stanbul 2004
165 18
ero crossing
1aximum
enter of gravity H
res hold
onst. frac.
110 120
umed effective
ops on rough
gger-pulses.
0.1
0.01
(in m) for the
obtained by
timated travel
m. Detection
ir = threshold,
we assume a
beam passes
then through
round. Again,
; however, the
s wide spread
gravity tend to
and maximum
r to maximum
te that in this
1g are able to
although their
).
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
0.2
0.15
brancht 4 L branch2 *- ground
o 01
©
0.05
i tree
ee bush —
0 = i I de 1
0 3 6 9 12 15 18 21 24 27 30
distance (m)
0.4 E I T T T: T T T
+ zero crossing
* maximum
B 0.3H| * center of gravity
a O threshold
2 v const. frac.
S 02r
©
5
wo! NS.
0 i L 4 1 { 1
0 20 40 60 80 100 120 140 160 180 200
time (ns)
Figure 4. Return pulse of a tree. Top: Assumed effective
scattering cross section of tree, including branches, low
vegetation and ground. Bottom: Reflected signal and derived
trigger-pulses.
In the third and final experiment, we investigate the detector
performance on a waveform resulting from the interaction of the
laser beam with a tilted roof, assuming a tilt angle of 45° and
pulse diameter of 1 m. As can be seen from Figure 5, top, the
scattering cross section has the form of a half-ellipse (the effect
of the tilt is to “stretch” the circular footprint into an ellipse; the
cross section is thus proportional to the width of the ellipse -
measured along the minor axis — as function of height.) Here,
the best result is obtained by constant fraction. This example
demonstrates, in particular, that zero crossing, although it
excels at resolving narrow, sharply peaked components of the
cross section, performs less well in the case of broad, plateau-
like maxima (such a constellation will, in general, lead to
premature triggering). Remarkable to note is that even in this
simple case the range values obtained by using different
detectors may vary by ~ 0.4 m (max compared to centre of
gravity), which is a large number given that the laser footprint is
Im.
1 i i i 1
5 9 10,5 12 13.5 15 16.5 18
distance (m)
I
Zero crossing
maximum
center of gravity H
threshold
const. frac.
x 4 +
4c
signal amplitude
50 $0 70 80 90 100 110 120
time (ns)
Figure 5. Return pulse of tilted roof. Top: Assumed effective
Scattering cross section of tilted roof. Bottom: Reflected signal
and derived trigger pulses.
4.3 Discussion
As illustrated by the above experiments, there is no such thing
as a single best detector; rather, the relative performance of the
detectors depends on several factors, such as the characteristics
of the effective scattering cross section, object distance and
noise level. For example, although zero crossing yields
excellent range resolution (discrimination between nearby
objects) and range accuracy under ideal conditions, it is
susceptible to noise (spurious trigger pulses) and slow gradients
(decrease in accuracy). Threshold may yield good accuracy for a
single object at a given distance, but poor range estimates (or no
trigger-pulse at all) if the amplitude of the back-scattered pulse
changes (i.e., due to an increase in object-sensor distance or
absorption/reflection of pulse energy by other objects).
Maximum, while not as accurate as zero crossing, gives
reasonably good results and can be expected to be more tolerant
against noise than zero crossing. Constant fraction seems to be
a good compromise between zero crossing and maximum.
The experiments presented in this section have mainly
qualitative character and are intended to illustrate the
dependency of the backscattered waveform on the scattering
cross sections of the illuminated objects, and to highlight the
respective strengths and weaknesses of various approaches to
pulse detection. Clearly, more elaborate experiments, both on
synthetic and real world data, will have to be conducted in order
to gain a better understanding of the physical principles
underlying waveform generation as well as the effects of noise
and scanner characteristics on detection performance.
5. OUTLOOK
In anticipation of the potential of full-waveform lasers for
vegetation mapping experimental systems have already been
built and tested by NASA (Blair et al, 1999). Soon, also
commercial full-waveform systems will become available.
RIEGL Laser Measurement Systems GmbH will soon offer a 2D
laser scanner for airborne applications (model RIEGL LMS-
Q560) with an optional data logger capability for recording the
digitised waveforms of both the transmitted laser pulse and the
echo signal (Riegl 2004). Laser pulse repetition rate is 50 kHz
and maximum range is typically 1500 m on targets with 80%
reflectivity. Sampling of the echo signal is carried out
simultaneously in two 8bit channels in order to cover a dynamic
range of about 40 dB optically. Sampling rate is 1GSamples/sec
for each channel. Samples within a configurable range gate
centred around detected targets are stored on redundant large
volume drives together with time stamp and scan angle data.
The maximum number of targets and the number of samples per
target to be logged can be defined by the user within limits
defined by the writing speed of the data logger and its capacity
limitation. Single target measurement accuracy is about 2 cm (1
sigma value).
As for each laser measurement also a fraction of the transmitted
laser pulse is sampled, the pulse shape of the transmitter pulse
can be evaluated from numerous measurements with high
resolution as shown in Figure 6. Pulse width of the sampled
pulse is about 4.5 ns (at 50% maximum amplitude) which is the
result of the convolution of the laser pulse and the pulse
response of the receiver. This measured pulse shape can be used
advantageously as the basis for the analysis of the echo signal of
complex targets.
