International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
we | 4.2 Building Modelling 5. EXPERIMENTAL RESULTS
the
ich | Modelling aims at extracting a three dimensional precision The proposed approach for data processing was not fully
lap | model of a building from the geometrically registered point implemented yet and hence only limited parts were verified
a clouds and images. The overall modelling processes are based on the relevant experiments. We summarize the results
ive | presented in Fig. 5. We first extract primitive features from the from acquisition, processing (mainly focusing on registration)
the raw data and produce derived features from the primitive and evaluation.
Gen features. From points, we can extract patches by grouping
points into sets of spatially coherent points and derive edges by 5.1 Test Site
the intersecting each pair of adjacent patches. From images, we can
ject extract edges using an edge operator such as the Kenny operator Fig. 6 shows the building used for the experimental results. This
and derive patches by grouping the edges to construct closed building comprises a complex structure that retains different
polygons. It is noticeable that the edges and patches can be appearances for each side.
is extracted independently from points and images. So, we can
m refine the edges and patches by comparing those generated from
85 points with those from images. The refined edges and patches
dis are grouped into a polyhedron representing a geometric (shape)
3 model. Each facet of the polyhedron is draped with the texture
derived from images. This draped polyhedron is the final result
und of these building modelling processes.
tem;
era: Particularly, extracting patches from points and grouping the
patches into polyhedron models are based on the three
d by dimensional perceptual organization processes developed by the
f the author in the previous research (Lee and Schenk, 2001a; Lee
and Schenk, 2001b; Lee and Schenk, 2002; Lee, 2002).
(a) Air vie
Points Images
extraction extraction
Patches Edges
derivation refinement derivation
JL Patches
(b) Eastern side
grouping Figure 6. Views of the building used for the experimental tests
t Wc | | 5.2 Acquisition Results
th the Shape Texture We used a terrestrial laser scanner (mode number: ILRIS-3D)
tment. | manufactured by Optech. Its main specifications are
Ngo summarized in Table 1. Based on these specifications, we
sie draping |. — — — — — — —31 expected about = 10 mm of accuracy for the acquired point
meters coordinates.
|- linear |
elus del Maximum range To 4 % target 350 m
ius Mode To 20 % target 800 m
b. 3 Range resolution : 10 mm
ground Figure 5. The overall processes of building modelling Roots e © Som ; mm
aes the : e > = = Spot spacing @ 50 m <10 mm
ground Spot position X-Y @ 50 m +7 mm
: sets of accuracy Z @ 50m +10 mm
clouds Field of view 40° (420°)
e to be Data sample rate 2,000 pts/s
987
Table 1. Main specifications of the laser scanner (ILRIS-3D)
