Isupova Machine Learning Methods for Behaviour Analysis and Anomaly Detection in Video

1. Auflage 2018
ISBN: 978-3-319-75508-3
Verlag: Springer International Publishing
Format: PDF
Kopierschutz: 1 - PDF Watermark

E-Book, Englisch, 144 Seiten

Reihe: Springer Theses

This thesis proposes machine learning methods for understanding scenes via behaviour analysis and online anomaly detection in video. The book introduces novel Bayesian topic models for detection of events that are different from typical activities and a novel framework for change point detection for identifying sudden behavioural changes.Behaviour analysis and anomaly detection are key components of intelligent vision systems. Anomaly detection can be considered from two perspectives: abnormal events can be defined as those that violate typical activities or as a sudden change in behaviour. Topic modelling and change-point detection methodologies, respectively, are employed to achieve these objectives.The thesis starts with the development of learning algorithms for a dynamic topic model, which extract topics that represent typical activities of a scene. These typical activities are used in a normality measure in anomaly detection decision-making. The book also proposes a novel anomaly localisation procedure. In the first topic model presented, a number of topics should be specified in advance. A novel dynamic nonparametric hierarchical Dirichlet process topic model is then developed where the number of topics is determined from data. Batch and online inference algorithms are developed.The latter part of the thesis considers behaviour analysis and anomaly detection within the change-point detection methodology. A novel general framework for change-point detection is introduced. Gaussian process time series data is considered. Statistical hypothesis tests are proposed for both offline and online data processing and multiple change point detection are proposed and theoretical properties of the tests are derived. The thesis is accompanied by open-source toolboxes that can be used by researchers and engineers.

Autoren/Hrsg.

Isupova, Olga

Inhaltsverzeichnis

1;Supervisor’s Foreword;6
2;Abstract;8
3;Acknowledgements;9
4;Contents;10
5;Acronyms;13
6;Notation;14
7;List of Figures;18
8;List of Tables;21
9;1 Introduction;22
9.1;1.1 Abnormal Behaviour Detection;23
9.1.1;1.1.1 Topic Modeling;23
9.1.2;1.1.2 Change Point Detection;24
9.2;1.2 Key Contributions and Outline;25
9.3;1.3 Disseminated Results;27
9.4;References;28
10;2 Background;29
10.1;2.1 Outline of Video Processing Methods;29
10.1.1;2.1.1 Object Detection;29
10.1.2;2.1.2 Object Tracking;34
10.2;2.2 Anomaly Detection;36
10.2.1;2.2.1 Video Representation;37
10.2.2;2.2.2 Behaviour Model;38
10.2.3;2.2.3 Normality Measure;40
10.3;2.3 Topic Modeling;41
10.3.1;2.3.1 Problem Formulation;41
10.3.2;2.3.2 Inference;42
10.3.3;2.3.3 Extensions of Conventional Models;45
10.3.4;2.3.4 Dynamic Topic Models;45
10.3.5;2.3.5 Topic Modeling Applied to Video Analytics;46
10.4;2.4 Change Point Detection;47
10.4.1;2.4.1 Change Point Detection in Time Series Data;47
10.4.2;2.4.2 Anomaly as Change Point Detection;48
10.5;2.5 Summary;49
10.6;References;49
11;3 Proposed Learning Algorithms for Markov Clustering Topic Model;56
11.1;3.1 Video Representation;57
11.2;3.2 Model;58
11.2.1;3.2.1 Motivation;58
11.2.2;3.2.2 Model Formulation;59
11.3;3.3 Parameter Learning;61
11.3.1;3.3.1 Expectation-Maximisation Learning;62
11.3.2;3.3.2 Variational Inference;65
11.3.3;3.3.3 Gibbs Sampling;67
11.3.4;3.3.4 Similarities and Differences of the Learning Algorithms;68
11.4;3.4 Anomaly Detection;68
11.4.1;3.4.1 Abnormal Documents Detection;69
11.4.2;3.4.2 Localisation of Anomalies;71
11.5;3.5 Performance Validation;72
11.5.1;3.5.1 Performance Measure;74
11.5.2;3.5.2 Parameter Learning;74
11.5.3;3.5.3 Anomaly Detection;75
11.6;3.6 Summary;81
11.7;References;83
12;4 Dynamic Hierarchical Dirichlet Process;84
12.1;4.1 Hierarchical Dirichlet Process Topic Model;85
12.1.1;4.1.1 Chinese Restaurant Franchise;86
12.2;4.2 Proposed Dynamic Hierarchical Dirichlet Process Topic Model;88
12.3;4.3 Inference;89
12.3.1;4.3.1 Batch Collapsed Gibbs Sampling;90
12.3.2;4.3.2 Online Inference;93
12.4;4.4 Anomaly Detection;94
12.5;4.5 Experiments;95
12.5.1;4.5.1 Synthetic Data;96
12.5.2;4.5.2 Real Video Data;97
12.6;4.6 Summary;100
12.7;References;100
13;5 Change Point Detection with Gaussian Processes;1
13.1;5.1 Problem Formulation;103
13.1.1;5.1.1 Data Model;103
13.1.2;5.1.2 Change Point Detection Problem Formulation;104
13.2;5.2 Gaussian Process Change Point Detection Approach Based on Likelihood Ratio Tests;105
13.2.1;5.2.1 Likelihood Ratio Test;105
13.2.2;5.2.2 Generalised Likelihood Ratio Test;106
13.2.3;5.2.3 Discussion;107
13.3;5.3 Gaussian Process Online Change Point Detection Approach Based on Likelihood Estimation;108
13.3.1;5.3.1 Test Formulation;108
13.3.2;5.3.2 Theoretical Evaluation of the Test;109
13.3.3;5.3.3 Test with Estimated Hyperparameters;111
13.3.4;5.3.4 Discussion;111
13.4;5.4 Performance Validation on Synthetic Data;112
13.4.1;5.4.1 Data Simulated by the Proposed Generative Model;113
13.4.2;5.4.2 Data Simulated by the GP-BOCPD Model;118
13.5;5.5 Numerical Experiments with Real Data;119
13.6;5.6 Summary;122
13.7;References;123
14;6 Conclusions and Future Work;124
14.1;6.1 Summary of Methods and Contributions;124
14.2;6.2 Suggestions for Future Work;126
14.2.1;6.2.1 Inference in Topic Modeling;126
14.2.2;6.2.2 Alternative Dynamics in Topic Modeling;126
14.2.3;6.2.3 Gaussian Process Change Point Detection;127
14.2.4;6.2.4 Potential Applications of the Proposed Statistical Methods;128
14.3;References;128
15;A EM for MCTM Derivation;130
16;Appendix B VB for MCTM Derivation;134
17;Appendix C Distributions of Quadratic Forms;137
18;C.1 Quadratic form of the ``Own'' Covariance Matrix;137
19;C.2 Quadratic form of an Arbitrary Symmetric Matrix;139
20;Appendix D Proofs of the Theorems for the Proposed Test Statistic;141
21;D.1 Proof of Theorem 5.1;141
22;D.2 Proof of Theorem 5.2;141
23;Appendix E Optimisation of Gaussian Process Covariance Function Hyperparameters;143
24;References;144