XIX-B1, 2012
Data
Data3D
ointCloud StereoModel
(b)
1etry (b) data
rocess through new
‘ation. In most cases
(barring case 2.1.1).
for formulating the
n to define inference
interpretation of the
knowledge schema
ene, with which the
ing. This prominent
es of algorithms to
hmic sequence.
JB detection
sification Method
knowledge schema
ture, and algorithms
selected. After the
> KB, the detected
bjects. As such, the
ithms best suited to
1 the more complex
ship to the simple
letected geometries
's of the objects to
| current version of
through semantic
point clouds have
005) and Munoz,
issociative Markov
Barinova, (2010),
, and Shapovalov,
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B1, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
Velizhev, & Barinova (2010) instead label them through
capturing geometric classes in context by designing node
features. However, these works do not incorporate semantics.
Koppula, Anand, Joachims, & Saxena (2011) present a more
semantic approach, whereby a graphical model that captures
features and their contextual relationships is presented. WiDOP
presents semantic rules to semantically annotate the objects
(Ben Hmida, Cruz, Nicolle, & Boochs, 2011). These rules are
executed through the extended SWRL, and use semantic
annotation to match the detected geometries with their probable
objects. The following example (equation 1) will detect all the
vertical geometries and annotate them as Mast if they are higher
than 6m.
3DProcessing. swrlb:VerticalElementDetection( ?vtr, ?dir) ^ Height(?x,
?ht) ^ swrlb:greaterThan(?ht, 6) -? Mast(?vrt) (1)
The domain ontology schema now hosts the first impressions of
the semantically annotated geometries. At this point the
annotations are still rough, and can be one of three types:
unambiguous, ambiguous or unknown.
Unambiguous: Geometries annotated to a single object.
Ambiguous: The same geometry can be qualified as two or
more objects.
Unknown: Geometries unclassifiable at this level of iteration.
The first iteration is likely to have a large number of
ambiguously annotated objects or even unknown objects. The
second iteration is needed to improve the result, wherein the
KB will now host more semantics. During the second iteration,
ASM uses unique characteristics to remove the ambiguity. The
mechanism under ASM investigates the rules which are unique
to each object in such ambiguity. It then uses these unique rules
to infer an algorithmic sequence for each of them. More precise
geometries are thus detected during this iterative stage and are
populated into the KB. The qualification through extended
geometries then repeats (equation 2).
BasicSignal( ?y) ^ BoundingBox 3D(?x) ^ hasHeight(?x, ?h)^
swrlb:greatThan(?h, 1) ^ swrlb:lessThan( ?h, 3) ^
3D swrib Topo:distance(?x, ?y, 100, 10) 2 SecondarySignal(?x) (2)
The iteration continues until all the ambiguity is removed and
objects are finally recognized and stable. In case of unknown or
ambiguous annotations, new knowledge about the scene or the
processing activities is fed into the KB. The first case (section
2.1) resembles the unambiguous annotations. The first level of
iteration is therefore unnecessary for this scenario. As objects
and their positions are known, the platform executes the
iteration from the second step and verifies.
3. IMPLEMENTATION
Figure 7 and 8 illustrate a typical site in the DB railroad system
and its 3D scan. The complexity in detecting objects in the
point cloud is not only due to the complex nature of the objects
but also due to the scan nature. The area is scanned using a
moving train; the objects are scanned only in one direction,
presenting challenges through occlusions.
93
Figure 7. A typical site of Deutsche Bahn system (source:
Christophe Leimkühler, Metronom Software)
Figure 8. The 3D scan data of the site
It must be emphasized here that we do not use the algorithmic
knowledge to estimate what objects should appear in the scene,
but rather the opposite. The characteristics of objects in the
scene evoke the appropriate algorithms for that object. In this
sense, the objects in the class DomainConcept are in the center
of the top level ontology schema (fig. 2). It hosts the semantics
of the objects first through hierarchical taxonomy (fig. 4) and
then through semantic rules for each specialized class under the
hierarchy. If we examine carefully, DomainConcept is a bridge
through which other knowledge domains can be explored. For
instance, the geometric characteristics of an object under
DomainConcept are related to the knowledge domain of
Geomerty through the relationship defining it (fig. 2). This can
also be extended to other knowledge domains, as we did with
data through class Data. The top level ontology in figure 2
provides a glimpse of such bridging and is not restricted to it.
3.1. Illustration
This section illustrates how underlying ASM within WiDOP
infers rules to derive an algorithmic sequence. We basically will
illustrate the principles discussed in section 2 through a case of
Deutsche Bahn (DB) with the underlying ASM in focus.
The property restriction rules play a major role in determining
the best algorithm. ASM determines this through inferring the
rules defined in DomainConcept (termed as DC in the DL
equations) to that defined in the class Algorithms. The platform
starts with the dominant rule of the scene. We presume the
dominancy through the number of occurrences of the rules,
with the higher the number, the more dominant the rule. An
example of this could the scene of a lecture room where most of
the objects have planar surfaces. In such cases the horizontal
plane detection algorithm will be preferred as a starting
algorithm.
This rule when inferred against specialized algorithms in class
Algorithms yield that algorithm HeightApproximation
(presented by HAA in the equation 4) is best suited for this case.
