at the same time, an accurate and up-to-date representation of its relevant subsections, in which simulation can take place.

Paper Details

Date Published: 5 October 2017
PDF: 14 pages
Proc. SPIE 10428, Earth Resources and Environmental Remote Sensing/GIS Applications VIII, 1042805 (5 October 2017); doi: 10.1117/12.2278237
Show Author Affiliations
Gisela Häufel, Fraunhofer-Institut für Optronik, Systemtechnik und Bildauswertung (Germany)
Dimitri Bulatov, Fraunhofer-Institut für Optronik, Systemtechnik und Bildauswertung (Germany)
Peter Solbrig, Fraunhofer-Institut für Optronik, Systemtechnik und Bildauswertung (Germany)

Published in SPIE Proceedings Vol. 10428:
Earth Resources and Environmental Remote Sensing/GIS Applications VIII
Ulrich Michel; Karsten Schulz, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?