Information Fusion Under Consideration of Conflicting Input Signals (eBook)

Dr.-Ing. Uwe Mönks studied Electrical Engineering and Information Technology at the OWL University of Applied Sciences (Lemgo), Halmstad University (Sweden), and Aalborg University (Denmark). Since 2009 he is employed at the Institute Industrial IT (inIT) as research associate with project leading responsibilities. During this time he completed his doctorate (Dr.-Ing.) in a cooperative graduation with Ruhr-University Bochum. His research interests are in the area of multisensor and information fusion, pattern recognition, and machine learning.

Kurzfassung 6
Abstract 8
Acknowledgements 9
Contents 10
List of Acronyms and Abbreviations 14
List of Symbols 16
1 Introduction 19
1.1 Motivation 24
1.2 Focus of the Work 25
1.3 Structure and Format 26
2 Scientific State of the Art 28
2.1 Information Fusion 28
2.1.1 Uncertainty 32
2.1.2 Conflict 34
2.2 Information Models 35
2.2.1 Probability Theory Fusion Approaches 36
2.2.2 Dempster-Shafer Theory of Evidence Fusion Approaches 39
2.2.3 Fuzzy Set Theory Fusion Approaches 43
2.2.4 Possibility Theory Fusion Approaches 46
2.2.5 Hybrid Information Fusion Approaches 47
2.2.6 Further Information Models 48
2.3 Human Group Decision-Making 49
2.4 Scientific Gap 50
2.5 Chapter Summary 52
3 Preliminaries 53
3.1 Modified-Fuzzy-Pattern-Classifier Membership Function Training 53
3.2 An Interconnection Between Dempster-Shafer, Fuzzy Set, and Possibility Theory 56
3.3 Two-Layer Conflict Solving 59
3.3.1 Conflict-Modified-DST 60
3.3.2 Group-Conflict-Redistribution 61
3.4 Fuzzy Aggregation 63
3.4.1 Ordered Weighted Averaging 64
3.4.2 Implicative Importance Weighted Ordered Weighted Averaging 66
3.5 Truncated Triangular Probability-Possibility Transform 67
3.6 Monitoring of Sensor Reliability 68
3.7 Chapter Summary 71
4 Multilayer Attribute-based Conflict-reducing Observation 72
4.1 The MACRO Architecture 73
4.2 Information Source Signal Conditioning 76
4.3 System State Representation 77
4.4 Fuzzy Basic Belief Assignment 80
4.5 Attribute Layer Fusion 82
4.5.1 Analysis of Two-Layer Conflict Solving 84
4.5.2 Balanced Two-Layer Conflict Solving 95
4.5.3 Fuzzified Balanced Two-Layer Conflict Solving 103
4.5.4 MACRO Attribute Layer Fusion 105
4.5.5 Conflicting Coefficient Behaviour 107
4.5.6 Conflict as a Measure of Importance 109
4.5.7 MACRO Attribute Structure 110
4.6 System Layer Fusion 110
4.6.1 Degree of Optimism 111
4.6.2 Attribute Importance 113
4.7 Sensor Defect Detection 114
4.7.1 Sensor Observation Determination 115
4.7.2 Measurement Scale Fuzzification 115
4.7.3 Majority Consistency Measure Adaptation 117
4.7.4 Groupwise Sensor Reliability Determination 118
4.7.5 Sensor Defect Decision Rule 119
4.8 Implementation Aspects 119
4.8.1 Matrix Notation 120
4.8.2 Matrix Decomposition 122
4.8.3 Computational Complexity 125
4.9 Chapter Summary 126
5 Evaluation 128
5.1 Implementations 129
5.2 Human Activity Recognition 129
5.2.1 Experiment Setup 131
5.2.2 Experiment Results 135
5.2.3 Discussion of the Results 141
5.3 Condition Monitoring Under Laboratory Conditions 142
5.3.1 Experiment Setup 145
5.3.2 PU static Data Set Results 148
5.3.3 PU manip Data Set Results 153
5.3.4 Discussion of the Results 160
5.4 Information Fusion Robustness Towards Noise 161
5.5 Sensor Defect Detection 164
5.5.1 PU static Data Set Results 165
5.5.2 PU manip Data Set Results 166
5.6 Chapter Summary 167
6 Summary 168
6.1 Conclusion 170
6.2 Future Work 172
6.2.1 Information Fusion System Design 173
6.2.2 Information Fusion System Composition and Adaptation 174
Appendix A Foundations of Probability Theory 177
Appendix B Foundations of Dempster-Shafer Theoryof Evidence 181
Appendix CFoundations of Fuzzy Set Theory 185
Appendix DProofs 187
D.1 Proofs of Section 4.4 187
D.2 Proofs of Section 4.5.1 188
D.3 Proofs of Section 4.5.2 191
D.4 Proofs of Section 4.5.3 191
D.5 Proofs of Section 4.8 196
Appendix E Compliance of the Fuzzy Basic Belief Assignment Approach with Dempster-Shafer Theory of Evidence 200
Appendix F Features Involved in Condition Monitoring Evaluation 204
F.1 Static Printing Unit Demonstrator Operation (PUstatic) 204
F.2 Manipulated Printing Unit Demonstrator Operation(PUmanip) 206
F.3 Noisy Manipulated Printing Unit Demonstrator Operation (PUmanip) 208
Appendix G Determination of OWA Weights with Desired Andness 213
Appendix HBrief Historical Background 215
H.1 Information Fusion 215
H.2 Fuzzy Set Theory 216
Bibliography 218
List of Figures 241
List of Tables 245
Theses 248