: : ; aor ; : : : ; : Interna 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 — 
The paper also addresses the specific issues related to the 5 rte | x apu 1a FLIGHT PATH 
identification, extraction, and tracking of forest fire hot spots : 
from the georeferenced LWIR images. The implementation of 
geo-referencing and its need in this system as well as some of 
the issues encountered while implementing the real time direct 
geo-referencing of the video is discussed. Finally a preliminary 
evaluation of the system's accuracy in real time through field 
test results is presented. 
  
Backgr 
HOTSPOT CENTER OFFSET 
    
  
  
  
  
1 Figure 2: Hot Spot Characteristic Diagram 
Figt 
The inset image is an enlargement of the hotspot from the 
2 background image. The background has been enhanced for 
3 clarity while the inset is the actual pixel values. At the flying The feat: 
altitude that this image was taken from, the fire pit imaged here from the 
would only be 1 pixel in size, but the hotspot from the fire in (not desc 
4. that pit is several pixels in size, because of the blooming effect extractior 
stated previously. The right side of Figure 2 shows the effect of peak tem 
the hot spot being obscured. The imager can see some of the for peak | 
heated air around the fire, but not the exact centre of the un- extracted 
obscured hot spot. As a clearer path to the hot spot occurs, a character 
> brighter hot spot is imaged. This will cause some error in the of grav 
actual centre for the hot spot. By monitoring the location of the sonespor 
8 brightest pixel within the identified hot spot, it may be possible These fez 
to weight the 3D intersection process, and also to potentially the proce 
separate actual hot spots from strong reflections and low level 
7 thermal sources such as roads or rock. 
2.2 Hot 
aids 1 The effect of an obscured hot spot is shown in the following : 
i t 3 EE By using 
I. Signal Synchronization Board excerpt from a video sequence, as depicted in Figure 3. Note features 
2. AHRS- Attitude and Heading Reference that these are inverted images, such that hotspots are black. To foxtures's 
System properly identify and locate small forest fire hot spots from tum todf 
3. Video Camera thermal imagery, a three-step process has been defined, as 
4. WADGPS Receiver explained next. Fire is .c 
5. Thermal Camera when the 
6. Inertial System the direc 
7. Data Logger may onl; 
various 
Figure 1- Hardware of the F* System accuratel 
Only the 
presentec 
2. EXTRACTION, IDENTIFICATION AND LOCATION 
OF FOREST FIRE HOTSPOTS FROM THERMAL N LEN f 
IMAGES Figure 3: Extracted Sub Image Highlighting oso 
A forest fire hot spot is hotter than its general surroundings. (Note EE NE A track unt 
Radiated heating takes a lot more time than convection and en > MR frames. 
conductive heating, and is dependent on several factors, maximur 
including the absorption rate, the heat capacity of the object, the 2.1 Initial Feature Extraction from Thermal Images Was CCI 
distance between the objects among others. For detecting fire, 
thermal imaging detects fire by the radiated energy. With a The first step of the process is to extract any features that are Figure 3 
small forest fire at the base of trees, and surrounded by several hotter than the general background. Figure 4 outlines this step. ug dá 
trees, the radiation is absorbed by these surrounding trees and A low-level threshold is set to identify any pixels and object au y'o 
not transmitted through. This leads to the fire "hot spots" being i ; for one 
that are of above average temperature from the background 
scur casionally b S ing forest. ; a : F Ë 
obscured occasionally by the surrounding forest correct image at the top left of Figure 3 creating a feature mask. 
value of 
what ide 
Also note there is a blooming effect around the fires. This 
occurs from the immediately surrounding air being heated. 
What this can cause is an incorrect determination of the hotspot 
center within the image when combined with the actual hot spot 
being obscured. Figure 2 is used to help explain these two 
characteristics.