X-B4, 2012 
  
ti-normal direction space 
r point by using 
en the basic unit 
Its. First possible 
he inner space of 
lyse whether the 
it belongs to the 
o evaluate the 
n progress could 
is named as the 
riangles of the 
ion requires this 
, the intersection 
lyhedron need to 
ibout the slope 
ing triangle and 
at the accessible 
is an acceptable 
rtical movement 
  
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B4, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
Polyhedron B 
  
   
Intersecting Triangle 
Plane of intersecting triangle 
  
Unit a 
  
  
  
Horizontal plane 
Figure 7. Accessibility evaluation of the unit belonged to the 
inner space of the building polyhedron 
4. EXPERIMENT 
The experiment area is the LaGuardia Airport in the City of 
New York. This area contains large groups of connected 
buildings, and as a public area it needs a proper evacuation plan 
under emergency situations. To provide the communication 
data for this simulation, we decide to generate the evacuation 
data from an existing Sketch-up file, and the skeleton of this file 
is shown in figure 8. 
   
Figure 8. Polyhedron-inner side evaluation for point by using 
the triangle plane 
After applying our extraction solution to the experiment area, 
ih Successfully generate a collection of accessible positions for 
¢ airport. To provide a general data view, the accessible 
position extraction result is shown in table 1. This table shows 
the building group is divided into 9 parts to finish the extraction 
task. The division of the whole area is expected to help 
accelerate the extraction analysis process. These building parts 
vary significantly both from size and direction. The largest part 
contains more than 16,000,000 units compared to the 7480 units 
in the smallest part. Furthermore, we provide the accessible data 
of part 3 in figure 9. 
  
Building Part Start and End Communication 
ID Position Unit Number 
  
3890.39241.15174.5700 7480 
: ge m 
3 Uem eR 
4 345,82880.56696.1506 2972490 
s Ui Smaezme 3931200 
6 i sus2Me Tess 
-607. 21- 
7 oso Se 
s ender 12066 
9 ores ou 
  
Table 1. The Extraction Result of the Experiment Area 
  
Figure 9. Extracted accessible positions in one region 
There is one more thing that should draw attentions. The time 
cost of the proposed method is depressive, and it uses about 90 
days to finish the extraction task. This long time period makes 
our solution not appropriate for some fast emergency response 
applications, and certainly we need to overcome this problem in 
the future work. We plan to use high frequency multi-core 
CPUs with multi-thread capability to finish this task. Since our 
current test platform only has a CPU with 2 cores and 4 threads 
capability, we could use two CPUs both with 8 cores and 16 
threads to reduce the execution time to less than 10 days 
theoretically.