SlideShare une entreprise Scribd logo
1  sur  31
Télécharger pour lire hors ligne
Rule-Based High-Level Situation Recognition
           from Incomplete Tracking Data
                    David Münch1, Joris IJsselmuiden1, Ann-Kristin Grosselfinger1,
                            Michael Arens1, and Rainer Stiefelhagen1,2
                    1   Fraunhofer IOSB, Germany, david.muench@iosb.fraunhofer.de
                               2 Karlsruhe Institute of Technology, Germany.




                        The 6th International Symposium on Rules - Rule ML2012,
                                 Montpellier, France, August 27-29, 2012


                                                   1
© Fraunhofer IOSB
Motivation
Motivation



•     Security systems: Persons
      and their behavior are
      the focus of attention:
      Person centric analysis
•     Threat detection
•     Visual surveillance
•     Activity logging
•     Video search
•     Driver Assistance Systems


        Input data is incomplete
        and noisy.


                                   2
© Fraunhofer IOSB
Overview




• Cognitive Vision System as a whole.

• High-level knowledge and situation recognition.

• Handling Incomplete Data.

• Experiments.



                              3
© Fraunhofer IOSB
Cognitive Vision System




                          4
© Fraunhofer IOSB
Cognitive Vision System




                          5
© Fraunhofer IOSB
Cognitive Vision System




                          6
© Fraunhofer IOSB
Cognitive Vision System




                          7
© Fraunhofer IOSB
Cognitive Vision System




                          8
© Fraunhofer IOSB
Cognitive Vision System




                          9
© Fraunhofer IOSB
Conceptual Layer – Conceptual Primitives Level




•     Quantitative information (from Quantitative Layer) is transformed into primitive
      conceptual knowledge (Logic predicates).
•     Dictionary of basic rules.
•     Mainly domain independent.
•     Support of uncertainty and vagueness.
•     The rules in the CPL are mostly concerned with spatial relations and temporal
      relations on short time intervals.




                                           10
© Fraunhofer IOSB
Dictionary of basic rules




   Dictionary of basic rules for every
   domain.



   Fuzzy Metric Temporal Logic
   (FMTL):

   Extension of first order logic by
   notions of fuzziness, time, and
   metrics on time.

   Inference engine: F-LIMETTE



                                         11
© Fraunhofer IOSB
“Numbers” mapped to Concepts




  Fuzzy membership functions 𝜇 𝑠𝑝𝑒𝑒𝑑 𝑣𝑎𝑙𝑢𝑒 for the subset {zero, small, normal,
  high, very_high} of discrete conceptual speed values.


                                  Figure from: H.-H. Nagel, Steps toward a Cognitive Vision System, 2004, AI Mag.


                                             12
© Fraunhofer IOSB
Conceptual Layer – Behavior Representation Level




How to represent the expected Situations?
•     Knowledge represented in Situation Graph Trees.
     • Graphically editable
     • Easy to extend and edit
     • Interpretable (vs. black box)




•     Exhaustive situation recognition.




                                          13
© Fraunhofer IOSB
Behavior Representation Level

Basic logic predicates from Conceptual Primitives Level are aggregated and
structured in Situation Graph Trees (SGT)
       high-level conceptual situations.

An SGT consists of situation schemes:
•     Can be start and/or end nodes.
•     Unique name.
•     State scheme (Precondition).
•     Action scheme (Postcondition).

Specialize a situation scheme:
• Prediction edges link to a possible subsequent situation scheme.
• Specialization edges link to more specific situation graphs in a hierarchical
 structure.

                                            14
© Fraunhofer IOSB
Behavior Representation Level




                                15
© Fraunhofer IOSB
Handling Incomplete Data – Low Level in Scene Domain Level




           Perfect input data.


                                                Truth value




                                                  Time

                                 16
© Fraunhofer IOSB
Handling Incomplete Data – Low Level in SDL




           Incomplete input data.


                                              Truth value




                                               Time

                                    17
© Fraunhofer IOSB
Handling Incomplete Data – Low Level in SDL



Interpolation of input data:




                                              Truth value




                                               Time

                               18
© Fraunhofer IOSB
Handling Incomplete Data – Low Level in SDL



Interpolation of input data:




                                              Time

                               19
© Fraunhofer IOSB
Handling Incomplete Data – High Level in BRL




                               20
© Fraunhofer IOSB
Handling Incomplete Data – High Level in BRL




                    What if is_open(trunk, Car) fails?




