Tracking with Particle Filter for High-dimensional Observation and State Spaces

Séverine Dubuisson is Assistant Professor in the Institute for Intelligent Systems and Robotics, at Pierre and Marie Curie University in Paris, France. Her research interests include tracking in video sequences, management of large state spaces and observation, analysis of the dynamics in video sequences and analysis and modeling of social interactions.

• NOTATIONS ix INTRODUCTION xiCHAPTER 1. VISUAL TRACKING BY PARTICLE FILTERING 11.1. Introduction 11.2. Theoretical models 21.2.1. Recursive Bayesian filtering 21.2.2. Sequential Monte-Carlo methods 41.2.3. Application to visual tracking 81.3. Limits and challenges 181.4. Scientific position 221.5. Managing large sizes in particle filtering 221.6. Conclusion 26CHAPTER 2. DATA REPRESENTATION MODELS 292.1. Introduction 292.2. Computation of the likelihood function    302.2.1. Exploitation of the spatial redundancy   312.2.2. Exploitation of the temporal redundancy 422.3. Representation of complex information    502.3.1. Representation of observations for movement detection, appearances and disappearances 502.3.2. Representation of deformations 532.3.3. Multifeature representation 562.4. Conclusion 75CHAPTER 3. TRACKING MODELS THAT FOCUS ON THE STATE SPACE 793.1. Introduction 793.2. Data association methods for multi-object tracking  803.2.1. Particle filter with adaptive classification  843.2.2. Energetic filter for data association    873.3. Introducing fuzzy information into the particle filter 953.3.1. Fuzzy representation   963.3.2. Fuzzy spatial relations  983.3.3. Integration of fuzzy spatial relations into the particle filter   993.4. Conjoint estimation of dynamic and static parameters 1143.5. Conclusion 119CHAPTER 4. MODELS OF TRACKING BY DECOMPOSITION OF THE STATE SPACE  1234.1. Introduction 1234.2. Ranked partitioned sampling       1264.3. Weighted partitioning with permutation of sub-particles 1334.3.1. Permutation of sub-samples 1354.3.2. Decrease the number of resamplings 1384.3.3. General algorithm and results 1384.4. Combinatorial resampling 1424.5. Conclusion 149CHAPTER 5. RESEARCH PERSPECTIVES IN TRACKING AND MANAGING LARGE SPACES 1515.1. Tracking for behavioral analysis: toward finer tracking of the “future” and the “now”  1535.2. Tracking for event detection: toward a top-down model 1565.3. Tracking to measure social interactions 159BIBLIOGRAPHY 163INDEX   197