Research and Future Perspectives on Intelligent Video Surveillance Systems Monique THONNAT Senior Scientist Head of Orion research team INRIA Sophia Antipolis FRANCE
Jan 18, 2016
Research and Future Perspectives on Intelligent Video Surveillance Systems
Monique THONNAT
Senior Scientist Head of Orion research team
INRIA Sophia AntipolisFRANCE
2
Introduction 1/4
Which Security Problems? Safety and security of goods and human beings
Safety: protection in case of accident or incident (e.g.: fall)
Security: protection against a malicious act (ex.: bomb) Evolution
1980: guarding, human surveillance 1990: start of video cameras setting up, remote
surveillance Highways, parking lots, subways, malls …
2000: explosion of video-surveillance Set of new laws Better understanding of the general public Threats of tragic events, terrorism acts
3
Introduction: control room 2/4
4
Introduction: issue 3/4
Definition: intelligent video surveillance automatic analysis of the video streams
Why ? Cost: 1 human operator /6-10 video streams Current paradox : the more there are video
cameras, the less these video camera are observed E.g. 600 000 video cameras in London need for the policemen to look at stored videos tapes
Effectiveness: duration of vigilance for an operator: > 1 or 2hours: % of the attention is lost
5
Introduction 4/4
Huge information flow Few pertinent information
Video streams
Video cameras Surveillance control rooms
Selection of the information Increase of the detection rate
Intelligent video
surveillance software
Intelligent video
surveillance software
Alarms
Video cameras
Video streams
Surveillance control rooms
6
Intelligent Video Surveillance 1/3demonstration
From object detection to complex event recognition(e.g.: violence)
7
Intelligent Video Surveillance Definition: Data captured by video surveillance cameras Real time and automated analysis of video sequences Video understanding= from people detection and
tracking to behavior recognition
Recognition of complex behaviors: of individuals (e.g. fraud, graffiti, vandalism, bank
attack) of small groups (e.g. fighting) of crowds (e.g. overcrowding) interactions of people and vehicles (e.g. aircraft
refueling)
Intelligent Video Surveillance 2/3
8
Intelligent Video Surveillance 3/3Typical problems
Metro station surveillance Surveillance inside trains
Building access control Airport monitoring
9
Video Understanding for Intelligent Video Surveillance 1/13
Definition:Cognitive Vision is a new research field mixing: Computer vision techniquesfor object detection, description, categorization and tracking Artificial intelligence techniquesfor knowledge acquisition, reasoning (e.g. spatial and temporal
reasoning,…), learning (e.g. categories, structures, parameters…)
Software engineering techniquesfor vision software design, integration, reusability, evaluation
Reference: http:www.eucognition.org/ecvision site and roadmap
10
Intelligent Videosurveillance:
How? A Cognitive vision approach for video understanding mixing:
computer vision: 4D analysis (3D + temporal analysis)
artificial intelligence: a priori knowledge (scenario,
environment)
software engineering: reusable software platform (VSIP)
Video Understanding for Intelligent Video Surveillance
2/13
11
Video Understanding for Intelligent Video Surveillance 3/13
4 D analysis:multi-cameras
tracking
Video understanding
People detection
and tracking
Scenario recognition
A PRIORI KNOWLEDGE:• 3d models of the environment • Camera calibration• Scenario Models
Alarms
People detection
and tracking
Interpretation of the videos from pixels to alarms
12
Video Understanding for Intelligent Video Surveillance
4/13
SegmentationSegmentation ClassificationClassification TrackingTracking Scenario RecognitionScenario Recognition
Alarms
access to forbidden
area
3D scene modelScenario models A priori Knowledge
Objective: Interpretation of videos from pixels to alarms
13
Behavior recognition: approach based on a priori knowledge
model of the empty scene (3D geometry and semantics)
models of predefined scenarios
a language for representing scenarios based on
combination of states and events more than 20 states and 20 events can be used
a reasoning mechanism for real time detection of states,
events and scenarios (e.g. temporal reasoning,
constraints solving techniques)
Video Understanding for Intelligent Video Surveillance
5/13
14
