Random Sets in Information Fusion an Overview

Abstract

“Information fusion” refers to a range of military applications requiring the effective pooling of data concerning multiple targets derived from a broad range of evidence types generated by diverse sensors/sources. Methodologically speaking, information fusion has two major aspects: multisource, multitarget estimation and inference using ambiguous observations (expert systems theory). In recent years some researchers have begun to investigate random set theory as a foundation for information fusion. This paper offers a brief history of the application of random sets to information fusion, especially the work of Mori et. al., Washburn, Goodman, and Nguyen. It also summarizes the author’s recent work suggesting that random set theory provides a systematic foundation for both multisource, multitarget estimation and expert-systems theory. The basic tool is a statistical theory of random finite sets which directly generalizes standard single-sensor, single-target statistics: density functions, a theory of differential and integral calculus for set functions, etc.