rt B2. Istanbul 2004
(1)
of the feature point in
on vector in mapping
n the body frame and
en the camera frame
m vector
e obtained from the
» for hot spots are
iction process in as
are fixed values that
nd Lr are obtained
b
item, an airborne test
1 Calgary in July 30,
vere set in fire pits at
stem was mounted in
ht services provided
Two multi pass test
ce over the test filed
both day and night
> fires that were used
verage of that target.
the right hand image
ly see the fire.
st Canopy Coverage
the aircraft. derived
ar real-time systems
trajectory for these
enced GPS (DGPS)
olution was derived
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
from the University of Calgary KINGSPAD'" software
(http://www kingpads.com/), El-Sheimy and Schwarz 1998).
The table clearly indicate that positional accuracy of the aircraft
is in the 1-2 level for both the WAAS and OmniStar systems.
Night Flight Day Flight
WAAS | OmniStar | WAAS | OmniStar
Mean 1.761 3.124 2.292 2.850
min 0.294 0.715 0.504 0.706
max 8.782 10.05 12.44 12.27
Std. Dev. 1.339 1.179 1.579 1.477
Table 2: 3D Error Statistics
Figure 8 show the real time attitude results for one of the test
flights. The reference attitude for these figures is the post
processed attitude derived from the University of Calgary
KINGSPADTM software. The figure clearly indicates that the
attitude accuracy is within 0.1 — 0.15 deg
Degrees Difference
05
20000 25000 30000 35000 40000 48000 50000 85000 60000
Video Frame
Figure 8: Night Flight Post Processed to Real Time Attitude
Comparison
Utilizing the extraction and identification process outlined in
Section 2, Table 3 and Table 4 lists the positional accuracy of
the F?D system in calculating the coordinates of 5 controlled
fire pits. The results in Table 3 were calculated using post-
processed georeferencing (position from DGPS) data, while the
results in Table 4 were calculated using the real-time
georeferencing (position from WADGPS) data.
Average Target
Altitude 1 2 3 4 5
Pass ] 404 m 2.4 8.1 2.1 4.6 1.8
Pags2 | 359m 7.3 8.4 5.0 1.8 6.0
Pass3| 942m 6.5 8.7
Pass4| 961m 10.9 j.5
Pass5| 364m 11.2| 11.6m 7.5 8.2 8.6
Table 3: Post processed 2D Position Errors (meters)
Average Target
Altitude 1 2 3 4 5
Pass | 404 m 23m: 7.7m {-2.8m 54m | 29m
Pass 2 359 m 63m | 69m 39m El'm [5.0m
Pass 3 942 m 53m 7.3m
Pass 4 961 m 13.8 m 5.1m
Pass 5 364 m Um ll7m| 30m |117m; 90m
Table 4: Real Time 2D Position Errors, meters
The results in Table 3 and 4 are extremely promising
considering the altitude and the image sizes (image size is
320x240 pixels). The blank cells in the tables indicate that the
target fire was not automatically detected. It is interesting to
note that the results using post-processed data is not always
better than the real time results for the same targets. The 2D
errors are in a similar range for the particular target pass
combination suggesting that there are some tracking errors still
occurring. Some of the other possible error sources include
redundant and multiple features being cross-referenced to the
same feature track, differences between the feature centre of
gravity co-ordinate and the maximum intensity co-ordinate, and
incorrect geometric calibration of the non-metric thermal
imager.
4. CONCLUSION
The FD system presented in this paper represents a substantial
leap in the speed of reporting identified forest fire hot spots. In
particular, by utilizing the latest in WADGPS/IMU system
integrated with new computer technology and custom image
processing algorithms, the georeferencing of small forest fire
hot spots is definitely possible in real time. Preliminary results
strongly suggest that small hot spots in a lightly forested area
can be readily detected in real time to within a 2-10 m level of
accuracy.
REFERENCES
Brass, J. A., Ambrosia, V. G., Rieker, J. D. and Robert, A. C.,
2001. , "Forest Fire Fighting From Real-Time Airborne Infrared
Remote Sensing", www.airplatforms.com/features/pdf/Canberra
Firefighting.pdf (accessed 23 Apr. 2004)
Canadian National Forestry Database Program,
http://nfdp.ccfm.org/framesqf e.htm (accessed 23 Apr. 2004)
Davies, D., Palmer, P. and Mirmehdi, M., "Detection and
Tracking of Very Small Low Contrast Objects”,
http://www.bmva.ac.uk/bmvc/1998/papers/d135/h135.htm
(accessed 23 Apr. 2004)
Ellum, C., 2001, "The Development of a Backpack Mobile
Mapping System", MSc. Thesis, Department of Geomatics
Engineering, University of Calgary.
El-Sheimy, N., 1996,"The Development of VISAT - A Mobile
Survey System for GIS Applications", PhD Thesis, Department
of Geomatics Engineering, University of Calgary.
