XIX-B3, 2012 
  
tions image showing 
2 triple-stereo-scene 
.3 and 5 (blue, green, 
ly calculated top-of- 
] (Fig. 1) as: 
(1) 
RT R=) 
—d,d 
(2) 
—R,f* RN, 
—d,d 
(3) 
jx) (4) 
rom 0 to 1 the binary 
v » 0.89, s » 0.59 
TM) is derived from 
For this the DSM is 
ady small unmatched 
ical grayscale open- 
(i.e. 20 x 8 x 0.5m) 
y is used filling also 
rescaling to original 
. 3. Subtracting the 
so called normalized 
es the heights of ob- 
connection between 
ight. The red marked 
> following steps. 
te height (here: 3m) 
a closing (radius of 
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B3, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
  
Figure 2: Classifications derived from pan sharpened TOA im- 
age, left: water (blue), vegetation (green), sealed/soil (red), 
dry vegetaion (yellow), right: water (blue), vegetation (green), 
shadow (red), shadow--vegetation (yellow, mostly near trees) 
  
   
  
  
100 — 200 300 400 500 600 700 800 900 1000 
Figure 3: Left: Overlay of DSM (green), DTM (blue) and nDEM 
(red) of of Tunis area with depicted profile line (red); right: pro- 
files along profile line: DSM (red), DTM (green) and nDEM 
(blue) (heights in meter above WGS84 ellipsoid) 
structuring element: 2 px) gives the high objects from the nDEM 
(cf. Fig. 4, left). Combining the classification and these high ob- 
jects mask gives the classes buildings (high and non-vegetation), 
trees (high and vegetation), ground (all non high objects includ- 
ing water, vegetation and non-vegetation areas) as shown in Fig. 4, 
right. 
  
Figure 4: Left: High objects mask, right: final classification, red: 
buildings, green: low vegetation, dark green: trees, blue: water 
3 THREE-DIMENSIONAL CITY MODEL 
RECONSTRUCTION 
In this section, we introduce two different approaches for three- 
dimensional city model reconstruction. In the experiments sec- 
tion, We compare these two different approaches and discuss about 
their detection capabilities, advantages and disadvantages. 
31 Active Shape Growing Based Approach 
The active shape growing approach is our first 3D city model 
reconstruction approach. This approach is based on three main 
modules as: active shape growing for shape detection, detecting 
building rooftop type, and finally we use detected the shapes and 
rooftop types to generate 3D models. 
325 
3.1.1 Box-Fitting for Shape Detection For detecting com- 
plex building shapes, herein we follow a similar methodology as 
represented in the previous study (Sirmacek et al., 2011). How- 
ever instead of using a binary active shape growing approach in 
each seed-point location, we propose a novel active shape grow- 
ing approach based on usage of three-dimensional information. 
To do so, after extracting (x, ys) seed point locations as we de- 
scribe in (Sirmacek et al., 2011) in detail, we start to grow our 
active rectangular shapes in each seed point location by regarding 
the hight information. We assume that (17, y;) array holds the 
pixel coordinates for nth edge of the virtual rectangular shape. It- 
eratively, we sweep each edge to the outwards growing direction 
if the edge pixels satisfy (max (D(x}, yy) —min(zy, yy) < 0) 
inequality (n € [1, 2, 3, 4]). Here, the ó threshold is the minimum 
building height that we would like to detect in the region. 
In our application we assume J as equal to 3, which means that 
we assume the buildings to be higher than 3 meters to be de- 
tected. When the growing process stops for each edge, we cal- 
culate the final energy value by using the equation that we rep- 
resent in Eqn. 5. In the equation, m(.) represents the mean 
value. For the same seed-point, we apply growing process for 
all 0 € [0,7/6, 1/3, 1/2, 21/3, ..., 27] angles with 04; — 7/6 
radian turning steps. As we discussed in detail in (Sirmacek et 
al., 2011), by reducing 04;; step sizes, we can obtain more ac- 
curate approximations, however in this case we need more com- 
putation time. After calculating Ee for all 0 angles we pick the 
estimated box which shows the highest Fe energy as detected 
building shape. Since most buildings appear like composition 
of rectangular building segments, it makes sense to extract rect- 
angular shapes on buildings. The main advantage of using the 
box-fitting approach is that approximate building shapes still can 
be found even if the building edges are not well-determined, or 
even if there are trees adjacent to the building facades. However, 
other region growing algorithms fail to extract an object shape in 
these cases, since the growing region can flow out easily when 
the parameters are not set precisely. 
4 
Eg — Ax m(Bp(z,y) x Di(z,y)) — >) m(D(a3,40)) © 
n=l 
For complex buildings, after fitting a chain of boxes, discontinu- 
ity between adjacent boxes should be smoothed. For this purpose, 
we simply benefit from morphological operations. First, we start 
with filling inside of the detected binary boxes with 1 value in 
Bg(x,y) binary mask. Then, we apply morphological dilation 
and erosion operations respectively to the detected boxes, using a 
disk shaped structuring element with radius 1. After this opera- 
tion small discontinuity between adjacent boxes can be smoothed. 
Final improved building shapes are kept in new B(z, y) binary 
mask. 
3.1.2 Classification of Building Rooftops After detecting build- 
ing groundfloor shapes, we focus on reconstruction of rooftops. 
For this purpose, we benefit from our previous ridge-line detec- 
tion approach which is presented in (Sirmacek et al., 2011). 
The ridge-line detection approach is based on derivative calcula- 
tion over the DSM. We use the following derivative filter. For a 
symmetric Gaussian function G(z,y) = exp(—(z* + y^)), itis 
possible to define basis filters Gpo and Gpz as follows, 
Gon = Glo) = ~emn(- +97) ©