:o.kr
a. The proposed
pproaches is that
| accuracy due to
1e. The approach
simplification of
es the domain of
mising.
issifies edges to
based on their
ity, parallelism,
Tempfli (2000)
ng elevation as
The core part of
ogical filter for
1 segments. The
J Or vegetation.
rated a 3D TIN
data for building
ated to serve the
e vertical walls.
tric properties of
> building wire-
larted by finding
populate a plane
plane parameter
lane parameters.
nd used to form
od for building
roposed method
| level processes.
aches is that we
5) into raw laser
ind accuracy due
e adjacency of
processing time.
oer begins with
ion, grouping for
ition of building
ctor format. To
d above changes
ind. pseudo-grid
diagram of the
and extraction
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
laser scanning data
pseudo- grid
manufacture
low- level
process :
noise removal
segmentation
T
tree removal
building extraction
high- level
process
second specia
points extraction
final special
points extraction
boundary
linearizing
Figure 1. The schematic diagram of the proposed approach for
building detection and extraction
2. BUILDING DETECTION AND EXTRACTION
The proposed approach is divided into two processes: low level
and high level process. The low level process consists of
pseudo-grid generation, noise removal and segmentation. The
high level process consists of grouping, tree removal and building
boundary extraction. In addition, each step changes the domain
of input data such as laser point domain and pseudo-grid domain
in order to achieve efficient data processing.
Figure 2 shows the change of data domain in the proposed
approach for building detection and extraction.
pseudo- grid space
group space
high-level
process nf
pseudo- grid space
extraction
grouping
i E
pseudo- grid space
low- level segmentation
process noise removal
pseudo- grid
l =
Figure 2. The change of data domain in the proposed approach
for building detection and extraction process
2.1 Low-level Process
Pseudo-Grid Generation
In many previous research for building detection and extraction,
irregularly distributed laser scanning data are converted into
grid form so as to enhance speed of data and then building
extraction is performed. In doing so, unwanted errors are
introduced in the process of interpolation. In order to avoid the
errors, we invented a concept of pseudo-grid that virtually
contains laser point data in each grid form.
The size of pseudo-grid is calculated with average point density
of laser point data. Once pseudo-grid is created, the raw laser
point data is assigned to each pseudo-grid shown in Figure 3.
e © e. Liiardata
e Pseudo-G rti
Fig. 3. Pseudo-grid generation
Since we create pseudo-grid, we don't need to convert laser
point data into regular grid form and don't introduce any errors
into the raw data through interpolation. In addition, the pseudo-
grid improves the adjacency among laser point data so as to
speed up the process such as building detection and extraction.
Noise removal
There are irregular random errors contained in raw laser point
data caused by instrument malfunction, natural phenomena and
so on. In this paper, we only consider random errors such as
outliers and remove them by statistical method.
Segmentation
It is defined here that segmentation is only to extract building
candidate points from laser data point cloud. We applied a local
maxima filter for segmentation.
2.2 High-level Process
Grouping
The process of grouping is performed on pseudo-grid domain
and defined as classifying laser point data as a group resulted
from segmentation above. After grouping, we can compute the
area and perimeter length of each group, which will be used for
building decision criteria.
Tree removal
After grouping, the laser points belonging to trees still exist as
building candidates. Those laser points could be removed by
two simple measures: minimum building area and circularity.
However, some of laser points belonging to trees can't be
eliminated if their size and shape are similar to buildings.
