Context-sensitive data fusion using structural equation modeling

Abstract

Context is used in data fusion - and in inferencing in general - to provide expectations and to constrain processing. It also is used to infer or refine desired information (“problem variables”) on the basis of other available information (exogenous ”context variables”). Context is used in refining data alignment and association as well as in target and situation state estimation. Relevant contexts are often not self-evident, but must be discovered or selected as a means to problem-solving. Therefore, context exploitation involves an integration of data fusion with planning and control functions. Structural Equation Modeling techniques are used to evaluate dependencies between latent and observed problem and context variables. Discovering and selecting useful context variables is an abductive data fusion/ management problem that can be characterized in a utility/ uncertainty framework. An adaptive evidence-accrual inference method - originally developed for Scene Understanding - is presented, whereby context variables are selected on the basis of (a) their utility in refining explicit problem variables, (b) the probability of evaluating these variables to within a given accuracy, given candidate system actions (data collection, mining or processing), and (c) the cost of such actions.