D Virtual World Retrieval Based on Ontology and Content

On the Web and on informatics systems, 3D virtual worlds used become big both in number and in size. Therefore, we propose in this paper a new method for retrieving 3D virtual worlds based on the semantics and the content. Firstly, we propose a new classified database of 3D virtual worlds given in VRML format. To achieve the semantic method, we construct an ontology that describes virtual worlds in various aspects including their contents (3D objects building a virtual world) and information about their contents (authors, file format, etc). This ontology is presented by OWL, the W3C recommended language. So as to extract desired 3D scenes from the proposed database we use the SPARQL query language. We propose finally a shape based method for searching desired virtual worlds by content in this database. This method is based on a new distance (metric) that define the similarity between virtual worlds. This method is evaluated using the recall vs. precision curves.


Introduction
With the development of software and 3D scanner that permit to build 3D objects, 3D virtual worlds including the 3D objects become more used around the web both in number and size.Those 3D virtual worlds are used in various domains like game, engineering, physical simulation etc. Considering this situation, finding desired 3D scenes from 3D virtual world database become difficult.The solution is to find new method for indexing and retrieving by semantic and by content those virtual worlds that contain sets of 3D objects.
Many researcher proposed yet search engines for 3D objects (Princeton search engine) (Informatics & Telematics Institute search engine) (ENIC search engine).Those search engines give user the possibility to search 3D objects by examples or by text key words from databases.The content based method are based in general on shape and use deformable (Niccolucci & D'Andrea, 2006) or non deformable 3D models (Chen et al., 2003;Ansary et al., 2007;Lmaati et al., 2009Lmaati et al., , 2010aLmaati et al., , 2010bLmaati et al., , 2010c)).The text based methods are in general based on key words describing 3D object given recently in XML format.
With the development of the semantic Web also called the intelligent web or next generation web or Web 3.0., RDF (2004) is used to represent information and to exchange knowledge on the Web.OWL, the recommended web ontology language is used for describing things of the web.SPARQL (2008) language an official W3C recommendation is an RDF (2004) query language that is used to retrieve and manipulate data stored by ontology.All these semantic web tools are important to improve the semantic 3D virtual world retrieval.
The semantic based methods are proposed in the literature (Horrocks, 2008;Gong et al., 2011;Kassimi et al., 2012;Niccolucci et al., 2006;Xin-ying et al., 2008).These methods use in general ontologies so as to describe 3D models.In our work we propose to describe and to retrieve using an ontology not just 3D models, but virtual worlds that contain sets of 3D models and other concepts building virtual world like graphical environments (illumination, fog, etc.) and animation of 3D models (translation, rotation, etc).
Practically, a set of 3D objects are used for building a 3D virtual world.Therefore, in order to find desired 3D virtual worlds from a database, we proceed to find 3D virtual worlds that contain the set of 3D objects constituting the virtual world query.Matching two 3D virtual worlds problematic is solved by matching two sets of 3D objects.For this, we define the distance (metric) between two sets of 3D objects using distances between 3D objects.
In this paper, we propose in the section 2 a database of virtual words.This database is a set of VRML (2001)

The class virtual wo
The class construct t

Ontology Structure
We are developing the 3D virtual world ontology so as to give description regarding the semantic (author of the virtual world, size and type of files used to build it) and the media content of the virtual world including objects given by meshes, objects given by basic forms (box), images used as texture or for the backgrounds, text, video and audio and the environment like (fog or illumination).This ontology describes as possible the used components to build the virtual worlds so as to retrieve them better.
The virtual world ontology that we propose is developed within the Protégé ( 2013) editor.This permit to generate the OWL ( 2004) file based on RDF/XML representation of the ontology.The graph given in Figure 4 shows the classes and subclasses used to describe a 3D virtual world.The classes: Meta-Info_semantic, Creation_Date, Author, Names_dif-Language and StorageProperty are subclasses of 'Thing'.Those classes represent the metadata about the representation of semantic information about the virtual world and its objects.
The class Animation_Object is subclass of 'Thing', it describes the object animations used to build the virtual world.
Since a variety of 3D-shape descriptors are proposed in the literature, we choose the Light Field (LF) (Chen et al., 2003), the 3D-DCT (Lmaati et al., 2010b) and the 3DCA (Lmaati et al., 2010c) as the 3D descriptors used by our ontology so as to store their components.The classes LFD, TD-DCT and ThreeDCA are subclasses of the class Global_features.
The image_Des_2D that permits to store the components of image descriptors has two subclasses: the Image_2D_Des_Color class and the Shape_Des_2D class.
In this section we give the different kinds of relationships between concepts for the proposed ontology as presented in Figure 6.A 3D virtual world presented by the class Virtual_world_3D can have the audio in the background and the text; those relations are given by the relationships has_Audio_BG and has_text respectively.The given the visual

Many wor
The

Figure
Figure Figure 8.

Fi
Figure 11.Two

s 9: this class 10: this class
Those classes are used to describe the virtual worlds which are subclass of the main class 'Thing'.The main class virtual_world_3D is a subclass of 'Thing', represent the virtual world description.The main class 'Object' present an abstraction of objects used to build the virtual world like 'table', 'tree', 'plane', etc.This class has subclasses: 1-'Complex_Obj-mesh' that presents the objects of the virtual world given by meshes.2-'Group_Basic_Forms_Obj'thatpresents the objects obtained by grouping a set of basic geometric forms like box and cylinder.3-Box,4-Cylinder, 5 -Cone, 6 -Indexed_Face_Set, 7 -Sphere, 8 -HalfCone, 9 -Torus, 10 -HalfCylinder and 11-HalfSphere.The class 'Image' presents the images used to build the virtual world background or other.The class texture presents the image used for texturing objects used in the virtual world.The class text presents the object text used in the virtual world as shown in the Figure2.
The classes: image, texture, text, and audio_BG are subclasses of the root class 'Thing'.They represent the abstraction of all media that can be part of a virtual world.The classes: Color, Illumination, Fog and Floor are subclasses of 'Thing'.They are used to describe the environment and the background of the virtual world.
The Virtual_world_3D has the Meta information presented by the class Meta-Info_semantic, it has also the storage property presented by the class StorageProperty; the graphics in the background (exp.image) description, environment (exp.fog) description and objects presented by the class Object.Those relations are presented by has_semanticDs, has_StorageProperty, has_GraphicsBG-env and has_Objects respectively.The Meta-Info_semantic is a class that has Author, Creation_Date and Names-dif_Languages; those relations are presented by has_Author, has_Creation_Date and has_Names respectively.
(Lmaati et al., 2010c)olor or texture, has storage property, has the Meta Information, can be animated and has a 3D shape description (example: ThreeDCA: 3DCA(Lmaati et al., 2010c)); those relationships are presented by has_Color, has_Texture, has_StorageProperty, has_semanticsDs, is_animatedBy, has_SD_3DCA respectively.The class Graphics_Env_BG_Des has illumination, can have Floor, can have Images or Textures and can have a Fog environment.Those relationships are presented by has_Illumination, has_Floor, uses_Image, has_Texture