                                          21
© Fraunhofer IOSB
Handling Incomplete Data – High Level in BRL




Hallucination (abduction) of missing evidence.




                    What if is_open(trunk, Car) fails?

                     Hallucinate is_open(trunk, Car) and continue!


                                          22
© Fraunhofer IOSB
VIRAT Video Dataset


  VIRAT Video Dataset Release 1.0



  Input data: annotated ground truth: persons and their situations.




  Place: 0000:                                Place: 0002:
  Videos: 02 03 04 06                         Videos: 00 06


                                         23
© Fraunhofer IOSB
Situations

1. Person loads object into car.
2. Person unloads object from car.
3. Person gets into car.
4. Person gets out of car.




                                          VIRAT Video Dataset Release 1.0




                                     24
© Fraunhofer IOSB
Evaluation




•     The annotated ground truth is regarded as complete information.
•     Randomly make gaps of a distinct length into the data.
•     Increase the amount of missing data in steps of 5%.
•     Repeat each experiment several times.




                                          25
© Fraunhofer IOSB
Evaluation


  Original,              F-Score
  unmodified


                         Gap size
                         5 seconds.




 Precision                Recall



Gap size                 Gap size
5 seconds.               5 seconds.



                    26
© Fraunhofer IOSB
Evaluation


  Original,              F-Score
  unmodified


                         Gap size
                         5 seconds.




 Precision                Recall



Gap size                 Gap size
5 seconds.               5 seconds.



                    27
© Fraunhofer IOSB
Evaluation


      ROC-curves for video (d) with gap size 5 seconds.


   TPR                                       TPR




                          FPR                                       FPR
           With interpolation and hallucination. Without interpolation and hallucination.


                                             28
© Fraunhofer IOSB
Evaluation




  Video (d),
  F-Score                Gap size
                         5 seconds.




Gap size                 Gap size
5 seconds.               5 seconds.



                    29
© Fraunhofer IOSB
Conclusion




•     low level: interpolation of data and its uncertainty.
      ordinary incomplete data.
•     high level: extension of the situation recognition inference algorithm
      (hallucination, abduction).
      high-level incomplete data such as occlusions.
•     Knowledge base for vehicle-centered situations.
•     Runs in real-time on off-the-shelf hardware.




                                             30
© Fraunhofer IOSB
Finis.




                      31
© Fraunhofer IOSB

Contenu connexe

En vedette

HARM: A Hybrid Rule-based Agent Reputation Model based on Temporal Defeasibl...
HARM: A Hybrid Rule-based Agent Reputation Model  based on Temporal Defeasibl...HARM: A Hybrid Rule-based Agent Reputation Model  based on Temporal Defeasibl...
HARM: A Hybrid Rule-based Agent Reputation Model based on Temporal Defeasibl...RuleML
 
KELPS LPS - A Logic-Based Framework for Reactive System30 aug 2012
KELPS LPS - A Logic-Based Framework for Reactive System30 aug 2012KELPS LPS - A Logic-Based Framework for Reactive System30 aug 2012
KELPS LPS - A Logic-Based Framework for Reactive System30 aug 2012RuleML
 
RuleML2015: Using Substitutive Itemset Mining Framework for Finding Synonymou...
RuleML2015: Using Substitutive Itemset Mining Framework for Finding Synonymou...RuleML2015: Using Substitutive Itemset Mining Framework for Finding Synonymou...
RuleML2015: Using Substitutive Itemset Mining Framework for Finding Synonymou...RuleML
 
RuleML2015: Rule-Based Exploration of Structured Data in the Browser
RuleML2015: Rule-Based Exploration of Structured Data in the BrowserRuleML2015: Rule-Based Exploration of Structured Data in the Browser
RuleML2015: Rule-Based Exploration of Structured Data in the BrowserRuleML
 
A Model Driven Reverse Engineering framework for extracting business rules ou...
A Model Driven Reverse Engineering framework for extracting business rules ou...A Model Driven Reverse Engineering framework for extracting business rules ou...
A Model Driven Reverse Engineering framework for extracting business rules ou...RuleML
 
