The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008
transfer, static or dynamic alignment. For maximum operation
flexibility the static initialization can be constrained to a very
short time if the approximate heading is specified using external
information (i.e. magnetic reading). Nevertheless, the
implemented modeling uses a customized version of the large-
heading error model (Kong et al., 1999) and tolerates well
larger initial uncertainties. Hence, it is possible to completely
initialize or re-initialize the system in-flight without imposing
much restriction on the dynamics (even for a helicopter).
Figure 4: Information window of the GIINAV module
graphical interface
The inertial navigation cannot completely monitor the integrity
of GPS positioning, therefore different strategies are currently
investigated to apply SBAS, RAIM and RTK technologies in a
cascade form as suggested in (Skaloud, 2006). Hence, the real
time absolute positioning accuracy depends on the employed
positioning mode (absolute, differential code and/or phase) and
therefore ranges from meter to sub-decimeter level. The
orientation accuracy is less dependent from the positioning
mode and typically ranges from 0.01-0.03 degree in roll and
pitch and 0.05-0.10 degree in heading when compared to the
CP-DGPS/INS post-processed smoothed solution. A detailed
evaluation the GIINAV performance is presented in section 7.
4. RT GEOREFERENCING (LIEOS)
The role of the georeferencing module is twofold: first, to
generate the laser-point-cloud while on a flight-line, second to
analyze its quality. The first task is a real-time operation
handled by the LIEOS module, while the second is a delayed
process handled by the LIAN (LIdar ANalyse) element. LIAN
is not a standalone application but a separate thread of lower
priorities that is entirely managed by LIEOS. Its functionality
will be described separately in the following section.
The inputs to LIEOS are the LiDAR line data served by the
ALS Data Logger (Figure 3) and the trajectory served by
GIINAV, both at predefined data rates. On the output, LIEOS
stores all laser point-cloud coordinates into a file and transmits
points related to swath-characteristic (i.e., boarders and nadir)
to HELIPOS for displaying. LIEOS supports different
projections and datum, choice of which is usually influenced by
the datum and projection on the map used for pilot guidance.
15:41
41:
Info
: Start Georef©renein« Po
LlltS
15:41
41:
Info
: ALS
results file for lir
>e i succès
sfully created...
15:41
44:
GPS
I inte :
316766.318780,
Line
Count :
402
15:41
49:
GPS
I ime :
316770.793300.
Line
Count :
502
15:41
54:
GPS
T ime :
316775.401780,
Line
Count :
603
15:41
59:
GPS
T ime :
316779.919850,
Line
Count :
703
15:42
04:
GPS
T ime :
316784.393210.
Line
Count :
803
15:42
10:
GPS
Time:
316789.002490,
Line
Count :
904
15:42
15:
GPS
T ime :
316793.521480,
Line
Count :
1004
15:42
18:
Info
: ALS
results file for line 1 closed
15:42
18:
Info
: End
Georeferencing
Points...
15:42
18:
LI8N
: Start new zone...
15:42
20:
MAN
: Added 328055 points
15:42
22:
MAN
: Analyse of line 1
terminated
Figure 5: LIEOS dos shell
The georeferencing algorithms implemented in the LIEOS were
optimized to allow processing throughput of ‘tens of thousands’
points per second considering that the computational load per
laser-return is influenced by several factors as: the frequency of
trajectory output, the selected coordinate system, choice of the
the georeferencing algorithm. These factors may vary per
system or its setup (e.g., scanner rates may vary from 10 to 180
kHz, trajectory rates from 0.01 to 2 kHz) while the availability
of processing power depends on the distribution of individual
applications and the processor speed. Hence, to allow general
use of this application, three georeferencing methods were
implemented and their choice is left upon the user. These are:
• Fast ( < 1 m),
• Approximate (< 1 cm),
• Rigorous.
The ‘fast option’ is an approximate method of sub-metric
accuracy that is especially advantageous if the point-cloud is
requested in the geographical coordinates. Despite its name, the
‘approximate’ method provides residual distortions at
subcentimeter level only (in most flight scenarios) and
regardless of the terrain characteristics (Legat, 2006; Skaloud
and Legat, 2008). Its choice is especially advantageous, if a) the
output is requested in national coordinates, b) the ratio
scanner/trajectory sampling is relatively high. Finally, the
‘rigorous’ method is also optimized for speed, but uses no
approximations. It performs the calculation of the laser point-
cloud coordinates in a Cartesian system and then applies its
rigorous transformation to the specified datum and projection.
Although this method is more computationally demanding, its
employment within the presented system requires no more than
10-15% of the total capability of the on-board processor.
5. LIDAR DATA ANALYSIS (LIAN)
As mentioned previously, LIAN runs as a separate thread
within the LIEOS module. Its purpose is to analyze the quality
and the completeness of the gathered laser data. The ALS data
is only transmitted to LIAN once the actual flight-line is over.
By pressing the offline/online button on the HELIPOS-GUI
(see Figure 8), the operator communicates to LIEOS, if the
system is
a) Online: The raw laser data is stored; RT georeferencing is
activated and swath boundaries are sent to HELIPOS
b) Offline: No ALS data is stored; the georeferenced point-
cloud of the previous flight-line is passed as one block to
LIAN.
The temporal splitting of the two main tasks (i.e. the RT
georeferencing vs. data analyze) allows keeping the CPU
requirements at reasonable level.
Once the georeferenced point-cloud of a strip has been passed
to LIAN, the program computes a density grid based on the 2D
laser point coordinates (see Figure 6). The rasterized data
coverage information is further processed to compute
a) the complete data extend (outer bound of all strips within
one flight zone) and
853
