In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B
243
AIRBORNE HYPERSPECTRAL IMAGE GEOREFERENCING
AIDED BY HIGH-RESOLUTION SATELLITE IMAGES
C. K. Toth ab *, J.H. Oh b , D. A. Grejner-Brzezinska b
a Center for Mapping, The Ohio State University, Columbus, Ohio, USA toth@cfm.ohio-state.edu
b Satellite Positioning and Inertial Navigation (SPIN) Laboratory, Dept, of Civil and Environmental Engineering and
Geodetic Science, The Ohio State University, Columbus, Ohio - (oh. 174, dbrzezinska)@osu.edu
Commission VII, WG VII/4
KEY WORDS: Georeferencing, Hyperspectral imagery, High-resolution satellite images, Image matching, Pushbroom camera
ABSTRACT:
Over the past decade, airborne hyperspectral systems have shown remarkable performance in identifying and classifying a variety of
ground objects, such as differentiating between minerals, vegetations, artificial materials, water, etc. Though the hyperspectral
imaging market is still relatively small, yet it is steadily growing. Currently, most of the high performance systems are of the
pushbroom camera type, and consequently, the sensor orientation of these systems heavily relies on the integrated GPS/INS (Global
Positioning System/Inertial Navigation System) based direct georeferencing solution. In this study, an indirect georeferencing
method is proposed that is based on utilizing robust image matching to high-resolution satellite imagery. This solution can be used in
circumstances where GPS/TNS-based georeferencing is not available or not feasible due to GPS signal loss and/or the lack of GPS
infrastructure. The proposed method is motivated by the attractive properties of state-of-the-art high-resolution satellite imagery,
including large swath width, high spatial and temporal resolution, and high positional accuracy. For robust image matching, a
combination of SURF (Speeded-Up Robust Features) and RANSAC (RANdom SAmple Consensus) is utilized, and the trajectory
modeling of the airborne hyperspectral pushbroom camera is based on the collinearity equation camera model with the Gauss-
Markov stochastic error model. Tests on simulation data showed encouraging performance results for the proposed approach.
1. INTRODUCTION
Over the past decade, airborne hyperspectral imaging (HSI)
systems have shown excellent performance in several
applications to identify and classify a broad range of ground
objects, including minerals, vegetations, artificial materials, and
water etc. Airborne HSI sensors measure the light from the
earth’s surface in high spectral resolution; typically, each pixel
of hyperspectral data contains dozens or hundreds of spectral
bands. HSI technology has been used in many commercial and
defence applications.
Airborne HSI systems are predominantly based on the
pushbroom camera model, which heavily relies on direct
georeferencing. Typically, the georeferencing solution,
including platform position and attitude data, is computed by an
Extended Kalman Filter (EKF), where first aircraft GPS data
are processed in DGPS (Differential GPS) mode based on a
nearby ground GPS base station (or network solution), and the
DGPS results are feed back to the EKF to control the INS
which provides the final georeferencing solution (Zhang et al.,
1994; Grejner-Brzezinska, 1999; Haala et al., 2000; Tuo and
Liu, 2005; Grejner-Brzezinska et al., 2005). In addition, sensor
alignment information, which is obtained by accurate boresight
calibration, is applied to the platform georeferencing data to
derive the exterior orientation parameters (EOPs) of the camera.
While GPS is generally available, there are certain rare
circumstances that direct georeferencing is not available, such
as in GPS denied environment. In fact, GPS signals could be
vulnerable to interference, such as jamming, broadcast television,
ultrawide-band communications, over-the-horizon radar and
cellular telephones (Carroll, 2001). In addition, there are remote,
inaccessible areas that lack a geodetic infrastructure and thus
GPS/INS-based georeferencing is not always feasible. In these
cases, EOPs have to be estimated through the indirect or image
referenced georeferencing technique.
Previously acquired and processed geospatial data are a good
source for ground control that can be used not only for airborne
image georeferencing (Dowman, 1998; Lin and Medioni, 2007;
Cariou and Chehdi, 2008; Oh et al., 2010), but also for aircraft
navigation (Oh et al., 2006; Conte and Doherty, 2009). The
requirements for such reference data include high positional
accuracy, and geometric and radiometric properties similar to
target airborne imagery. From the various geospatial images
obtained from different sensors, high resolution satellite images
have clear advantages due to their uniform global accessibility.
Since high resolutions satellite images meet the main
requirements in terms of the spatial and spectral resolution and
positioning accuracy, this study proposes their use for
georeferencing of airborne pushbroom imagery. Note that the
spatial resolution of satellite images is lower than that of
airborne imagery, yet the currently allowed 50 cm satellite
image resolution has a good potential for image matching.
Figure 1 depicts a GPS denied situation. At the epoch t 0 , direct
georeferencing becomes unavailable, as GPS is denied and
there is no external information until tj. Between tj and t 2 , there
is reference image data available and using common features,
georeferencing is possible. Between the epochs t 2 and t 3 , the
image referenced georeferencing may not be feasible when
there are not enough image features in the reference data, such
as in forested areas (Oh et al., 2010).
* Corresponding author.