Belief fusion, pignistic probabilities, and information content in fusing tracking attributes

Abstract

In the design of information fusion systems, the reduction of computational complexity is a key design parameter for real-time implementations. One way to simplify the computations is to decompose the system into subsystems of noncorrelated informational components, such as a qualitative informational component, a quantitative informational component, and a complement informational component. A probability information content (PIC) variable assigns an information content value to any set of system or sub-system probability distributions. The PIC variable is the normalized entropy computed from the probability distribution. This article derives a PIC variable for a subsystem represented by the complement probabilities. This article also derives a relationship between the PIC variable of sub-system components and the system informational PIC variable. A series of pignistic probability transforms are presented that estimate the probability for any belief data set. The generalized belief fusion method of combining independent multi-source beliefs is presented.