towards interoperability in tracking systems: an ontology-based approach juan gómez romero miguel...

Post on 28-Dec-2015

221 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

TOWARDS INTEROPERABILITY IN TRACKING SYSTEMS:AN ONTOLOGY-BASED APPROACH

Juan Gómez RomeroMiguel A. PatricioJesús GarcíaJosé M. Molina

Applied A.I. Research Group (GIAA)University Carlos III of Madrid

the problem

To provide means to facilitate communication, interoperability,

scalability and extensibility of multi-camera tracking systems

CS-MAS: multi-camera agent-based tracking system

FusionAgents

Track.Agents

Track.Agents

CS-MAS

data communication

Tracking data: Track identification Physical properties (2D or 3D-space):

Size, Position, Color, Velocity, etc.

Estimated properties (Kalman, etc.) Size, Position, Velocity, etc.

State Active, Occluded, Grouped, etc.

multi-camera tracking systems: CS-MAS

FusionAgents

Track.Agents

Track.Agents

CS-MAS

Variable Value

Track_ID 1

In_Frame 2

From_Camera ABX56

Width 50

Height 70

Pos_X 324

… …

example

The pizza delivery example

communication problems

Problem: Misunderstandings!

Different individuals involved: Different vocabulary Different assumptions Different background knowledge

Solution: Use a formal language to describe pizzas

Knowledge representation: Ontologies

ontologies

“Formal, explicit specifications of a shared conceptualization” [1] An ontology is a knowledge model which describes from a common

perspective the objects in a common domain using a language that can be processed automatically

Based on Description Logics (DLs) DLs are a family of logics to represent structured knowledge

Basic constructs: Concepts, Relations, Individuals, Axioms

Standard The Web Ontology Language (OWL)

[1] R. Studer, V. R. Benjamins, & D. Fensel. “Knowledge engineering: principles and methods”. In: Data Knowledge Engineering 25.1-2 (1998). Pp. 161–197.

example: the pizza ontology

Manchester Pizza Ontology: http://www.co-ode.org/ontologies/pizza/pizza.owl

American Pizza Class: http://www.co-ode.org/ontologies/pizza/pizza.owl#American Is a:

NamedPizza hasCountryOfOrigin value America hasTopping only

(MozzarellaTopping or PeperoniSausageTopping or TomatoTopping)

hasTopping some MozzarellaTopping hasTopping some PeperoniSausageTopping hasTopping some TomatoTopping

advantages of the use of ontologies

Understanding among agents: Different interpretations are not possible

Decoupling of internal and external representations A pizza image can have associated a formal description

Extensibility of the architecture Different pizza companies can communicate; delivery

could be extended between districts

advantages of the use of ontologies

Obtaining implicit knowledge by reasoning Pepper is a Spicy ingredient; pizzas with pepper are Spicy

pizzas

Support for high-level information interpretation It can be deduced, using DL inference, that a client likes spicy

pizzas and special offers can be sent

Improved data manipulation and querying Ontologies have associated query languages (e.g. SPARQL)

Implementation of mash-up applications A web page with suggestions to the clients based on their

preferences

proposal

Use of ontologies to describe the tracking information exchanged

between the agents of CS-MAS

Tasks: Development of the TREND (Tracking Entities

Description) ontology Use of the TREND ontology as the communication

language of the agents

TREND ontology: basic classes

TREND ontology: track states

TREND ontology: properties representation

example: Contents of CS-MAS messages

<Track rdf:about="#track12"> <hasSnapshot rdf:resource="#snapshot_A"/></Track>

<TrackSnapshot rdf:about="#snapshot_12_A "> <rdf:type rdf:resource="#ActiveTrackSnapshot"/> <isValidIn rdf:resource="#frame1"/> <isValidIn rdf:resource="#frame2"/> <hasActualProperties rdf:resource="#actual_properties_12_A"/> <hasPredictedProperties rdf:resource="#predicted_properties_12_A "/></TrackSnapshot>

… (continues)

example: Contents of CS-MAS messages

<Track rdf:about="#track12"> <hasSnapshot rdf:resource="#snapshot_B"/></Track>

<TrackSnapshot rdf:about="#snapshot_12_B "> <rdf:type rdf:resource="#ActiveTrackSnapshot"/> <isValidIn rdf:resource="#frame3"/> <hasActualProperties rdf:resource="#actual_properties_12_B"/> <hasPredictedProperties rdf:resource="#predicted_properties_12_B "/></TrackSnapshot>

summary & future work

Ontology for describing the tracking data interchanged by the agents of CS-MAS (a multi-camera tracking system)

Common vocabulary advantages: understandability, extensibility, interoperability

Research directions: Fully integration of TREND in CS-MAS Implementation of software tools exploiting TREND, e.g. a visualization

tool to present the temporal evolution of tracks of the image High-level interpretation of data

Interpretation of the scene in terms of objects, events, etc.

Define, based on TREND, more abstract descriptive ontologies

end

Thank you!

jgomez@inf.uc3m.es

top related