Distributed Video Sensor Networks

✍ Scribed by Bir Bhanu (editor), Chinya V. Ravishankar (editor), Amit K. Roy-Chowdhury (editor), Hamid Aghajan (editor), Demetri Terzopoulos (editor)

• Publisher: Springer
• Year: 2011
• Tongue: English
• Leaves: 476
• Category: Library

✦ Synopsis

Large-scale video networks are of increasing importance in a wide range of applications. However, the development of automated techniques for aggregating and interpreting information from multiple video streams in real-life scenarios is a challenging area of research.

Collecting the work of leading researchers from a broad range of disciplines, this timely text/reference offers an in-depth survey of the state of the art in distributed camera networks. The book addresses a broad spectrum of critical issues in this highly interdisciplinary field: current challenges and future directions; video processing and video understanding; simulation, graphics, cognition and video networks; wireless video sensor networks, communications and control; embedded cameras and real-time video analysis; applications of distributed video networks; and educational opportunities and curriculum-development.

Topics and features: presents an overview of research in areas of motion analysis, invariants, multiple cameras for detection, object tracking and recognition, and activities in video networks; provides real-world applications of distributed video networks, including force protection, wide area activities, port security, and recognition in night-time environments; describes the challenges in graphics and simulation, covering virtual vision, network security, human activities, cognitive architecture, and displays; examines issues of multimedia networks, registration, control of cameras (in simulations and real networks), localization and bounds on tracking; discusses system aspects of video networks, with chapters on providing testbed environments, data collection on activities, new integrated sensors for airborne sensors, face recognition, and building sentient spaces; investigates educational opportunities and curriculum development from the perspective of computer science and electrical engineering.

This unique text will be of great interest to researchers and graduate students of computer vision and pattern recognition, computer graphics and simulation, image processing and embedded systems, and communications, networks and controls. The large number of example applications will also appeal to application engineers.

✦ Table of Contents

Distributed Video Sensor Networks
Preface
Contents
Introduction
Distributed Video Sensor Networks and Research Challenges
Report on NSF/ARO/ONR Workshop on Distributed Camera Networks: Research Challenges and Future Directions
Introduction
Workshop Recommendations
A. Video Processing and Video Understanding
Calibration Risk and Payoffs
Nuisance Factors Risks and Payoffs
B. Simulation, Graphics, Cognition and Video Networks
C. Wireless Video Sensor Networks, Communications and Control
D. Distributed Embedded Cameras and Real Time Video Analysis
E. Applications
F. Educational Opportunities and Curriculum Development
Suggested Major Research Topics
Topic 1: Robust Scalable Video Networks for Wide Area Analysis
Disciplines Involved
Research Concentration
Topic 2: Active, Distributed and Communication Aware Video Sensor Networks
Disciplines Involved
Research Concentration
Topic 3: Large-scale Heterogeneous Sensor Networks for Wide Area Analysis
Disciplines Involved
Research Concentration
List of Attendees in Alphabetical Order
Groups and Group Leaders
Talks with Titles and Presenters
Video Processing and Understanding
Motion Analysis: Past, Present and Future
Introduction to Motion: An Early History
Motion: Highlights from Philosophy, Psychology and Neurobiology
Motion in Computer Vision: The Beginnings
Optical Flow-Based Motion Detection
Human Actions and Activities
Motion: Future
References
Projective Joint Invariants for Matching Curves in Camera Networks
Introduction
Related Work
Our Approach
Problem Formulation and Preliminaries
Joint-Invariant Signatures
Toward Local Signatures
Slices and Sections of Signature Manifold
Correspondence and Equivalence from Matching Sections
Picking Pivot Points
Matching Performance
Discussion
References
Multiple-View Object Recognition in Smart Camera Networks
Introduction
Contributions
Encoding Multiple-View Features via Sparse Representation
Random Projections
Enforcing Nonnegativity in l1-Minimization
Estimation of Joint Sparse Signals
System Implementation
Experiment
Conclusion and Discussion
References
A Comparison of Techniques for Camera Selection and Hand-Off in a Video Network
Introduction
Related Work and Contributions
Comparison for Existing Works
Our Contributions
Theoretical Comparison
Descriptions of the Key Ideas of Selected Approaches
The Utility-Based Game Theoretic Approach
The Co-occurrence to Occurrence Ratio (COR) Approach
The Constraint Satisfaction Problem (CSP) Approach
The Fuzzy-based Approach
Experimental Results
Data
Tracking
Parameters
Experimental Results and Analysis
Conclusions and Future Work
References
Distributed Sensing and Processing for Multi-Camera Networks
Introduction
Robust Statistical Inference
Computationally Efficient and Distributed Algorithms
Opportunistic and Parsimonious Sensing
Statistical Inference for Tracking
Homography
Detection
Multi-View Fusion and Tracking
Efficient Particle Filtering
Particle Filtering: A Brief Overview
Metropolis Hastings Algorithm
Particle Filtering with IMHA-Based Resampling
Experimental Results
Compressive Sensing
Compressive Sensing
Compressive Background Subtraction
Multi-View Ground-Plane Tracking
Conclusions and Future Directions
Distributed Bayesian Inference
Manifold-Based Dimensionality Reduction (NLDR)
References
Tracking of Multiple Objects over Camera Networks with Overlapping and Non-overlapping Views
Introduction
Related Work
Tracking within a Single Camera
Detection of Occlusion and Segmentation Errors
Measurement Selection via Segmented VOs
Measurement Selection via Adaptive Particle Sampling
Adaptive Appearance
Tracking Across Multiple Cameras
Establish Field of View (FOV) Lines
Extracting Landmark Points
Findling Matching Landmark Points
Aligning Two Images
Brightness Calibration of Neighboring Cameras
Consistent Labeling Across Cameras
Experimental Results
Performance of Tracking within a Single Camera
Performance of Tracking Across Multiple Cameras
Conclusion
References
Toward Robust Online Visual Tracking
Introduction
Appearance Modeling for Visual Tracking
Learning Nonlinear Appearance Manifold
Learning Nonlinear Manifold Online
Online Update of Submanifold
Leveraging Prior Knowledge with Online Learning
Learning Detectors Online for Visual Tracking
Multiple Instance Learning
Learning Detectors with Online Multiple Instance Boosting
Articulated Objects
Conclusions
References
Modeling Patterns of Activity and Detecting Abnormal Events with Low-Level Co-occurrences
Introduction
Context, Overview and Notations
Context
Overview and Notation
Our Method
Training Phase
Nominal Model
Learning the Co-occurrence Matrix
A Specific Case for Co-occurrence
Complexity Issues & Conditional Independence
Observation Phase
Abnormal Model
Abnormality Detection
Experimental Results
Conclusion
References
Use of Context in Video Processing
Introduction
Case Study: Environment Discovery
Environmental Context
Camera Priors in Activity Recognition
Object Recognition Through User Activities
User-Based Context
Adapting Domain Knowledge from User Feedback
Conclusion
References
Simulation, Graphics, Cognition and Video Networks
Virtual Vision
Introduction
The Case for Virtual Vision
Related Work
Smart Camera Nodes
Synthetic Video
Visual Processing
Camera Node Behavioral Controller
Surveillance Systems
Active Camera Scheduling
Collaborative Persistent Surveillance
Conclusions
References
Virtualization and Programming Support for Video Sensor Networks with Application to Wireless and Physical Security
Motivation
Related Work
Wireless Intrusion Detection
Intrusion Detection Systems
Wireless Intrusion Detection Systems
SNBench Overview
Enabling Wireless Monitoring
WifiAlertSensor
WifiActivitySensor
WifiResponder
Deployment Environment
Service Programming Primer
Wireless Security Services
Future Work and Conclusions
Wireless Access Lists from Physical Data
snBench as a Complete, Turn-Key Network Security Solution
In Conclusion
References
Simulating Human Activities for Synthetic Inputs to Sensor Systems
Overview
The CAROSA System
Related Work
Parameterized Representations
Resource Management
Roles and Groups
Scenario Authoring
Example Simulation
CAROSA Summary
Input to Distributed Sensor Networks
Summary
References
Cognitive Sensor Networks
Introduction
Cognition
The Domain Theory Hypothesis
Sensor Networks
Domain Theory-Based Perception
Symmetry Theory in Signal Processing
Conclusion
References
Ubiquitous Displays: A Distributed Network of Active Displays
Introduction
State of the Art: Centralized Displays
Passive Multi-Displays
Single Active Displays
Bottleneck of Centralized Systems
Disruptive Change in Display Metaphor
Active Displays
Interaction with Environment
Interaction with User and Data
Initial Progress
Distributed Self-Calibration of Planar Display Walls
Color Registration Amenable to Parallelization
Distributed Interaction with 2D Applications on Planar Display Walls
Projector Camera Self Calibration Techniques
Registering Constrained Non-Planar Displays Using a Single Uncalibrated Camera
Conclusion
References
Wireless Video Sensor Networks, Communications and Control
Research Challenges for Wireless Multimedia Sensor Networks
Introduction
Applications of Wireless Multimedia Sensor Networks
Network Architecture
Factors Influencing the Design of Multimedia Sensor Networks
Application Layer
Multimedia Encoding Techniques
Distributed Video Coding
Video Encoding Based on Compressed Sensing
Collaborative In-network Processing
Transport Layer Protocols
UDP Based Protocols
TCP and TCP Friendly Schemes for WMSNs
Distortion-Minimizing Rate Control
Network Layer
MAC Layer
Contention-Based Protocols
Contention-free Single Channel Protocols
Physical Layer
Conclusions
References
Camera Control and Geo-Registration for Video Sensor Networks
Introduction
Related Work
PTZ Camera Viewspace Control Model
Scene-Based Camera Geo-Registration and Mapping
Operational Interface
Summary
References
Persistent Observation of Dynamic Scenes in an Active Camera Network
Introduction
Technical Rationale
Necessity of Collaboration in an Active Camera Network
Necessity of a Decentralized Strategy
Relation to Previous Work
Cooperative Target Acquisition Using Game Theory
Motivation for Game-Theoretic Formulation
Precise Problem Statement and Notation
Game-Theory Fundamentals
Choice of Utility Functions
Target Utility
Global Utility
Camera Utility
Negotiation Mechanisms
Application of SAP Negotiation Mechanism
Experimental Results
Conclusion
References
Proactive PTZ Camera Control
Introduction
Related Work
Proactive Camera Control
Problem Statement
Finding Good State Sequences
PTZ Camera Relevance
State Sequence Quality
Planning
Finding an Optimal Sequence
Results
Conclusions and Future Work
References
Distributed Consensus Algorithms for Image-Based Localization in Camera Sensor Networks
Introduction
Chapter Contributions
Related Work
Review of Average-Consensus Algorithms
Distributed Object Localization
Consensus on SO(3)
Consensus on so(3)
Simulation Results
Distributed CSN Localization
Estimation of the Rotations
Estimation of the Translations
Complete Estimation
Simulation Results
Conclusion
References
Conditional Posterior Cramér-Rao Lower Bound and its Applications in Adaptive Sensor Management
Introduction
Information Theoretic Measures
PCRLB
Conditional PCRLB for Recursive Nonlinear Filtering
CRLB
Unconditional PCRLB
Conditional PCRLB
Recursive Conditional PCRLB
C-PCRLB-Based Sensor Management
Adaptive Sensor Selection for Iterative Source Localization
Dynamic Sensor Selection for Tracking
Applications in Camera Network Management
References
Distributed Embedded Cameras and Real-Time Video Analysis
VideoWeb: Optimizing a Wireless Camera Network for Real-time Surveillance
Introduction
Related Work and Contributions
Building the Camera Network
Choosing the Type of Network
Choosing the Right Camera
Choosing and Configuring the Network Hardware
The VideoWeb Wireless Camera Network
Experiments for Performance Characterization and Optimization of the Video Network
Optimizing Camera Configuration
Multi-objective Optimization Using Pareto Efficiency
Inferiority and Non-Inferiority
Data Collection
Evaluation Results
Conclusions
References
VideoWeb Dataset for Multi-camera Activities and Non-verbal Communication
Introduction
Data Collection
Purpose and Significance of Data
Environment for Data
Contents of Data
Examples
Ground-Truth Annotations
Availability of the Data
Conclusions
References
Wide-Area Persistent Airborne Video: Architecture and Challenges
Introduction
Spatio-temporal Reflectance Variations
Wide Aperture Imaging Model of Camera Arrays
Seamless Stitchable Camera Arrays
Geometric Properties of WFOV Imaging Arrays
Physical Considerations Governing Camera-Array-based WFOV Virtual Focal Planes
Accommodating Dynamic Variations in Operational Camera Arrays Using Pose Information
Summary and Conclusions
References
Collaborative Face Recognition Using a Network of Embedded Cameras
Introduction
Contributions and Main Results
Outline of the Paper
Related Work
Experimental Setup
System Model
Assembly of Camera Platform
Assembly of Embedded Camera Network
Software Implementation
Experimentation
System Performance
Network Performance
Face Recognition Performance
Real-time Capability
Conclusions and Future Work
References
SATware: A Semantic Approach for Building Sentient Spaces
Introduction
SATware: An Middleware Framework for Sentient Spaces
A Programming Model for Pervasive Applications
Virtual Sensors: Bridging Application Needs to Raw Sensor Streams
Query Processing in SATware
Supporting Scalability through Semantic Scheduling
Supporting Robustness through Sensor Recalibration
Conclusions
References
Applications of Distributed Video Networks
Video Analytics for Force Protection
Aerial Video Analysis
Stabilization
Object Detection
Tracking
Super-resolution
Tracking from Fixed Ground Based Cameras
Person Detection from Moving Platforms
Biometrics at a Distance
Face Capture and Recognition
Iris Recognition
Whole-Body Re-Identification
Facial Analysis
Summary
References
Recognizing Activity Structures in Massive Numbers of Simple Events Over Large Areas
Introduction
Spatial Structure
Temporal Structure
Event-Linkage Structure
Short Event-Sequence Structure
Network Structure
Summary
References
Distributed Sensor Networks for Visual Surveillance
Introduction
Technical Challenges in Large Sensor Networks
System Design and Components
Auto Camera Calibration and Geo-Registration
Video Processing
Efficient Processing of High-Resolution Imagery
Context Learning
Environmental Conditions
Scene Types and Elements
Target Property Models
Data Fusion and Event Inference
User Interface
Results
References
Ascertaining Human Identity in Night Environments
Introduction
Color-NIR Cross-Spectral Iris Matching
Multispectral Iris Dataset and Data Used in Simulations
Proposed Predictive Model
Recognition Performance
Short Wave Infrared Face Verification
SWIR Data Collection
Methodology
Preprocessing and Normalization
Matching Experiments
Results
SWIR Face Verification
Gait Curves for Human Recognition in a Night-Time Environment
Methodology
Preprocessing and Silhouette Extraction
Spatio-temporal Feature Extraction
Experiments and Results
Classification of Human Gait
Discussion
Soft Biometrics-Body Measurement
Summary
References
Educational Opportunities and Curriculum Development
Educational Opportunities in Video Sensor Networks
Introduction
Computational Sensor Networks
Engineering Background for Video Sensor Networks
Course Organization
Support Technology for Instruction
Conclusion
Recommended Courses and Topics
Machine Vision
Sensor Networks
Hardware Systems
Software Systems
References
Index