Geography Reference
In-Depth Information
feature within the footprint and locating that elevation at
the centre of the footprint. Although this may not create
large errors when comparing measured and modeled
water-surface elevations, it adds bias on the river section
with over estimation of bottom elevation. Using these
sections for hydraulic modeling can affect our ability to
predict velocity conditions, critical to riverine habitats.
Consequently, there is also a benefit to more accurately
defining meso-scale topographic features of the river
bed for volume comparisons related to morphological
changes. This can be achieved by using a narrow footprint
(NFP) system, such as the EAARL system, or as the four
X-Y-Z points obtained from a single footprint with the
HawkEye (Figure 7.3).
Consequently, a better altimetric (decimetric) and
planimetric (metric) accuracy is obtained in LiDAR data
than in photogrammetry data (Baltsavias 1999).
7.5 River characterisation from LiDAR
signals
7.5.1 Altimetryandtopography
7.5.1.1 Emergent terrain topography
The last returns, or echos, of an emitted laser pulse do not
necessarily reflect from the ground, especially if the target
is dense. In addition, some negative outliers originating
from multipath reflection may be positioned below the
ground surface (Kobler et al., 2006). Consequently, it
is necessary to mathematically separate the bare-earth
reflection from other reflections to produce digital ele-
vation models (DEM). As shown in Figure 7.5a, the last
echoes are first extracted from the entire point cloud, and
they are filtered to keep only bare-earth echoes. These
latter points are interpolated to produce a raster DEM.
Several filtering techniques have been developed that
can be summarised in four groups (Sithole and Vos-
selman, 2004). The first group, called 'slope-based',
compares the slope between two points to a given thresh-
old. One of the two points is then classified as an object
if the slope exceeds the threshold or as bare-earth if it
does not. The second group, called 'block-minimum',
directly produces a raster DEM by selecting the mini-
mal height value of points in each raster cell. The third
group, called 'surface-based', constructs a surface from
a minimum number of selected points. The surface is
iteratively made more complex by the addition of succes-
sive points selected by certain distance criteria. The main
method belonging to this group, called 'iterative TIN', is
the most commonly used method for producing a DEM
from LiDAR data (Axelsson, 2000). The last group, called
'clustering/segmentation', studies the positions of point
clusters by comparing them to their neighbours to detect
objects with noticeable edges and elevations. A raster
DEM is then computed using interpolation techniques
(kriging, inverse distance weighting), with the splines give
the best results (Brovelli et al., 2004).
In general, all of these filters produce satisfactory
results for low complexity landscapes (low slope, sparse
vegetation, small buildings). However, landscapes with
bare-earth discontinuities, such as mountainous areas
or
7.4.2 Geodeticpositioning
Ranging LiDAR-delivered data are usually X,Y,Z points.
These geodetic positions result from a geometrical chain
that includes several processes from optical and electronic
time delays up to aircraft attitude and location.
An accurate distance measurement between the LiDAR
system and the target is needed, which results from a time
waveform or pulse transform as explained in Section
7.2.1.1. The calibration of the timing system is to sub-
nanosecond accuracy because a 1-ns error in time results
in an altimetrical error of about 15 cm.
The LiDAR system positioning in time as well as laser
beam angle must also be precisely determined. This is
done by first using Kinematic GPS (KGPS) onboard, with
the on-the-fly (OTF) technique, carrier-phase ambiguity
resolution and correction with a near-ground base DGPS
station. This step prevents erroneous initialisations, and
it provides highly precise (sub-decimeter accuracy) hor-
izontal and vertical positions for aircraft with respect to
the WGS-84 ellipsoid. Additionally, the aircraft's attitude
and, thus, the laser beam angles are collected in time
through an inertial measurement unit (IMU) for pitch,
roll, and yaw registration. As Guenther (1985) has stated,
'at a 400-m altitude and with a nominal 20-degree nadir
angle, a system angle error of 0.05 degrees (
1 mrad),
which equates to a nadir angle error of 0.10 degrees,
would yield a 25-cm error in the vertical height of the
aircraft'. If 0.01 degrees is the acceptable limit for angle
accuracy, ' this can only be accomplished by applying
an inverse algorithm to flight data collected occasionally
for the purpose of angle calibration' (Guenther et al.,
1999). Moreover, as GPS and IMU collect data at a 1-Hz
frequency, they need to be properly interpolated in time
to the laser frequency.
<
steep
riverbanks
still
pose
problems
for
filtering
algorithms.
Search WWH ::
Custom Search