|
Current
Methods for Developing Digital Terrain Models
John Althausen, Aimee Baldwin, and Kurt Schwoppe
Over the last two decades, technologies have been
developed that are revolutionizing the way geospatial
scientists derive and update terrain models. These
advances include the systems that collect the terrain data
as well as the hardware and software utilized to
manipulate, manage, and analyze it. Traditional methods of
generating terrain models, including manual photogrammetry
and field surveying, are rapidly leaving the marketplace
and are being replaced by processes that combine data from
remote sensing systems with image processing and/or
softcopy photogrammetry software. This article looks at
current trends in the remote sensing industry for deriving
DTMs including the three most popular approaches for
capturing terrain data: optical, laser, and microwave
remote sensing.
High-Resolution Optical Data
& Softcopy Photogrammetry
The use of high-resolution digital data along with
softcopy photogrammetric techniques, to generate a DTM, is
probably the most familiar. The softcopy approach to
extract DTMs is built on the same principles and theories
that photogrammetrists have been practicing “manually”
for years.
The principle is straightforward. First, two images
are acquired of the same area with slightly different
viewing perspectives (stereo-overlap). These images are
then aligned and geometrically matched so that a
mathematical (triangulation) model can be obtained. The
analyst then has the opportunity to view the modeled
stereopair in 3D to manually extract the terrain
information or use automated stereocorrelation tools and
an existing lower-resolution terrain model to extract a
new DTM. The automated process is typically used when the
surface landscape needs to be extracted.
“Above-the-landscape” features (e.g., buildings) are
typically derived manually (using stereo extraction tools)
and then “placed” on the DTM (Figure 1).
The same problems and pitfalls (e.g., clouds,
shadows, etc.) that an analyst finds when interpreting
optical data are prevalent when using this type of
remotely sensed data for DTM extraction. Care must be
taken when evaluating the DTM product and error
assessments are always a wise, and necessary, final step.
Typically, an analyst will use a terrain editing toolbox
to manually “fix” areas in the derived DTM where
problems exist.
Synthetic Aperture Radar &
Interferometric Process
Due to the geometric robustness of most Synthetic
Aperture Radar (SAR) data, DTMs generated using
interferometric techniques (which compare two images taken
at slightly different locations) have the potential to be
highly accurate and reliable. However, using SAR data to
generate a DTM requires special analysis tools and an
understanding of how SAR pulses interact with Earth
objects. Most dedicated image processing software packages
on the market today provide the capabilities to process
Interferometric Synthetic Aperture Radar (IFSAR) data and
derive DTMs from them (Figure 2).
Data collection for IFSAR is generally carried out
in one of two approaches; two antennas, one-pass or single
antenna, repeat-pass. Airborne IFSAR typically
incorporates the two antennas, single-pass approach while
satellite IFSAR is usually done with a single antenna,
repeat-pass approach. The recent Shuttle Radar Topography
Mission (SRTM) incorporated the two antennas, single-pass
approach to collect global IFSAR data currently trickling
into the marketplace.
The appropriateness of a pair of images for
generating an IFSAR Digital Stereo Model (DSM) is measured
by the “coherence” between the two. Poor coherence is
caused when the backscatter is different on the two
images, thus phase unwrapping (calculating the absolute
phase difference between the two images) to extract
heights cannot be performed. Coherence is usually stable
in single-pass IFSAR and can be troublesome in repeat-pass
IFSAR due to a greater likelihood of environmental change
between the two images.
One of the issues that challenge the IFSAR user is
the frequency response between the different SAR
wavelengths. Shorter wavelengths typically scatter off
partially penetrable objects while longer wavelengths
generally reach and echo off of non-penetrable surfaces.
Combining the two wavelength domains, as well as
incorporating polarimetric information, allows for
improved 3D exploitation as well as the potential
extraction of bare soil DTMs. The more “open” the
environment, the better the chance to extract a bare soil
DTM.
Laser Pulsing &
Range/Reflectance Measurements
The most recent technology being applied to DTM
generation is airborne LiDAR. LiDAR systems carry a
transmitting laser of a specific wavelength and a
receiver, an optical telescope. Different kinds of lasers
are used depending on the power and wavelength required.
The lasers may be either continuous wave (like a light
bulb) or pulsed (like a strobe light). The LiDAR receiving
system records the scattered light returned at fixed time
intervals.
LiDAR systems typically use extremely sensitive
detectors called photomultiplier tubes to detect the
backscattered light. The photo-counts received are
recorded for fixed time intervals during the return pulse.
The times are then converted to range bins (heights) since
the speed of light is taken as constant.
LiDAR systems capture multiple reflections (Figure
3), caused by objects which are smaller than the footprint
in different ranges. Early returns are typically used to
measure and model partially penetrable objects (e.g.,
tree canopies), while latter returns are exploited
to measure non-penetrable surfaces (e.g., the land). Thus
for DSM/DTM generation, the latter returns are typically
the ones of interest (Figure 4).
Similarities, Comparisons &
Integration
IFSAR, LiDAR, and photogrammetric data sets are
currently used for many geospatial applications, including
the derivation of DTMs.
Many times, a single mission will require the
collection of more than one data set, collected on
different platforms. Integration is possible and can
strengthen the final DTM product. IFSAR can provide
geometric fidelity to the model, LiDAR can supply accurate
“point-level” data, while high-resolution
photogrammetric data can provide realistic terrain
visualization (Table
1).
Though the accuracy of all three techniques is
well-established, there are still sources of error which
are not implicit or quantified. Calibration of the
instrumentation, and a better understanding of the
geography, may help produce better DTMs, but the fact
remains that there will be artifacts in all three data
collects that will always be part of the derived DTMs.
Analysts must always expect the unexpected.
About the Authors
John Althausen is the Senior Remote Sensing
Scientist for Leica Geosystems’ Defense Solutions. He
conducts training courses in DTM generation using
photogrammetric, IFSAR, and LiDAR technologies. You can
reach John at [email protected].
Aimee Baldwin is a Project Manager with Leica
Geosystems’ Defense Solutions. You can reach Aimee at
aimee. [email protected].
Kurt Schwoppe is Vice President of Leica Geosystems
GIS & Mapping, LLC., Defense Solutions office in
Alexandria, Virginia. You can reach Kurt at [email protected].
Acknowledgments
The authors wish to thank the European Space Agency
for providing the ERS-1 and ERS-2 imagery utilized to
generate the IFSAR DTM.
Back
|