RuleML2015: GRAAL - a toolkit for query answering with existential rules
RuleML2015:  GRAAL - a toolkit for query answering with existential rulesRuleML2015:  GRAAL - a toolkit for query answering with existential rules
RuleML2015: GRAAL - a toolkit for query answering with existential rulesRuleML
 
RuleML 2015: Ontology Reasoning using Rules in an eHealth Context
RuleML 2015: Ontology Reasoning using Rules in an eHealth ContextRuleML 2015: Ontology Reasoning using Rules in an eHealth Context
RuleML 2015: Ontology Reasoning using Rules in an eHealth ContextRuleML
 
RuleML2015: Using PSL to Extend and Evaluate Event Ontologies
RuleML2015: Using PSL to Extend and Evaluate Event OntologiesRuleML2015: Using PSL to Extend and Evaluate Event Ontologies
RuleML2015: Using PSL to Extend and Evaluate Event OntologiesRuleML
 
RuleML2015: Semantics of Notation3 Logic: A Solution for Implicit Quantifica...
RuleML2015:  Semantics of Notation3 Logic: A Solution for Implicit Quantifica...RuleML2015:  Semantics of Notation3 Logic: A Solution for Implicit Quantifica...
RuleML2015: Semantics of Notation3 Logic: A Solution for Implicit Quantifica...RuleML
 
RuleML2015 PSOA RuleML: Integrated Object-Relational Data and Rules
RuleML2015 PSOA RuleML: Integrated Object-Relational Data and RulesRuleML2015 PSOA RuleML: Integrated Object-Relational Data and Rules
RuleML2015 PSOA RuleML: Integrated Object-Relational Data and RulesRuleML
 
Ruleml2012 - A production rule-based framework for causal and epistemic reaso...
Ruleml2012 - A production rule-based framework for causal and epistemic reaso...Ruleml2012 - A production rule-based framework for causal and epistemic reaso...
Ruleml2012 - A production rule-based framework for causal and epistemic reaso...RuleML
 

En vedette (13)

HARM: A Hybrid Rule-based Agent Reputation Model based on Temporal Defeasibl...
HARM: A Hybrid Rule-based Agent Reputation Model  based on Temporal Defeasibl...HARM: A Hybrid Rule-based Agent Reputation Model  based on Temporal Defeasibl...
HARM: A Hybrid Rule-based Agent Reputation Model based on Temporal Defeasibl...
 
KELPS LPS - A Logic-Based Framework for Reactive System30 aug 2012
KELPS LPS - A Logic-Based Framework for Reactive System30 aug 2012KELPS LPS - A Logic-Based Framework for Reactive System30 aug 2012
KELPS LPS - A Logic-Based Framework for Reactive System30 aug 2012
 
RuleML2015: Using Substitutive Itemset Mining Framework for Finding Synonymou...
RuleML2015: Using Substitutive Itemset Mining Framework for Finding Synonymou...RuleML2015: Using Substitutive Itemset Mining Framework for Finding Synonymou...
RuleML2015: Using Substitutive Itemset Mining Framework for Finding Synonymou...
 
RuleML2015: Rule-Based Exploration of Structured Data in the Browser
RuleML2015: Rule-Based Exploration of Structured Data in the BrowserRuleML2015: Rule-Based Exploration of Structured Data in the Browser
RuleML2015: Rule-Based Exploration of Structured Data in the Browser
 
A Model Driven Reverse Engineering framework for extracting business rules ou...
A Model Driven Reverse Engineering framework for extracting business rules ou...A Model Driven Reverse Engineering framework for extracting business rules ou...
A Model Driven Reverse Engineering framework for extracting business rules ou...
 
RuleML2015: GRAAL - a toolkit for query answering with existential rules
RuleML2015:  GRAAL - a toolkit for query answering with existential rulesRuleML2015:  GRAAL - a toolkit for query answering with existential rules
RuleML2015: GRAAL - a toolkit for query answering with existential rules
 
RuleML 2015: Ontology Reasoning using Rules in an eHealth Context
RuleML 2015: Ontology Reasoning using Rules in an eHealth ContextRuleML 2015: Ontology Reasoning using Rules in an eHealth Context
RuleML 2015: Ontology Reasoning using Rules in an eHealth Context
 
RuleML2015: Using PSL to Extend and Evaluate Event Ontologies
RuleML2015: Using PSL to Extend and Evaluate Event OntologiesRuleML2015: Using PSL to Extend and Evaluate Event Ontologies
RuleML2015: Using PSL to Extend and Evaluate Event Ontologies
 
RuleML2015: Semantics of Notation3 Logic: A Solution for Implicit Quantifica...
RuleML2015:  Semantics of Notation3 Logic: A Solution for Implicit Quantifica...RuleML2015:  Semantics of Notation3 Logic: A Solution for Implicit Quantifica...
RuleML2015: Semantics of Notation3 Logic: A Solution for Implicit Quantifica...
 
