BJECTS
Prudente, SP,
. The method is
| representation
relevant image
connected road
Is. Experiments
extracting road
ons. Section 2
ic road segment
idiometric and
| are presented
Ire perspectives
IC ROAD
(traction that is
bject is a local
| fragment. The
Xther according
i
id the way they
described with
line segments
ristics that are
in any road
onvention the
e superior one
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
is called candidate. For each road object case, the endpoints of
both straight line segments (base and candidate) are
orthogonally projected from one to each other, resulting only in
two points projected between endpoints. For example, in the
figure 1(a) the endpoints of the candidate straight line segment
are projected into two points of the base straight line segment.
The opposite occurs with case 2 (figure 1(b)). In relation to
cases 3 and 4, as respectively illustrated in figures 1(c) and
1(d), only one endpoint of a straight line segment is projected
between endpoints of other straight line segment, and vice-
versa. In all cases, two end points belonging to the base and/or
candidate straight line segment and two projected endpoints are
combined to build quadrilaterals very close in shape to
rectangles. Each road object gives rise to a quadrilateral, being
each one identified as crosshatched area in the figure 1. The
axis of each quadrilateral coincides with a short road centreline.
The building of the four road objects is based on a rule set
constructed from a priori road knowledge. The main rules used
to identify and build road objects are described below:
. Anti-parallelism rule: According to this rule, two image
gradient vectors taken at two opposite road edge points, and
belonging to the same road cross section, are in
approximately opposite directions. Beside this, they are
approximately orthogonal to the road edges. This also means
that if the road edges are approximated by polygons, the
image gradient vectors computed at edge pixels fitted to each
straight line segment (of a polygon) are close to parallel.
Thus, a compact and effective representation for the image
gradient vectors, computed for each straight line segment, is
the mean image gradient vector;
. Parallelism and proximity rule: by this rule, two straight
line segments, base and candidate, are compatible to a road
object if they were approximately parallel and sufficient
close to each other;
. Homogeneity rule: the road pixel grey levels do not vary too
much, at least within short road segments. Thus, the area
inside each quadrilateral must be approximately
homogeneous;
4. Contrast rule: roads usually contrast sharply with the
background, meaning that each road object quadrilateral and
its background must show a high contrast;
.Superposition rule: a base and candidate straight line
segments are compatible if only if two of their endpoints can
be orthogonally projected onto each other. It is just this rule
that gives rise to four cases of road object depicted in the
figure 1. For example, in the case 1 the two endpoints of the
candidate straight line segment are orthogonally projected
onto the base straight line segment, giving rise to the
quadrilateral of road object of case 1;
. Fragmentation rule: as roads are usually smooth curves,
polygons composed by short straight lines are not usually
related to roads. For examples, image noise can generate
short and isolate polygons. However, parts of polygons with
very short straight line segments can be extracted from a road
crossing where the curvature is much more accentuated.
Another case is related to very perturbed road edges (by
shadow or obstruction, for example), which may give rise to
many short straight line segments connected to form a
polygon. In these places the road objects are difficult to be
formed, as the two first rules are hardly satisfied at all. Thus,
cases involving short straight line segments are not
considered and possible extraction fails (for examples, road
crossings not extracted) are left to be handled by other
strategies, which are based on previously extracted road
bo
[o]
wn
c
segments and other road knowledge, as e.g. context - relation
between roads and other objects like trees and building.
The order of application of the rules presented above is
important, mainly when the base and candidate straight line
segments are incompatible, as it can avoid in most cases the
verification of all rules for road object construction. The first
rule to be applied is the sixth as it allows parts of or whole
polygons potentially not related with road objects to be
eliminated. The next rule to be applied is the fifth, avoiding the
use of another set of rules in the case this rule is not satisfied. In
the following, the order of rule to be applied is the 2™ rule, the
1* rule, the 3? rule, and the 4" rule. A road object is accepted if
all rules are satisfied.
2.2 Road Segment Extraction by Grouping Road Objects
As described above, the road objects are constructed by
combining the base and candidate straight line segments, which
in turn belong to polygons representing all relevant image
edges. Each road object is a local representation for the longest
straight segment of a road segment. Thus, the problem we have
in hands is how to connect the road objects to form the road
segments.
2" Case | | 2" Case |
1* Case
3" Case
(a)
2" Case 1* Case 1* Case
3"! Case
(c) (d)
i
2" Case
Figure 2. Connections between road objects
Figure 2 shows the possible connections to the left and to the
right between the road objects. Figure 2(a) shows that 1* case
road object can connect to the left with the 2"* and 3 cases and
to the right with the 2" and 4™ cases. The 2" case road object
(figure 2(b)) can connect to the left with the 1* and 4" cases
and to the left with the 1* and 3" cases. Note that the 3" and 4™
cases (figures 2(c) and 2(d) respectively) can connect
themselves to both the left and the right cases.
[n order to construct a road segment by combining road objects,
two polygons are selected and their straight line segments are
combined two-by-two and the resulting road objects are
connected sequentially. The advantage of using the connection
rules is that the construction of any new road object is limited to
one or two cases (figure 2). The great problem of the polygon
combination is the large search space if no heuristic is used. For
high-resolution images, an efficient way for drastically reducing
the search space is to use strategies based on the space scale,
which allow the elimination of most part of previously
extracted polygons (Baumgartner et al., 1999).
