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 
The top level classes represent knowledge in their specific 
areas. We can see there are other top level classes beyond 
Algorithms and DomainConcept in figure 2. There are other 
knowledge domains having influences on these two main 
domain classes. Actually, WiDOP uses two such top level 
knowledge classes, however it is possible to extend if required. 
The interrelation between the knowledge domains completes 
the knowledge set. This provides the foundation for enabling 
various knowledge sources to interact and to use it for different 
purposes, like deciding upon a processing strategy (sequence of 
algorithms) or upon the nature of elements being found during 
processing. These inter-relations in knowledge terms are object 
properties and shown through the arcs in figure 2. They are 
basically catalysts for knowledge generation, as they act in the 
knowledge framework and appear in property restrictions or 
domain rules realized by a rule language like SWRL. Such rules 
are important, as they allow for the integration of external 
processing components, called built-ins, which are necessary to 
do conventional numerical processing. This happens when 
invoking algorithms or topological operations (Ben Hmida, 
Cruz, Nicolle, & Boochs, 2011). 
Figure 3 illustrates the taxonomical hierarchical structure of the 
class Algorithms. The algorithms are currently classified into 
three major sub-classes. This could be extended if there are 
requirements. Most of the algorithms used here belong to one of 
those three classes. 
  
Algorithms 
DataProcessing Geometric Topologic 
Segmentation Detection Representation Perpendicular Parallel Anglular 
Figure . Taxonomical hierarchical semantics of class 
Algorithms 
DomainConcept 
| 
Building£lement Furnitures 
i 
i Ï 
Signals Mast Schaltanlage.... Telefone Tische Schrenk........... Stuhl 
Figure . Taxonomical hierarchical semantics of class 
DomainConcept (scene Deutsche Bahn) 
Other classes in the top level ontology schema (figure 2) 
function as supporting classes. 
The objects from the scene are categorized to their respective 
classes under DomainConcept (figure 4). This structure is 
designed for the objects found in the Deutsche Bahn (‘DB’) 
scene. It can easily be replaced by other domains provided that 
the top level structure is respected. 
They represent knowledge domains defining the characteristics 
of the main classes. Two top level classes are currently defined, 
although this is extensible. Class Geometry (fig. 5a.) is defined 
to represent geometric characteristics. Class Data (5b) 
represents data characteristics. Both these classes are used 
within rules which map algorithms to scene knowledge. 
92 
Geometry Data 
_1D _2D _3D Data2D Data3D 
Point2D Line2D Area2D Point3D Line3D Area3D Image PointCloud StereoModel 
(a) (b) 
Figure . (a) Class hierarchy of geometry (b) data 
2.2.3. The Iteration 
ISCM detects and refines the detection process through new 
gained knowledge at every step of the iteration. In most cases 
the degree of knowledge is limited initially (barring case 2.1.1). 
The knowledge schema provides a basis for formulating the 
processing strategy, and provides a platform to define inference 
rules. These rules are based on the expert interpretation of the 
scene and the algorithmic behavior. The knowledge schema 
presents the prominent rule defining the scene, with which the 
ASM uses to begin the algorithmic processing. This prominent 
rule is inferred against the semantic rules of algorithms to 
evoke the most suitable algorithm or algorithmic sequence. 
Scene Algorithm Knowledge 2 us Q ; 
  
Processing Rules Strategy 
  
  
Fs 
Jj i í 
Doniain On tology 
Object Population i 
: BB detection } 
   
Objects’ 
Annotations 
Population into 
ontology 
Figure 6. Detailed Iterative Semantic Classification Method 
Again using the DB example, first the knowledge schema 
determines that the objects are vertical in nature, and algorithms 
suitable for vertical geometry detection are selected. After the 
algorithmic results are populated into the KB, the detected 
geometries are qualified to their respective objects. As such, the 
prominent rule evokes the first set of algorithms best suited to 
detect the simple and dominant objects, then the more complex 
objects are detected through their relationship to the simple 
ones through further iterations. 
Qualification follows the population. The detected geometries 
are inferred against the geometric semantics of the objects to 
initiate the first set of approximations. The current version of 
WiDOP provides this approximation through semantic 
annotation. Labeling mechanisms for 3D point clouds have 
been researched by Anguelov et al. (2005) and Munoz, 
Vandapel, & Hebert (2009), who use an associative Markov 
network to label. Shapovalov, Velizhev, & Barinova, (2010), 
Golovinskiy, Kim, & Funkhouser (2009), and Shapovalov, 
Veli 
capt 
feat 
Kop 
sem 
feat 
pres 
(Bei 
exe 
ann 
obje 
vert 
than 
3DP 
?ht) 
The 
the 
anni 
unai 
Un: 
mor 
Uni 
The 
amb 
secc 
KB 
AS? 
mec 
to e; 
to ir 
geoi 
pop! 
geo 
Basi 
swrl 
The 
obje 
amt 
proc 
2.1) 
iter: 
and 
iterz 
Fig 
and 
poir 
but 
pres