Finite set statistics with applications to data fusion

Abstract

Data fusion is concerned with the following general problem: Locate and identify objects of many different types on the basis of different kinds of evidence. The evidence is collected on an ongoing basis by many sensors having varying capabilities. Some evidence might be (i) probabilistic, e.g., estimated location with Gaussian distribution supplied by radar, (ii) incomplete probabilistic, e.g., partially specified location distribution supplied by sonar, (iii) linguistic, e.g., sighting by pilot (“sub is probably in region A but may also have been in region B”), (iv) conditional, e.g., if A is observed then B is observed with probability p(B∣A),etc.