Chen Yuen Teoh
Attributes
Pictorial Illustration
Property of Road Model Addressed
E
Separation between a
pair of edges
————————
D
+X
ZN
Seperation = x
Deals with basic road properties that a
road segment must have a minimum and
a maximum of width
Difference in Spatial
direction
E E
dir y
Difference = (dir x - dir y)
The road property that a road segment
must be running in parallel direction or
near parallel direction
Difference in Gradient
direction
oradient dir X
—
gradient dir x
Difference = (gradient dir x -
gradient dir y)
The road property that a road segment
has opposite gradient directions
Average enclosed
intensity
Intensity —
Average Intensity
= avg(Intensity of all pixels
within the enclosed
boundary)
Image property for road model, by having
smooth higher intensity from border
to border. within a road segment
Difference in
intensity with the
surrounding
Intensity y
Intensity x
Difference = (intensity x -
intensity y)
(a) Level
Addressing the fact that road surfaces
have higher colour intensity than other
surface material
—
Attributes
Pictorial Illustration
Property of Road Model Addressed
Minimum Seperation
Minimum Seperation =
max(x, y)
Deals with basic road property that
2 linkng road segments should have
minimal separation, normally caused
by noise.
Gap Difference
Difference = max(x, y)
Deal with basic road properties that
2 linking road segments should not have
big difference in width.
Spatial Direction dir a dirc | Deals with basic road property that
ee »- € t .
Difference 2 linking road segments should have
Er c — — | similar spatial direction, i.e.: changes is
dir b dir d angle or direction should be minimal.
Difference =
max((dir a-dir c),(dir b-dir d))
Width Difference Ep T TW Deals with basic road properties that
Ix Y y 2 linking road segments should not have
MN ES ——
wide changes in width ..
Intensity Difference
intensity x
intensity y
Intensity — (Intensity x -
Intensity y)
(b) Level 2
Figure 2: Features targeted at Different Levels
Exp. | Combination Level | L1 | L2 || Figure
Z| LU/LI/LZ/L3 NE | NE | NE #
Preprocessed 6481 3 (b)
CL/No/No/No
CL/ML/No/No
CL/ML/ML/No
WN —
2644
2644
2644
252
252
2 (a)
2 (b)
86 2 (c)
Deals with image property of road
model that linking two road segments,
the intensity of these two candidates
should have similar intensity, i.e.:
minimal intensity difference.
Table 1: Result SummaryML: Machine Learning; CL: Clustering; Li: Level i of the System; NE: Number of Edges
884
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.