Information Technology Reference
In-Depth Information
(a) Edge detection on
R
component (b) Edge detection on
u
∗
component
Fig. 8.
Result of Canny edge detection on different colour components
Fig. 9.
Subgraphs given after matching procedure
(1.)
d
(
v
i
,v
j
)=
(
x
j
−
x
i
)
2
+(
y
j
−
y
i
)
2
<
5
,
(2.)
C
new
,...
(
x
i
,y
i
)=true,
(3.)
C
new
,...
(
x
j
,y
j
)=true,
(4.)
a finite path between
v
i
and
v
j
.
C
new
,...
indicates either
C
new
,r
or
C
new
,u
.
5
is a tolerance value, which depends
on the resolution and average size of the objects. We apply
5
= 30.
These conditions guarantee that only vertices connected just in the newer edge
map are matched. Like in the lower right part of Figure 9 two closely located
buildings are separated correctly.
We obtain a graph composed of many separate subgraphs, which can be seen in
Figure 9. Each of these connected subgraph is supposed to represent a building.
However, there might be some unmatched keypoints, indicating noise. To discard
them, we select subgraphs having at least two vertices.
To determine the contour of the subgraph-represented buildings, we used the
aformentioned GVF snake method. The convex hull of the vertices in the sub-
graphs is applied as the initial contour.
∃
Search WWH ::
Custom Search