6A-4-3
required ground resolution. Line CCD sensors can
overcome the limited coverage problem, but at the
price of more complex modeling of the sensor
geometry, which has further implications on the
surface extraction performance. To compensate for
both the small coverage of the frame sensors and the
need of complex modeling with line CCDs, the use of
GPS/INS-based orientation with these imagers is
mandatory (Toth 1997). A new feature of CCD sensors
is a much higher gray-scale intensity range, easily
extending up to 12 or 14 bits.
Road/Building Extraction. Currently one of the most
popular research areas in digital photogrammetry is
definitely algorithmic research on extracting man
made objects, such as roads and buildings, from aerial
imagery (Fritsch and Ameri 1998; Haala et. al. 1998;
Wang 1998). Identifying and modeling these objects
can help resolve ambiguities in the matching process
and can increase the robustness of surface extraction
methods far beyond simply reconstructing a “draped”
surface.
Laser Scanning. Scanning laser-ranging (LIDAR)
sensors have substantially improved over recent years
and are about to enter map production. In fact, they are
rapidly becoming a de facto first choice for fast digital
surface data acquisition. Operational systems can
easily provide a large number of elevation spots with
excellent vertical accuracy, which depends primarily
only on the positioning performance of the GPS/INS
module. An unmatched feature of the laser scanning
sensors is the capability to process multiple returns,
thus being able to separate vegetation from terrain
surface and other objects (Pfeifer et. al. 1998). A
weakness of the technique is the strong signal
dependence on the surface slope.
Hyperspectral Imagery/SAR. These new sensor
techniques offer an additional dimension to the surface
extraction and interpretation tasks by providing
material-specific signatures, which can further enhance
the performance of the matching process (Tonjes and
Growe 1998). Interferometric SAR can itself provide
range data, and it is viewed as the future tool for fast
and large area terrain data acquisition.
3. TEST FLIGHT WITH LIDAR AND 4K BY 4K
DIGITAL CAMERA
To support our surface extraction research on integrating
LIDAR data with stereo imagery, a recent test flight was
organized to simultaneously collect laser data and direct
digital monochrome frame images over a well-surveyed
area at Hagerstown, MD in cooperation with EarthData
Technologies. A similar test flight was conducted in the
Northern Florida region a year earlier. However, the
Florida flight did not allow for the concurrent operation of
the two imaging sensors, and the project area did not have
acceptable surface data necessary for ground truth
comparisons. The surface extraction experiences using
only the 4K by 4K direct oriented imagery are reported
elsewhere (Toth and Grejner-Brzezinska 1998).
For the Hagerstown test flight, EarthData Technologies
provided the LIDAR system and all necessary flight
support, while the complete AIMS™ prototype, including
the 4K by 4K digital camera and the GPS/INS system,
came from OSU. In several missions, LIDAR and direct
digital image data were collected at different flying
heights over a one square kilometer area. A representative
image taken at 850 m AGL is shown in Figure 1. The
Hagerstown test site has been routinely used for
calibration and system performance evaluation tests and
has excellent and accurate photogrammetrically derived
ground-truth data, including a topographic surface and
man-made objects (mostly residential buildings).
Figure 1. Typical 4K by 4K image taken at 850 m AGL
over Hagerstown, MD.
Traditional aerial-triangulation was performed on images
collected at the lower flying height (the GSD for higher
altitude flights was around 1 m, preventing any control
point identification). The aerial-triangulation served two
purposes: to determine the best orientation of the block
and consequently, through those results, to provide the
boresight misalignment for the GPS/INS to deliver the
exterior orientation data for the remaining blocks. The 50
mm lens equipped 4K by 4K camera was laboratory-
calibrated prior to the test Eights (Seedahmed et. al.
1998). The difficulty of measuring control points as
illustrated in Figure 2 comes from two sources. First, the
limited angular resolving power provides only a very
coarse pixel resolution on the ground. Second, the very
asymmetrical target pattern is troublesome for computer-
assisted point extraction, which is typically tailored
toward symmetrical objects.