RuleML2015 PSOA RuleML: Integrated Object-Relational Data and Rules
RuleML2015 PSOA RuleML: Integrated Object-Relational Data and RulesRuleML2015 PSOA RuleML: Integrated Object-Relational Data and Rules
RuleML2015 PSOA RuleML: Integrated Object-Relational Data and Rules
 
Ruleml2012 - A production rule-based framework for causal and epistemic reaso...
Ruleml2012 - A production rule-based framework for causal and epistemic reaso...Ruleml2012 - A production rule-based framework for causal and epistemic reaso...
Ruleml2012 - A production rule-based framework for causal and epistemic reaso...
 
Voz pasiva recetas
Voz pasiva recetasVoz pasiva recetas
Voz pasiva recetas
 
Curso de francés Basico
Curso de francés BasicoCurso de francés Basico
Curso de francés Basico
 

Plus de RuleML

Aggregates in Recursion: Issues and Solutions
Aggregates in Recursion: Issues and SolutionsAggregates in Recursion: Issues and Solutions
Aggregates in Recursion: Issues and SolutionsRuleML
 
A software agent controlling 2 robot arms in co-operating concurrent tasks
A software agent controlling 2 robot arms in co-operating concurrent tasksA software agent controlling 2 robot arms in co-operating concurrent tasks
A software agent controlling 2 robot arms in co-operating concurrent tasksRuleML
 
Port Clearance Rules in PSOA RuleML: From Controlled-English Regulation to Ob...
Port Clearance Rules in PSOA RuleML: From Controlled-English Regulation to Ob...Port Clearance Rules in PSOA RuleML: From Controlled-English Regulation to Ob...
Port Clearance Rules in PSOA RuleML: From Controlled-English Regulation to Ob...RuleML
 
RuleML 2015: When Processes Rule Events
RuleML 2015: When Processes Rule EventsRuleML 2015: When Processes Rule Events
RuleML 2015: When Processes Rule EventsRuleML
 
Challenge@RuleML2015 Developing Situation-Aware Applications for Disaster Man...
Challenge@RuleML2015 Developing Situation-Aware Applications for Disaster Man...Challenge@RuleML2015 Developing Situation-Aware Applications for Disaster Man...
Challenge@RuleML2015 Developing Situation-Aware Applications for Disaster Man...RuleML
 
Rule Generalization Strategies in Incremental Learning of Disjunctive Concepts
Rule Generalization Strategies in Incremental Learning of Disjunctive ConceptsRule Generalization Strategies in Incremental Learning of Disjunctive Concepts
Rule Generalization Strategies in Incremental Learning of Disjunctive ConceptsRuleML
 
RuleML 2015 Constraint Handling Rules - What Else?
RuleML 2015 Constraint Handling Rules - What Else?RuleML 2015 Constraint Handling Rules - What Else?
RuleML 2015 Constraint Handling Rules - What Else?RuleML
 
RuleML2015 The Herbrand Manifesto - Thinking Inside the Box
RuleML2015 The Herbrand Manifesto - Thinking Inside the Box RuleML2015 The Herbrand Manifesto - Thinking Inside the Box
RuleML2015 The Herbrand Manifesto - Thinking Inside the Box RuleML
 
Industry@RuleML2015: Norwegian State of Estate A Reporting Service for the St...
Industry@RuleML2015: Norwegian State of Estate A Reporting Service for the St...Industry@RuleML2015: Norwegian State of Estate A Reporting Service for the St...
Industry@RuleML2015: Norwegian State of Estate A Reporting Service for the St...RuleML
 
A Service for Improving the Assignments of Common Agriculture Policy Funds to...
A Service for Improving the Assignments of Common Agriculture Policy Funds to...A Service for Improving the Assignments of Common Agriculture Policy Funds to...
A Service for Improving the Assignments of Common Agriculture Policy Funds to...RuleML
 
