ING
)7
y short time
applications
/s to collect
it. Thus the
quantitative
york on the
veling. The
ed pulse is
and point
he platform
nother type
ble through
urn signal
icies in the
of error in
ated totally
olution and
nd the laser
y. only the
her objects
ice they do
1 and steep
s especially
nds on the
ror will be
o the slope
lope is less
he gradient
in can be
(ical object,
| amplitude
data might
d due to its
regions of
litions (air
fluence the
)rous sense
olation and
anning and
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
2.3 GPS and INS positioning and orientation techniques
Unlike photogrammetry, the laser scanning system does not
rely on triangulation in exposure station positioning. Rather it
depends exclusively on the airborne GPS and INS to provide an
accurate position and orientation of the platform. Differential
carrier phase positioning (DGPS) in kinematic mode is capable
of generating an accurate position with a precision of several
centimeters. Satellite geometry, which is quantified by the
positional dilution of precision (PDOP), plays a major role in
GPS positioning reliability. Poor satellite geometry, high
PDOP, generates inaccurate GPS positioning. Poor positioning
can be avoided by optimizing the survey time and having at
least one visible satellite in each of the four quadrants in order
to be well distributed across the sky (Mikhail et al., 2001). On
the other hand, a minimum of four visible satellites is needed to
position a receiver using the DGPS system. Also, inaccurate
orbits might produce significant errors, especially when
observing longer GPS baselines.
Multipath, when more than one signal arrives at a receiver via
more than one path, affects the vertical component of GPS
observations. This can generate a height error of several
centimeters based on the multipath configration. More on this
issue and its treatment can be found in (Georgiadou and
Kleusberg, 1988). An error in resolving the phase ambiguity,
the number of integer cycles from the antenna to the satellite, is
another source of error in GPS positioning. Signal propagation
(troposphere and ionosphere) and uncertainty in calculating
atmospheric transmission delays also affect the attainable GPS
accuracy. There are many methods to address this problem and
they can be found in (Grewal, et al., 2001; Gonzalea, 1998; and
Mikhail et al., 2001).
Error in INS attitude data can be described by these factors:
misalignment with the platform or the GPS system, biases in
the accelerometers, gyrodrift, non-orthogonality of the axes,
and gravity modeling error. These factors can accumulate a
significant amount of angular error with long mission times
3. TEST DATA SET
The data used in this research was collected using an Optech
ALTM 1210 LIDAR sensor operated by Woolpert Consultants
on April 2001. It was flown over the Purdue University campus
with an approximate area of 3,500,000m^. It measures
approximately 1700m in easting and 2000m in northing with an
approximate density of one data point per square meter. The
data consists of fourteen strips flown in the north south
direction. Each of which has an approximate length of 2000m
and width of 200m. The average flying height over terrain was
about 600m. Therefore the angular extent of a swath is
approximately 19 degrees.
4. RELATIVE ACCURACY OF DATA STRIPS
Airborne laser scanning data are acquired in a strip-wise pattern
With a strip width varying depending on the chosen scan angle
and the flying height. Usually, those strips are flown in parallel
and overlapping until the entire region of interest has been
covered. Overlap between strips (as shown in figure 1) provides
à mean to evaluate the relative accuracy between them. The
imprecision in system positioning, orientation, and ranging may
cause the same point to have two different heights if scanned at
two different times, which always happen in neighboring
overlapping data strips. These points are considered as tie
points in strip adjustment to adjust strips and eliminate or at
least minimize relative error between them. However, the
discrepancy between tie points from adjacent strips gives an
indication of the relative offsets without any strong conclusion
of the absolute error. In this section, the height discrepancy is
examined between adjacent strips.
E AT
c 188 i ves Eos 1
= - : : P o p
— Zar rte .
©
'q 187.5
X
|
187 Lo i A x l i x i 1 i
350 360 370 380 390 400 410
: Eastina + 913000 in m
Figure 1: Profile of the overlap region between two adjacent
strips.
4.1 Height relative offset
The relative height offsets are obtained by measuring the height
discrepancies between overlapping regions from adjacent strips.
Height offset can be computed between totally overlapped
footprints from the two strips, which hardly exist, or points
within a limited distance. Another way is to construct two
different horizontal planes in a flat area, one from each strip;
and compare these surfaces. Reflectance data can also be used
to match features between the two strips (Burman, 2000;
Vosselman, 2002), however, this approach is not always
successful especially with low-density data and large laser
footprints on the ground.
In the test data in this research, the percentage overlap between
adjacent strips was designed to be about 30% of the swath
width which is about 200m. However, due to the real
conditions during the data collection, such as wind, overlap
areas between strips ranges from less than 30m to as wide as
100m. The test data the data consists of 14 strips. Therefore, 13
overlap regions were examined in this research to quantify the
relative height discrepancy. Each region has a length of 1,500m
and they are oriented in the North-South direction as shown in
figure 2. The fact that the data is fairly dense (one spot height
per square meter) and the overlapped regions included in the
testing are large increases the likelihood of having coincident
and closeby data points. Therefore, the relative height accuracy
was obtained in a straightforward approach by computing the
difference between totally or partially overlapped data points.
Strip 1
overlapped region: (1,2).
Strip 2
overlapped: reglon (2,3)
Strip 3
overlapped region: (3,4): :
Strip 4
- ‘overlapped regloh (4,5) :
Strip 5
overlapped region (5,0
Strip 6
Figure 2: Data strips and overlap regions between them.