3. FLOWCHART OF PROCESSING 
The data flow diagram of processing is shown as follows. 
  
= 
  
-——— —— wd P € 
  
  
— 
| [4.1] Data Acquisition from 
| | Airborne Laser Scanner | | 
  
  
  
e: — SE = 
Raw Image Raw Data | 
| ( Digital Camera ) | ( Random Points )J 
  
  
  
  
| 
1 
I 
1 
1 
1 
1 
1 
1 
I 
! 
| 
: 
mq —_— ! 
vx NM Y | 
| S x TR ra ae en ! 
| [4.2] Ground Filtering [ i T i 
| [4.3b] Removing Noise | | H3] sers | 
L Cr DEM? bom ee | 
i 
EN zT 40 1 
ET = TT TN ! 
e "um Eu m | 
MM Es | IET I 
Digital Surface | | Digital Elevatio | ! 
[. Model ( DEM ) J D. Mode ( DSM ) ; 
Y — COL | 
— A | 
| Ortho Image | 
| Creation | ! 
1 
  
po il 
[5.4] 
  
    
    
| Building Footprints | 
L zl 
  
-— ic 
  
[ 5. Building Outline Extraction ] i er fr 1 
  
(DSM-DEM) | 
Te 
m E 
| Laying Objects 
L Heigth Image ) 
NE Nee 
| [5.2] Contour Polygons | 
Calculation | 
| [5.1] Subtruction | 
  
  
  
  
" € 
| [5.3] Building Footprints | 
| Extraction | 
  
1 
Reshape fi —— eu 
| Rough | 
| Building Footprints J 
\ 
a a om a a ———————_——_—_—— 0 0 2 2 7 0 2 0‘ ‘0 “0 0 A0 0 8 
eus esse: [ 6. 3D-City Models Creation] ---------4---- 
a 
NES [6] Cre ano = 
Rm = reation of | 1 
Building Footprints | 3D-City Models | ; 
L d E E - zd ! 
e parit ; 
vy | 
~~ | 
3D City Model | 
Lt A | 
| 
  
  
  
  
4. PRE- PROCESSING 
4.1 Data Acquisition from Airborne Laser Scanner 
The measurement conditions when it takes with the laser 
scanner are shown in table 3. In the case of these conditions, the 
sampling interval of measurement and the number of points (all 
and each returns) within the scope of 1km? are shown in table 4. 
  
Measurement Area 
Shinjuku (JAPAN) 
  
  
  
  
Flight Altitude 9000 [feet] (Approx. 2743 [m]) 
Flight Speed 110 [knot] (Approx. 209 [km/h]) 
FOV 16.0 [deg] 
Scan Rate 19.5 [Hz] 
  
Table 3 Measurement Conditions 
  
Points Interval (Along) 
Approx. 2.5 [m] 
  
Points Interval (Cross) 
Approx. 2.0 [m] 
  
Number of Points 
233573[points/1km’] 
  
1% - 5" Return (Ratio) 
  
  
1% 83.6% / 2": 13.9% 
3rd. 2 206 / 455 0,2294 
5%. 0.03% 
  
  
4.2 Ground Filtering 
fallows. 
  
  
The ortho image is corrected using the DEM, therefore the high 
buildings are slightly lying on the ortho image. In order to 
reduce the influence to a minimum, the aero photographs with 
large overlap are used. And the central part of aero photograph 
is used to create the ortho image. It is hard for the extraction of 
the exact building footprints only from the laser point clouds, 
because the measuring interval of the laser points is rough. But 
in this study, the building footprints are corrected with the ortho 
image, and the 3D city models are created from the building 
footprints. 
Figure 1 Flowchart of the Processing 
Table 4 Characteristics of Random Points 
The laser point clouds are divided into tiles for the filter 
processing, and then the points are classified into the ground 
and the others based on the statistic values within the tile. The 
tile size is changed several times according to the size of lying 
object on the ground. The results of filter processing are as 
  
Figure 2 DSM (2m Grid) 
  
Figure 3 DEM (2m Grid) 
—222— 
  
JE bs A N 3] 7 — d