Preface

Multisensor data fusion seeks to combine information from multiple sources (including sensors, human reports, and data from the Internet) to achieve inferences that cannot be obtained from a single sensor or source, or whose quality exceeds that of an inference drawn from any single source. Historically, the U.S. Department of Defense (DoD) and other defense establishments around the world invested enormous amounts of funds for data fusion systems in applications such as automatic target recognition, target tracking, automated situation assessment, identification-friend-foe-neutral (IFFN) systems, and smart weapons. This investment has resulted in an extensive legacy of data fusion technological capability that combines process models, algorithms, evolving tool kits, and system engineering methodologies (e.g., system design and algorithm selection). Of particular note is the U.S. Joint Directors of Laboratories (JDL) Data Fusion Working Group process model, which has been the foundation for characterizing hierarchical levels of fusion processing, categorized fusion functions, and candidate algorithm approaches. The JDL model continues to evolve and expand to meet the demands of DoD and non-DoD applications.

During the 1990s, the data fusion research and development community commonly termed multisensor fusion as an emerging technology, meaning that there were numerous outstanding issues related to algorithm design and selection, fusion system architectures, test, and evaluation, establishing the utility of fusion systems, and other issues. Indeed, the creation of the JDL model in the early 1990s was in part motivated by a need to establish common definitions and understandings of this emerging technology to guide researchers and system designers. Although all issues have not yet been resolved, multisensor fusion can no longer be considered as an emerging technology. Fusion development has had to keep pace with a proliferation of micro- and nanoscale sensors, increasing speed and availability of wireless communications, and increased computing speeds, enabling the assembly of information from sensors, models, and human input. Fusion research and applications have adapted to service-oriented architectures, pushed the boundaries of situational modeling of human behavior, and expanded into fields such as chemical and biological sensing, medical diagnostics, environmental engineering, crisis management, monitoring and control of manufacturing processes, and intelligent buildings. In the defense and national security arenas, requirements and capabilities have moved from traditional force-on-force, nation-state-centric problems to a broad array of new and yet more challenging problems of asymmetric warfare, regional conflict, counterterrorism, and a host of specialized problems. The information spaces of these application domains are extremely large, and their demands for timely and sufficiently accurate inferences have required extraordinary creativity and agility from the data fusion community to respond with a multitude of inventive solutions.

This second edition to the original Handbook of Multisensor Data Fusion captures the latest data fusion concepts and techniques drawn from a broad array of disciplines including statistical estimation; signal and image processing; artificial intelligence; and biological, social, and information sciences. Leading experts in the fusion community have contributed to this new edition. This new edition has been expanded to 31 chapters and adds a wide range of new topics that represent the new technology directions. Many of the chapters brought into this edition from the first edition have been rewritten and updated. New material involves extensions into service-oriented networks and data mining to expand applicability to the dynamic growing field of information technology; automated detection fusion that adapts decision thresholds based on the variable performance of multiple sensors; an introduction of particle filtering, which provides a look into the background theories that gave rise to this important direction in target tracking; random set theory, which has been completely rewritten to provide an extensive development of this generalization of Bayesian and non-Bayesian approaches; situation and impact assessment theory and concepts, which have been significantly extended to build on research directions that have been previously glossed over; new techniques in visualization that have been developed; and finally, a new chapter in commercial off-the-shelf (COTS) software tools, which provides the reader with a wealth of fusion research tools and techniques. Additionally, areas of application have expanded to include electromagnetic systems, chemical and biological sensors, and army command and combat identification techniques. Finally, an appendix provides links to current web sites related to multisensor fusion.

This book follows the same theme of the first edition, dividing the material into four topical themes: Introduction to Multisensor Data Fusion provides the latest fundamental concepts and techniques; Advanced Tracking and Association Methods builds on the fundamentals to provide important research directions of fusion processing; Automated Reasoning and Visualization for Situation and Threat Refinement pushes the state-of-the-art concepts of fusion to situational behavior modeling, cognitive methods for human–machine interaction, and software sources for the reader to build on current technologies; and finally, Sample Applications demonstrate that the field of fusion research is no longer an emerging field, but a mature, strong contributor to information processing and understanding.