Datalog+-Track Introduction & Reasoning on UML Class Diagrams via Datalog+-
Datalog+-Track Introduction & Reasoning on UML Class Diagrams via Datalog+-Datalog+-Track Introduction & Reasoning on UML Class Diagrams via Datalog+-
Datalog+-Track Introduction & Reasoning on UML Class Diagrams via Datalog+-RuleML
 
RuleML2015: Binary Frontier-guarded ASP with Function Symbols
RuleML2015: Binary Frontier-guarded ASP with Function SymbolsRuleML2015: Binary Frontier-guarded ASP with Function Symbols
RuleML2015: Binary Frontier-guarded ASP with Function SymbolsRuleML
 
RuleML2015: API4KP Metamodel: A Meta-API for Heterogeneous Knowledge Platforms
RuleML2015: API4KP Metamodel: A Meta-API for Heterogeneous Knowledge PlatformsRuleML2015: API4KP Metamodel: A Meta-API for Heterogeneous Knowledge Platforms
RuleML2015: API4KP Metamodel: A Meta-API for Heterogeneous Knowledge PlatformsRuleML
 
RuleML2015: Ontology-Based Multidimensional Contexts with Applications to Qua...
RuleML2015: Ontology-Based Multidimensional Contexts with Applications to Qua...RuleML2015: Ontology-Based Multidimensional Contexts with Applications to Qua...
RuleML2015: Ontology-Based Multidimensional Contexts with Applications to Qua...RuleML
 
RuleML2015: Compact representation of conditional probability for rule-based...
RuleML2015:  Compact representation of conditional probability for rule-based...RuleML2015:  Compact representation of conditional probability for rule-based...
RuleML2015: Compact representation of conditional probability for rule-based...RuleML
 
RuleML2015: Learning Characteristic Rules in Geographic Information Systems
RuleML2015: Learning Characteristic Rules in Geographic Information SystemsRuleML2015: Learning Characteristic Rules in Geographic Information Systems
RuleML2015: Learning Characteristic Rules in Geographic Information SystemsRuleML
 
RuleML2015: User Extensible System to Identify Problems in OWL Ontologies and...
RuleML2015: User Extensible System to Identify Problems in OWL Ontologies and...RuleML2015: User Extensible System to Identify Problems in OWL Ontologies and...
RuleML2015: User Extensible System to Identify Problems in OWL Ontologies and...RuleML
 
RuleML2015: Representing Flexible Role-Based Access Control Policies Using Ob...
RuleML2015: Representing Flexible Role-Based Access Control Policies Using Ob...RuleML2015: Representing Flexible Role-Based Access Control Policies Using Ob...
RuleML2015: Representing Flexible Role-Based Access Control Policies Using Ob...RuleML
 
RuleML2015: Rule Generalization Strategies in Incremental Learning of Disjunc...
RuleML2015: Rule Generalization Strategies in Incremental Learning of Disjunc...RuleML2015: Rule Generalization Strategies in Incremental Learning of Disjunc...
RuleML2015: Rule Generalization Strategies in Incremental Learning of Disjunc...RuleML
 
Industry@RuleML2015 DataGraft
Industry@RuleML2015 DataGraftIndustry@RuleML2015 DataGraft
Industry@RuleML2015 DataGraftRuleML
 

Plus de RuleML (20)

Aggregates in Recursion: Issues and Solutions
Aggregates in Recursion: Issues and SolutionsAggregates in Recursion: Issues and Solutions
Aggregates in Recursion: Issues and Solutions
 
A software agent controlling 2 robot arms in co-operating concurrent tasks
A software agent controlling 2 robot arms in co-operating concurrent tasksA software agent controlling 2 robot arms in co-operating concurrent tasks
A software agent controlling 2 robot arms in co-operating concurrent tasks
 
Port Clearance Rules in PSOA RuleML: From Controlled-English Regulation to Ob...
Port Clearance Rules in PSOA RuleML: From Controlled-English Regulation to Ob...Port Clearance Rules in PSOA RuleML: From Controlled-English Regulation to Ob...
Port Clearance Rules in PSOA RuleML: From Controlled-English Regulation to Ob...
 
