Low-level multi-INT sensor fusion using entropic measures of dependence

An information-theoretic method of low-level multi-INT sensor fusion is presented, the end product of which is the entropic map, i.e. a collection of Gaussian clusters of information relevant to a given target signature formed over a geographical basis. The method is designed to be computationally efficient with minimal side-information. To that end, an unbiased estimate of information from finite data is derived along with a data-dependent, information-optimal measurement partition. A method for the determination of the information-optimal sensor suite is given for a possibly geographically dependent target signature. Finally, it is shown that a multi-relational entropic measure of dependence can be superior to suboptimal error-based techniques of estimation of multiple sensor measurements of a real process.