3D Scene Model: Barcelona Metro Station Sagrada Famiglia mezzanine (European project ADVISOR)
Video Understanding for Intelligent Video Surveillance
6/13
15
States, Events and Scenarios : State: a spatio-temporal property involving one or several actors on a time interval
Ex : « close», « walking», « seated»
Event: a significant change of states
Ex : « enters», « stands up», « leaves »
Scenario: a long term symbolic application dependent activity
Ex : « fighting», « vandalism»
Video Understanding for Intelligent Video Surveillance
7/13
16
Vandalism scenario description :
Scenario(vandalism_against_ticket_machine, Physical_objects((p : Person), (eq : Equipment, Name = “Ticket_Machine”) )
Components ((event s1: p moves_close_to eq) (state s2: p stays_at eq)
(event s3: p moves_away_from eq) (event s4: p moves_close_to eq)
(state s5: p stays_at eq) ) Constraints ((s1 != s4) (s2 != s5)
(s1 before s2) (s2 before s3) (s3 before s4) (s4 before s5) ) ) )
Video Understanding for Intelligent Video Surveillance
8/13
Scenario Recognition : Temporal constraints
17
Video Understanding for Intelligent Video Surveillance 9/13Vandalism in metro (Nuremberg, Germany)
18
Video Understanding for Intelligent Video Surveillance 10/13
3d Scene Model of 2 bank agencies
objet du contexte
mur et portezone d’accès
salle du coffrerue
rue
salle automates
zone d’entrée de l’agence
zone des distributeurs
zone de jour/nuit
zone devant le guichet
zone derrière le guichet
zone d’accès au bureau du
directeur
zone de jour
ported’entrée
porte salleautomates
armoire
guichet
commode
19
Video Understanding for Intelligent Video Surveillance
11/13Bank Monitoring: Bank attack scenario description :
scenario Bank_attack_one_robber_one_employeephysical_objects: ((employee : Person), (robber : Person), z1: Back_Counter, z2: Entrance_Zone, z3: Front_Counter, z4: Safe, d: Safe_door)
components: (State c1 : Inside_zone(employee, z1)) (Event c2 : Changes_zone(robber, z2,z3))
(State c3 : Inside_zone(employee, z4)) (State c4 : Inside_zone(robber, z4))) constraints : ((c2 during c1) (c2 before c3) (c1 before c3) (c2 before c4) (c4 during c3) (d is open))
20
Video Understanding for Intelligent Video Surveillance
12/13
bank monitoring Recognition of a bank attack scenario: an employee is e behind a counter, an aggressor enters, goes behind the counter, then he goes with the emplyee towards theSTR (secured technical room), they enter in the STR then they leave ethe STR and they go toward the exit of the agency.
Examples : Brussels and Barcelona Metro Surveillance
Exit zone
Jumping over barrier
Blocking
Overcrowding
Fighting
Group
behavior
Crowd
behavior
Individual
behavior
Groupbehavior
Video Understanding for Intelligent Video Surveillance
13/13
21
22
Impact: Visual surveillance of metro stations, bank agencies, trains,
buildings and airports
5 European projects (PASSWORDS, AVS-PV, AVS-RTPW,
ADVISOR, AVITRACK)
4 contracts with End-users companies (metro, bank, trains)
2 transfer activities with Bull (Paris) and Vigitec (Brussels)
Cooperation over more than 11 years with partners
Creation in 2005 of a start-up Keeneo www.keeneo.com
Conclusion 1/5
23
Conclusion 2/5
Hypotheses: fixed cameras 3D model of the empty scene predefined behavior models
Results:
+ Behavior understanding for Individuals, Groups of people, Crowd or Vehicles
+ an operational language for video understanding (more than 20 states and events)
+ a real-time platform (10 to 25 frames/s)
24
Conclusion 3/5 Current issues
Systems have poor performances over time, can be hardly modified and do not use enough a priori knowledge
shadowsstrong perspectivetiny objects
close view
clutterlightingconditions
25
Knowledge Acquisition Design of learning techniques to complement a
priori knowledge: Frequent events, scenario model learning European project CARETAKER
Object description Fine human shape description:3D posture models Crowd description: European project SERKET
Reusability is still an issue for vision programs Video analysis robustness Dynamic configuration of programs and parameters
Conclusion: Where we go 4/5
26
Conclusion 5/5 Posture Recognition
Current image and binary image
Instantaneous posture
Postures recognised along the time
27
Conclusion 5/5 Crowd Behavior
Motion direction detection
Abnormal direction of people in a crowd
28
Video Understanding demo?
Airport Apron Monitoring “Unloading Operation” European AVITRACK project