RuleML 2015: When Processes Rule Events
RuleML 2015: When Processes Rule EventsRuleML 2015: When Processes Rule Events
RuleML 2015: When Processes Rule Events
 
Challenge@RuleML2015 Developing Situation-Aware Applications for Disaster Man...
Challenge@RuleML2015 Developing Situation-Aware Applications for Disaster Man...Challenge@RuleML2015 Developing Situation-Aware Applications for Disaster Man...
Challenge@RuleML2015 Developing Situation-Aware Applications for Disaster Man...
 
Rule Generalization Strategies in Incremental Learning of Disjunctive Concepts
Rule Generalization Strategies in Incremental Learning of Disjunctive ConceptsRule Generalization Strategies in Incremental Learning of Disjunctive Concepts
Rule Generalization Strategies in Incremental Learning of Disjunctive Concepts
 
RuleML 2015 Constraint Handling Rules - What Else?
RuleML 2015 Constraint Handling Rules - What Else?RuleML 2015 Constraint Handling Rules - What Else?
RuleML 2015 Constraint Handling Rules - What Else?
 
RuleML2015 The Herbrand Manifesto - Thinking Inside the Box
RuleML2015 The Herbrand Manifesto - Thinking Inside the Box RuleML2015 The Herbrand Manifesto - Thinking Inside the Box
RuleML2015 The Herbrand Manifesto - Thinking Inside the Box
 
Industry@RuleML2015: Norwegian State of Estate A Reporting Service for the St...
Industry@RuleML2015: Norwegian State of Estate A Reporting Service for the St...Industry@RuleML2015: Norwegian State of Estate A Reporting Service for the St...
Industry@RuleML2015: Norwegian State of Estate A Reporting Service for the St...
 
A Service for Improving the Assignments of Common Agriculture Policy Funds to...
A Service for Improving the Assignments of Common Agriculture Policy Funds to...A Service for Improving the Assignments of Common Agriculture Policy Funds to...
A Service for Improving the Assignments of Common Agriculture Policy Funds to...
 
Datalog+-Track Introduction & Reasoning on UML Class Diagrams via Datalog+-
Datalog+-Track Introduction & Reasoning on UML Class Diagrams via Datalog+-Datalog+-Track Introduction & Reasoning on UML Class Diagrams via Datalog+-
Datalog+-Track Introduction & Reasoning on UML Class Diagrams via Datalog+-
 
RuleML2015: Binary Frontier-guarded ASP with Function Symbols
RuleML2015: Binary Frontier-guarded ASP with Function SymbolsRuleML2015: Binary Frontier-guarded ASP with Function Symbols
RuleML2015: Binary Frontier-guarded ASP with Function Symbols
 
RuleML2015: API4KP Metamodel: A Meta-API for Heterogeneous Knowledge Platforms
RuleML2015: API4KP Metamodel: A Meta-API for Heterogeneous Knowledge PlatformsRuleML2015: API4KP Metamodel: A Meta-API for Heterogeneous Knowledge Platforms
RuleML2015: API4KP Metamodel: A Meta-API for Heterogeneous Knowledge Platforms
 
RuleML2015: Ontology-Based Multidimensional Contexts with Applications to Qua...
RuleML2015: Ontology-Based Multidimensional Contexts with Applications to Qua...RuleML2015: Ontology-Based Multidimensional Contexts with Applications to Qua...
RuleML2015: Ontology-Based Multidimensional Contexts with Applications to Qua...
 
RuleML2015: Compact representation of conditional probability for rule-based...
RuleML2015:  Compact representation of conditional probability for rule-based...RuleML2015:  Compact representation of conditional probability for rule-based...
RuleML2015: Compact representation of conditional probability for rule-based...
 
RuleML2015: Learning Characteristic Rules in Geographic Information Systems
RuleML2015: Learning Characteristic Rules in Geographic Information SystemsRuleML2015: Learning Characteristic Rules in Geographic Information Systems
RuleML2015: Learning Characteristic Rules in Geographic Information Systems
 
RuleML2015: User Extensible System to Identify Problems in OWL Ontologies and...
RuleML2015: User Extensible System to Identify Problems in OWL Ontologies and...RuleML2015: User Extensible System to Identify Problems in OWL Ontologies and...
RuleML2015: User Extensible System to Identify Problems in OWL Ontologies and...
 
