International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
building area [m?] | height [m] | length [m] | width [m]
type
One- 90-115 3-8.5 13-15
family
house
row-house 60-110 3-8.5 8-12
small 90-160 6-12 10-17
more-
family
house
large 140-280 10-18 14-24 10-16
more-
family
house
tower 280-900 27-60 20-75 17-30
block
Table 3: Adapted values
This method needs a lot of manual processing and therefore is
very time-consuming. Furthermore, it has to be adapted to
different settlement types of a city, e.g. the values differ from
city centre to suburban area.
4.2 Clustering
In the second approach we use an unsupervised classification
method, namely clustering for the automatic search of groups
with similar attributes. Instead of adjusting the values to
classify most of the buildings, all buildings are used and every
building is assigned to a cluster.
[n this step we used the program package WEKA (WEKA,
2000). The table 4 shows the mean value and the standard
deviation for some clusters.
After all the buildings are assigned to a cluster, each cluster has
to be assigned to the appropriate building type. This is less
time-consuming than the iterative method described in 4.1.
Cluster | Height Area Length Width | Building
[m] [m?] [m] [m] Type
Mean Mean Mean Mean
(StdDev) | (StdDev) | (StdDev) | (StdDev)
0 3.78 20.42 5.85 3.74 Garage
(1.04) (3.08) (2.64) (2.10)
1 10.58 253.69 19.94 15.23 LMFH
(2.07) (19.44) (1.56) (0.90)
2 14.31 177.74 18.10 10.36 LMFH
(2.40) (15.85) | (0.92) (0.89)
3 9.82 174.50 17.20 10.30 LMFH
(1.10) (14.81) (1.19) (0.70)
12 7.33 125.10 13.10 10.60 SMFH
(1.47) 0561) 10:37) (0.71)
13 8.46 91.64 10.74 8.96 SMFH
(0.40) (4.91) (0.39) (0.15)
18 41.27 368.42 25.65 18.39 Tower
(4.79) (13.70) |(1.41) (0.91) block
Table 4: Cluster and assigned building types.
This method also delivers a more detailed building typology
because different characteristics for the same building type are
considered.
718
Figure 4 shows an example of the clustering On the left are
blocks of large more-family houses. In the upper left corner are
one-family houses (pale red). On the right are office buildings
and stores.
Len, S um NS ee i, LU p Ta, c fd
€ T oi MV. abs Be Be, © < AREAS
A PR Bret CIS Mq.
ia Le LEA AI uf e 2 s "A
- “9 MAS ve v :
" PS E. P A | 7 M €
in ads A VAM 5%,
Bee AN x.
$ De Ç 3 g :
Se EN 5 504 px
it a +2 Se p, A
d C CN v 22
3 00039 SE) C d ANS GA y
de) æ vo a. AN S
d Ss c A
VAN NOS
WO wo DR 9
S
2
= 2m
EKD AN
A E
Figure 4: Results for the city centre of Stuttgart.
It also reveals that there are larger areas with the same building
types.
5. IDENTIFYING SETTLEMENT TYPES
The settlement types are composed of certain building types
and they vary in density and the arrangement of buildings.
To identify these different settlement types, clusters of objects
with similar properties and characteristic spatial distribution
have to be found. One possibility to do so is to use spatial
clustering (Anders, 2002). Here, not only the similarity between
the object is taken into account, but also the similarity in the
spatial density. Also here, however, the final assignment to a
certain settlement type has to be done manually after the
automatic clustering.
The other possibility is to use information from ATKIS. The
streets separate the surface into many small areas. Mostly these
small areas coincide with the building blocks. Furthermore,
these areas are assigned specific settlement types. Depending
on this type and further block characteristics, e.g. building
types, density or average distance to road, the assignment t0
settlement types from the settlement typology can be done.
Then neighbouring blocks with similar characteristics can be
merged.
6. RESULTS
Figure 5 shows the result of our method using laser scanning,
ground plans from ALK and specific heat coefficients from the
building typology. Figure 6 shows the heat demand from af
existing heat atlas. The values differ about 10 to 20 per cent.
Internati
i s
MV
& — ; - | f" MW o 2^".
Figure 5
AF où Uu e A A E NS ASS
. Figure 6:
