Sensor and Data Fusion Sensor and Data Fusion Edited by Dr. ir. Nada Milisavljević I-Tech IV Published by In-Teh In-Teh is Croatian branch of I-Tech Education and Publishing KG, Vienna, Austria. Abstracting and non-profit use of the material is permitted with credit to the source. Statements and opinions expressed in the chapters are these of the individual contributors and not necessarily those of the editors or publisher. No responsibility is accepted for the accuracy of information contained in the published articles. Publisher assumes no responsibility liability for any damage or injury to persons or property arising out of the use of any materials, instructions, methods or ideas contained inside. After this work has been published by the In-Teh, authors have the right to republish it, in whole or part, in any publication of which they are an author or editor, and the make other personal use of the work. © 2009 In-teh www.in-teh.org Additional copies can be obtained from: publication@ars-journal.com First published February 2009 Printed in Croatia p. cm. ISBN 978-3-902613-52-3 1. Sensor and Data Fusion, Dr. ir. Nada Milisavljević Preface Data fusion is a research area that is growing rapidly due to the fact that it provides means for combining pieces of information coming from different sources/sensors, resulting in ameliorated overall system performance (improved decision making, increased detection capabilities, diminished number of false alarms, improved reliability in various situations at hand) with respect to separate sensors/sources. Different data fusion methods have been developed in order to optimize the overall system output in a variety of applications for which data fusion might be useful: security (humanitarian, military), medical diagnosis, environmental monitoring, remote sensing, robotics Generally speaking, there is no fusion approach that works better than the others, but depending mainly on the types (quality, quantity) of data, some approaches might be better suited to a particular problem than the others. Actually, the choice of the combination method which is best-suited for a particular application is related to types of sources/sensors, types of data (numerical, symbolic, knowledge-based; maps, legends, historical information…), amounts of data available etc. As a consequence, various data fusion techniques have been investigated for years, such as probability theory, fuzzy logic, possibility theory, evidence theory (Dempster-Shafer, belief functions). In addition, depending mainly on the types of data processing involved and on the problem itself, several levels of data fusion exist in general (e.g., pixel level, feature level, decision level, or object assessment, situation assessment, impact assessment). In different data fusion books and articles, we can also find various data fusion architectures, having three main classes: centralized, decentralized and hybrid. As a result of this variety of techniques, architectures, levels, etc., data fusion is able to bring solutions in various areas of diverse disciplines. The goal of this book is to provide highlights of the current research in the field of data fusion. The book consists of twenty-five research papers, addressing various problems in areas such as: target tracking (including adaptive sensor management, data association and road obstacle tracking), obstacle detection for the railway traffic, real-time traffic state estimation, air traffic control, automotive applications (e.g., car safety and driver assistance), robotic systems, smoke detectors and home security, industrial instrumentation and process monitoring, remote sensing (vegetation indices, update of scarce high resolution images with time series of coarser images, land cover classification), medical imaging, anomaly detection and behavior prediction, environmental monitoring including forest fires and electromagnetic pollution, change detection, (distributed, wireless) sensor networks, etc. The list of possible applications is, actually, large, since most of the methodologies presented in the book can be adapted easily to a variety of problems and situations. The techniques involved cover a wide range of classical or novel methods, from different Bayesian-based approaches via Dempster-Shafer evidence theory and fuzzy logic to artificial neural networks and multi-agent based fusion methods. Depending mainly on the technique, VI different strategies for including expert knowledge and other collateral sources of information in the fusion process are also investigated. Altogether, the book aims to provide a valuable source of up-to-date data fusion methods, systems, applications and tools. As such, it may be useful to researchers, engineers, computer scientists, as well as to undergraduate and graduate students who are interested in the latest developments in the data fusion field. The editor is thankful to the contributors for their precious work towards the realization of this book as well as to Dr. Vedran Kordić for his valuable help. January 2009 Editor Dr. ir. Nada Milisavljević Department of Communication, Information, Systems and Sensors Royal Military Academy, Brussels, Belgium Contents Preface V 1. Advanced Sensor and Dynamics Models with an Application to Sensor Management 001 Wolfgang Koch 2. Target Data Association Using a Fuzzy-Logic Based Approach 035 Stephen Stubberud and Kathleen Kramer 3. Data Fusion Performance Evaluation for Dissimilar Sensors: Application to Road Obstacle Tracking 057 Blanc Christophe, Checchin Paul, Gidel Samuel and Trassoudaine Laurent 4. IR Barrier Data Integration for Obstacle Detection 071 J. Jesús García, Jesús Ureña, Manuel Mazo and Álvaro Hernández 5. A Model of Federated Evidence Fusion for Real-Time Traffic State Estimation 089 Qing-Jie Kong and Yuncai Liu 6. Multi Sensor Data Fusion Architectures for Air Traffic Control Applications 103 Baud Olivier, Gomord Pierre, Honoré Nicolas, Ostorero Loïc, Taupin Olivier and Tubery Philippe 7. Sensor Data Fusion in Automotive Applications 123 Panagiotis Lytrivis, George Thomaidis and Angelos Amditis 8. Multisensor Data Fusion Strategies for Advanced Driver Assistance Systems 141 Mahdi Rezaei Ghahroudi and Reza Sabzevari 9. Trajectory Generation and Object Tracking of Mobile Robot Using Multiple Image Fusion 167 TaeSeok Jin and Hideki Hashimoto VIII 10. Multisensory Data Fusion for Ubiquitous Robotics Services 177 Ren C. Luo and Ogst Chen 11. Design of an Intelligent Housing System Using Sensor Data Fusion Approaches 191 Arezou_Moussavi Khalkhali, Behzad_ Moshiri, Hamid Reza_ Momeni 12. Model-based Data Fusion in Industrial Process Instrumentation 201 Gerald Steiner 13. Multi-Sensor Data Fusion in Presence of Uncertainty and Inconsistency in Data 225 Manish Kumar and Devendra P. Garg 14. Updating Scarce High Resolution Images with Time Series of Coarser Images: a Bayesian Data Fusion Solution 245 Dominique Fasbender, Valérie Obsomer, Patrick Bogaert and Pierre Defourny 15. Multi-Sensor & Temporal Data Fusion for Cloud-Free Vegetation Index Composites 263 Bijay Shrestha, Charles O’Hara and Preeti Mali 16. Three Strategies for Fusion of Land Cover Classification Results of Polarimetric SAR Data 277 Nada Milisavljević, Isabelle Bloch, Vito Alberga and Giuseppe Satalino 17. Multilevel Information Fusion: A Mixed Fuzzy Logic/Geometrical Approach with Applications in Brain Image Processing 299 Julien Montagner and Vincent Barra 18. Anomaly Detection & Behavior Prediction: Higher-Level Fusion Based on Computational Neuroscientific Principles 323 Bradley J. Rhodes, Neil A. Bomberger, Majid Zandipour, Lauren H. Stolzar, Denis Garagic, James R. Dankert and Michael Seibert 19. A Biologically Based Framework for Distributed Sensory Fusion and Data Processing 337 Ferro M. and Pioggia G. 20. Agent Based Sensor and Data Fusion in Forest Fire Observer 365 Ljiljana Šerić, Darko Stipaničev and Maja Štula IX 21. A Sensor Data Fusion Procedure for Environmental Monitoring Applications by a Configurable Network of Smart Web-Sensors 379 Claudio De Capua and Rosario Morello 22. Monitoring Changes in Operational Scenarios via Data Fusion in Sensor Networks 401 Papantoni-Kazakos, Dr. Titsa and Burrell, Dr. Anthony 23. Elements of Sequential Detection with Applications to Sensor Networks 417 Stefano Marano and Vincenzo Matta 24. Parameter Estimation Over Noisy Communication Channels in Distributed Sensor Networks 437 Thakshila Wimalajeewa, Sudharman K. Jayaweera and Carlos Mosquera 25. Monte Carlo Methods for Node Self-Localization and Nonlinear Target Tracking in Wireless Sensor Networks 471 Joaquín Míguez, Luis Arnaiz and Antonio Artés-Rodríguez [...]... no means trivial B Aspects of Sensor and Data Fusion Among the primary technical prerequisites for sensor data and information fusion are communication links with a sufficient bandwidth, small latency, and robustness against failure or jamming Moreover, the transformation of the sensor data into a common coordinate system requires a precise space-time registration of the sensors, including their mutual... Sensor and Data Fusion given all sensor data and a priori information available In those applications data acquisition and tracking are completely decoupled For phased-array radar, however, the current signal-to-noise ratio of the object (i.e the detection probability) strongly depends on the correct positioning of the pencil-beam, which is now taken into the responsibility of the tracking system Sensor. .. planning data, tactics) and human observer reports (HUMINT: Human Intelligence) is also important information in the fusion process [4] The exploitation of context information of any kind can significantly improve the fusion system performance Fig 1 Sensor data and information fusion for situation pictures: overview of characteristic aspects and their mutual interrelation Advanced Sensor and Dynamics... wide range and may be chosen individually for each track Of special interest are military air situations where both agile objects and objects significantly differing in their radar cross section must be taken into account Unless properly handled, such situations can be highly allocation time- and energyconsuming In this context, advanced sensor and dynamics models for combined tracking and sensor management...1 Advanced Sensor and Dynamics Models with an Application to Sensor Management Wolfgang Koch German Defence Research Establishment (FGAN e.V.) Germany 1 Introduction The methods provided by sensor and data fusion [14] are important tools for fusing large sets of mutually complementary data end efficiently exploiting the sensor systems available A challenging exploitation... analyzing the ‘time series’ created by the sensor data • Problem: The sensor information is inaccurate, incomplete, and possibly even ambiguous Moreover, the objects’ temporal evolution is usually not well-known • Approach: Interpret sensor measurements and object state vectors as random variables Describe by probability density functions (pdfs) what is known about these random variables • Solution: Derive... revisit time and the corresponding radar beam position, rangeand Doppler-gates, or the type of the radar wave forms to be transmitted Track search requests require the setting of appropriate revisit intervals, search sectors and patterns, and other radar parameters In the dwell scheduling unit these preparations are transformed into antenna commands, by which the radar sensor is allocated and radar energy... road map extraction In the case of multifunctional sensors, feedback exists from the tracking system to the process of sensor data acquisition (sensor management) D A Characteristic Application: Sensor Management Modern multifunctional agile-beam radar based on phased-array technology is an excellent example for a sensor system that requires sophisticated sensor management algorithms This is particularly... tracking/fusion system along with its relation to the underlying sensors After passing a detection process, essentially working as a means of data rate reduction, the signal processing provides 4 Sensor and Data Fusion estimates of parameters characterizing the waveforms received at the sensors’ front ends (e.g radar antennas) From these estimates sensor reports are created, i.e measured quantities possibly... error covariance matrix The Bayesian formalism and the sensor model (likelihood function) obviously define how a negative sensor output, i.e a missing detection is to be processed In the case of well-separated objects in the presence of false returns and imperfect detection, the nk sensor data Zk are also not longer uniquely interpretable Let ik = 0 denote the data interpretation hypothesis that the object . TaeSeok Jin and Hideki Hashimoto VIII 10 . Multisensory Data Fusion for Ubiquitous Robotics Services 17 7 Ren C. Luo and Ogst Chen 11 . Design. of Sensor and Data Fusion Among the primary technical prerequisites for sensor data and information fusion are communication links with a sufficient bandwidth,