2008
41
ASPECTS OF THE STANDARDIZATION OF SENSOR AND DATA FUSION OF
REMOTE SENSING DATA
R. Hoffmann, R. Reulke
Humboldt-University Berlin, Department of Computer Science, Computer Vision
Unter den Linden 6, 10099 Berlin, Germany
(rhoffmann, reulke)@informatik.hu-berlin.de
Commission I, WG 1/1
KEY WORDS: Standards, Fusion, Algorithms, Sensor, Classification, Remote Sensing
ABSTRACT:
Owing to continuous developments in technology and new requirements single sensor measurements are faced with constraints in
their accuracy and thus in the applicability of their results. An innovative method of improving the geometric and radiometric data
quality is the fusion of data obtained from different digital sensors. Sensor fusion means merging data obtained from different
individual physical sensors to provide more comprehensive information from a single ’logical’ or ’virtual’ sensor. Airborne and
spacebome high-resolution digital cameras, laser scanners, hyperspectral systems, radar and InSAR systems have been considered in
this context.Common fusion methods are resolution improvement, fusion of altitude or distance information and texture information
(orthophoto generation), pan-sharpening and tracking. The derivation of orientation information from fusion of different sensors is
not regarded in this paper.Assessment criteria for fusion results of data obtained from different sources have so far only been
established in a few cases. Within the INS project (German Institute for Standardization DIN, 2008) the scientific bases for
standardization has been developed to provide manufacturers and users with rules for the quality of the end products - also with
regard to the international market. Therefore sensors and methods for achieving possible fusion products were introduced and
discussed. This formed the basis for the development of a working document for a draft standard for requirements for geometric
fusion methods.
KURZFASSUNG:
Messungen mit einzelnen Sensoren kommen durch die ständige technologische Entwicklung und neue Anforderungen an die
Grenzen ihrer Genauigkeit und damit auch der Verwendbarkeit dieser Ergebnisse. Eine innovative Methode zur Verbesserung der
geometrischen und radiometrischen Datenqualität ist die Fusion von Daten unterschiedlicher digitaler Sensoren. Dabei werden die
Daten von verschiedenen individuellen physikalischen Sensoren zu einer umfassenderen Information eines „logischen“ oder
„virtuellen“ Sensors zusammengefügt. In diesem Zusammenhang wurden flugzeug- und satellitengetragene hochauflösende
Kameras, Laserscanner, Hyperspektralscanner, Radar- und InSAR-Systeme berücksichtigt.Typische Fusionsverfahren sind die
Auflösungsverbesserung, die Verbindung von Höheninformation und bildhafter Information (mit Parallelen zur
Orthophotoerstellung), Pan-sharpening sowie Objektverfolgung. Die Ableitung von Orientierungsinformationen aus der Fusion
unterschiedlicher Sensoren wird hier nicht betrachtet.Bislang existieren nur in wenigen Fällen Kriterien, die eine Beurteilung der
Ergebnisse der Fusion von Daten unterschiedlicher Quellen erlauben. Im INS-Projekt (German Institute for Standardization DIN,
2008) sollten die wissenschaftlichen Grundlagen erarbeitet werden, um Herstellern und Anwendern - auch in Hinblick auf den
internationalen Markt - Festlegungen über die Qualität der Endprodukte an die Hand geben zu können. Deshalb wurden Sensoren
und Verfahren für potenzielle Fusionsprodukte vorgestellt und diskutiert, um auf dieser Grundlage einen Norm-Entwurf für
Anforderungen an geometrische Fusionsverfahren zu erarbeiten.
1. INTRODUCTION
In order to improve measurement accuracy and therefore also to
extend the applicability of sensors it is advisable to merge data
obtained from different individual physical sensors so that more
comprehensive information is provided by a single “logical”
sensor. The measurement data of a virtual sensor is already
calibrated and has a well-known spatial and temporal relation
between each single dataset. Currently many sensors,
applications and techniques exist. Scientific literature has been
reporting on it for more than 25 years (Haydn et al., 1982). One
of the first applications was the integration of the low-resolution
multispectral bands of Landsat data with the high-resolution
panchromatic band of SPOT data (pan-sharpening). This
opened up new possibilities for simultaneous processing of
remote sensing data. With the introduction of digital
photogrammetric airborne cameras different applications, which
were only discussed in the scientific community, are becoming
economically important. Especially the simultaneous
measurements with multispectral bands also open up
possibilities for remote sensing applications apart from
photogrammetric products. Thus, fusion products are already
available and integrated into a workflow for higher level
photogrammetric and remote sensing products. So this new
generation of digital cameras with multispectral bands is
leading towards integration between photogrammetry and
remote sensing. At the same time, high resolution multispectral
data pose a new challenge for the remote sensing community.