RuleML2015: Representing Flexible Role-Based Access Control Policies Using Ob...
RuleML2015: Representing Flexible Role-Based Access Control Policies Using Ob...RuleML2015: Representing Flexible Role-Based Access Control Policies Using Ob...
RuleML2015: Representing Flexible Role-Based Access Control Policies Using Ob...
 
RuleML2015: Rule Generalization Strategies in Incremental Learning of Disjunc...
RuleML2015: Rule Generalization Strategies in Incremental Learning of Disjunc...RuleML2015: Rule Generalization Strategies in Incremental Learning of Disjunc...
RuleML2015: Rule Generalization Strategies in Incremental Learning of Disjunc...
 
Industry@RuleML2015 DataGraft
Industry@RuleML2015 DataGraftIndustry@RuleML2015 DataGraft
Industry@RuleML2015 DataGraft
 

Dernier

Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreternaman860154
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonAnna Loughnan Colquhoun
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptxHampshireHUG
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slidespraypatel2
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024The Digital Insurer
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonetsnaman860154
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountPuma Security, LLC
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfEnterprise Knowledge
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationRadu Cotescu
 
Boost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityBoost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityPrincipled Technologies
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsEnterprise Knowledge
 
Salesforce Community Group Quito, Salesforce 101
Salesforce Community Group Quito, Salesforce 101Salesforce Community Group Quito, Salesforce 101
Salesforce Community Group Quito, Salesforce 101Paola De la Torre
 
Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024The Digital Insurer
 
Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Allon Mureinik
 
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking MenDelhi Call girls
 
Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024The Digital Insurer
 
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j
 
Histor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slideHistor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slidevu2urc
 
08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking MenDelhi Call girls
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024Results
 

Dernier (20)

Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreter
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt Robison
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slides
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonets
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path Mount
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organization
 
Boost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityBoost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivity
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI Solutions
 
Salesforce Community Group Quito, Salesforce 101
Salesforce Community Group Quito, Salesforce 101Salesforce Community Group Quito, Salesforce 101
Salesforce Community Group Quito, Salesforce 101
 
Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024
 
Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)
 
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
 
Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024
 
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
 
Histor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slideHistor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slide
 
08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024
 

Ruleml2012 - Rule-based high-level situation recognition from incomplete tracking data

