A semi-supervised learning framework based on spatio-temporal semantic events for maritime anomaly detection and behaviour analysis
1. 1/24
A semi-supervised learning framework basedA semi-supervised learning framework based
on spatio-temporal semantic eventson spatio-temporal semantic events
for maritime anomaly detection and behaviour analysisfor maritime anomaly detection and behaviour analysis
Arnaud VandecasteeleArnaud Vandecasteele
Rodolphe DevillersRodolphe Devillers
Aldo NapoliAldo Napoli
CoastGIS - GIS and New Technologies - June 20
2. 2/24
Background & Research problems
Maritime domain
Problem
Semantic Event Modelling
What is an ontology ?
Simple Event Model
Vessels behaviours analysis
Prototype & examples
Prototype architecture
Components of the architecture
Examples
3. 3/24
Context
Economic
90% of world trade is transported by sea
In Europe 90% of oil and gas are transported by sea
Illegal Fishing
Only 6% of illegal fishing frauds are detected
88% of fishing stocks in the EU are overexploited
Illegal immigration
55% of illegal border crossing immigration is done by
sea (EU)
3000 illegal ''known'' immigrants lost their life at sea
every year
Source : ICC International Maritime Bureau
Maritime domain
Background & Research problems >Background & Research problems > Semantic Event Modelling > Prototype & examples
4. 4/24
Poor interface
Data Overflow
Few information
Large surveillance area
High maritime traffic density
Cognitive Overflow
No tools for automatic detection
Maritime information system
Background & Research problems >Background & Research problems > Semantic Event Modelling > Prototype & examples
5. 5/24
?
ImproveImprove Detection & Analysis
Better understandingunderstanding
for maritime surveillance
High volume of data
Heterogeneous data
and knowledge
Distributed
data and knowledge
Analysis of
complex information
Research problem
Background & Research problems >Background & Research problems > Semantic Event Modelling > Prototype & examples
6. 6/24
Improve Understanding
An enriched formalization with spatial capabilities offers a better
way to describe and analyze the behaviour of the vessels1
Formalize
expert knowledge
Automated
spatial reasoning
Spatial Ontologies
2
Integrate the spatial
dimension into ontologies
Automatic detection
of suspicious events
Automatic identification
of abnormal behaviours
Research problem
Background & Research problems >Background & Research problems > Semantic Event Modelling > Prototype & examples
12. 12/24
Actor TypeRole Type Event Type Object Type Place Type
ROLE
ACTOR
PLACE
EVENT
Time-Stamped Entity
subClassOfsubClassOf
subClassOf
subClassOfsubClassOf
Takes place in
Time
Stamp
Participates in
as role
(begins in place - ends in place)
hasRole
Takes place in
Takes place in
Involves inParticipates in
Has role
type
Has Actor
type
Has Event
type
Has object
type
Has place
type
Has Actr
type
Linked together with types
Background & Research problems > Semantic Event Modelling >Semantic Event Modelling > Prototype & examples
16. 16/24
Dataset :
More than 5 millions of AIS positions
Between February and December 2009
Information
Position, timestamp, heading, speed...
http://www.chorochronos.org/?q=node/9
Data from the French Naval Academy Resarch Lab
Background & Research problems > Semantic Event Modelling > Prototype & examplesPrototype & examples
21. 21/24
Timeline to navigate
through time
Time widget
to animate the data
3D Web Mapping
interface
Visualization of the results
Background & Research problems > Semantic Event Modelling > Prototype & examplesPrototype & examples
22. 22/24
2D View
3D View
Example of acceleration events
Background & Research problems > Semantic Event Modelling > Prototype & examplesPrototype & examples
23. 23/24
Conclusion
Ontologies provide a richer way to describe events
A richer description can provide a better understanding of a situation
A semantic model linked to a webmapping interface has been created
This prototype offers an interface to explore semantic events
More events type must be added
Vessels must be linked to the timeline