Mathematics of data fusion (Q1366688): Difference between revisions

Data fusion is the name which has been given to variety of interrelated expert-system problems which have arisen primarily in military applications. However, in a more general sense data fusion refers to a broad range of problems which require the combination of quite diverse types of information, provided by a wide variety of sensors, in order to make decisions and initiate actions. This information is typically uncertain in at least one of the many possible senses of the word probabilistic, imprecise, fuzzy, contingent, and so on. This problem domain includes not only military applications but also statistics, decision theory, decision-support systems expert systems, engineering and so on. At the current time, a sometimes bewildering variety of techniques have and are being applied to data fusion problems: statistical estimation and decision theory, Bayesian nets, fuzzy logic, Dempster-Shafer evidence theory, rule-based approaches neural nets, genetic algorithms, wavalet transform theory, and so on. The current theoretical developments in the field of data fusion have been moving towards: development of representation and management calculi for reasoning in the presence of uncertainty, development of spatial and contextual reasoning processes suitable for assessment of activity and intent on the basis of emporal and spatial behaviour, development of integrated smart sensors with soft-decision signal processing of use in systems that perform numeric and symbolic reasoning in uncertainty. In light of current advances in expert-system techniques, cognitive science, and so on, for the authors it seems desirable to incorporate into a single volume recent research work involving the representation and combination of disparate information types (statistical, linguistic, imprecise, contingent, etc.). In particular, their own researches in fuzzy logic, random set theory, and conditional event algebra have come to a sufficient degree of fruition that a book-length treatment seems warranted. Thus, this book is intended to be both an update on research progress in data fusion and an introduction to potentially powerful new techniques. As such, is consists of three parts: Introduction to data fusion and standard techniques (Part 1); The random set approach to data fusion (Part 2); and Use of conditional and relational event algebra in data fusion (Part 3). It is intended as a reference book for researchers and practioners in data fusion or expert system theory, or for graduate students as a text for a research seminar or a graduate-level course.