  • 1. Rule-Based High-Level Situation Recognition from Incomplete Tracking Data David Münch1, Joris IJsselmuiden1, Ann-Kristin Grosselfinger1, Michael Arens1, and Rainer Stiefelhagen1,2 1 Fraunhofer IOSB, Germany, david.muench@iosb.fraunhofer.de 2 Karlsruhe Institute of Technology, Germany. The 6th International Symposium on Rules - Rule ML2012, Montpellier, France, August 27-29, 2012 1 © Fraunhofer IOSB
  • 2. Motivation Motivation • Security systems: Persons and their behavior are the focus of attention: Person centric analysis • Threat detection • Visual surveillance • Activity logging • Video search • Driver Assistance Systems Input data is incomplete and noisy. 2 © Fraunhofer IOSB
  • 3. Overview • Cognitive Vision System as a whole. • High-level knowledge and situation recognition. • Handling Incomplete Data. • Experiments. 3 © Fraunhofer IOSB
  • 4. Cognitive Vision System 4 © Fraunhofer IOSB
  • 5. Cognitive Vision System 5 © Fraunhofer IOSB
  • 6. Cognitive Vision System 6 © Fraunhofer IOSB
  • 7. Cognitive Vision System 7 © Fraunhofer IOSB
  • 8. Cognitive Vision System 8 © Fraunhofer IOSB
  • 9. Cognitive Vision System 9 © Fraunhofer IOSB
  • 10. Conceptual Layer – Conceptual Primitives Level • Quantitative information (from Quantitative Layer) is transformed into primitive conceptual knowledge (Logic predicates). • Dictionary of basic rules. • Mainly domain independent. • Support of uncertainty and vagueness. • The rules in the CPL are mostly concerned with spatial relations and temporal relations on short time intervals. 10 © Fraunhofer IOSB
  • 11. Dictionary of basic rules Dictionary of basic rules for every domain. Fuzzy Metric Temporal Logic (FMTL): Extension of first order logic by notions of fuzziness, time, and metrics on time. Inference engine: F-LIMETTE 11 © Fraunhofer IOSB
  • 12. “Numbers” mapped to Concepts Fuzzy membership functions 𝜇 𝑠𝑝𝑒𝑒𝑑 𝑣𝑎𝑙𝑢𝑒 for the subset {zero, small, normal, high, very_high} of discrete conceptual speed values. Figure from: H.-H. Nagel, Steps toward a Cognitive Vision System, 2004, AI Mag. 12 © Fraunhofer IOSB
  • 13. Conceptual Layer – Behavior Representation Level How to represent the expected Situations? • Knowledge represented in Situation Graph Trees. • Graphically editable • Easy to extend and edit • Interpretable (vs. black box) • Exhaustive situation recognition. 13 © Fraunhofer IOSB
  • 14. Behavior Representation Level Basic logic predicates from Conceptual Primitives Level are aggregated and structured in Situation Graph Trees (SGT)  high-level conceptual situations. An SGT consists of situation schemes: • Can be start and/or end nodes. • Unique name. • State scheme (Precondition). • Action scheme (Postcondition). Specialize a situation scheme: • Prediction edges link to a possible subsequent situation scheme. • Specialization edges link to more specific situation graphs in a hierarchical structure. 14 © Fraunhofer IOSB
  • 15. Behavior Representation Level 15 © Fraunhofer IOSB
  • 16. Handling Incomplete Data – Low Level in Scene Domain Level Perfect input data. Truth value Time 16 © Fraunhofer IOSB
  • 17. Handling Incomplete Data – Low Level in SDL Incomplete input data. Truth value Time 17 © Fraunhofer IOSB
  • 18. Handling Incomplete Data – Low Level in SDL Interpolation of input data: Truth value Time 18 © Fraunhofer IOSB
  • 19. Handling Incomplete Data – Low Level in SDL Interpolation of input data: Time 19 © Fraunhofer IOSB
  • 20. Handling Incomplete Data – High Level in BRL 20 © Fraunhofer IOSB
  • 21. Handling Incomplete Data – High Level in BRL What if is_open(trunk, Car) fails? 21 © Fraunhofer IOSB
  • 22. Handling Incomplete Data – High Level in BRL Hallucination (abduction) of missing evidence. What if is_open(trunk, Car) fails?  Hallucinate is_open(trunk, Car) and continue! 22 © Fraunhofer IOSB
  • 23. VIRAT Video Dataset VIRAT Video Dataset Release 1.0 Input data: annotated ground truth: persons and their situations. Place: 0000: Place: 0002: Videos: 02 03 04 06 Videos: 00 06 23 © Fraunhofer IOSB
  • 24. Situations 1. Person loads object into car. 2. Person unloads object from car. 3. Person gets into car. 4. Person gets out of car. VIRAT Video Dataset Release 1.0 24 © Fraunhofer IOSB
  • 25. Evaluation • The annotated ground truth is regarded as complete information. • Randomly make gaps of a distinct length into the data. • Increase the amount of missing data in steps of 5%. • Repeat each experiment several times. 25 © Fraunhofer IOSB
  • 26. Evaluation Original, F-Score unmodified Gap size 5 seconds. Precision Recall Gap size Gap size 5 seconds. 5 seconds. 26 © Fraunhofer IOSB
  • 27. Evaluation Original, F-Score unmodified Gap size 5 seconds. Precision Recall Gap size Gap size 5 seconds. 5 seconds. 27 © Fraunhofer IOSB
  • 28. Evaluation ROC-curves for video (d) with gap size 5 seconds. TPR TPR FPR FPR With interpolation and hallucination. Without interpolation and hallucination. 28 © Fraunhofer IOSB
  • 29. Evaluation Video (d), F-Score Gap size 5 seconds. Gap size Gap size 5 seconds. 5 seconds. 29 © Fraunhofer IOSB
  • 30. Conclusion • low level: interpolation of data and its uncertainty.  ordinary incomplete data. • high level: extension of the situation recognition inference algorithm (hallucination, abduction).  high-level incomplete data such as occlusions. • Knowledge base for vehicle-centered situations. • Runs in real-time on off-the-shelf hardware. 30 © Fraunhofer IOSB
  • 31. Finis. 31 © Fraunhofer IOSB