In region surveillance applications, sensors oftentimes accumulate an overwhelmingly large amount of data, making it infeasible to process all of the collected data in real-time. For example, a multi-channel synthetic aperture radar (SAR) flown on an airborne platform could receive on the order of 10 GBits of data per second. This data can be exploited in a number of ways (e.g., constructing a detected image, applying an ATR algorithm, or performing moving target processing) each of which requires significant computational resources. Given the enormous amount of data and the correspondingly large number of potential exploitation algorithms, there simply are not enough computational resources to process all of the data with all possible exploitation algorithms. The natural question then becomes one of how to most effectively utilize limited processing resources so as to facilitate real time exploitation of the collected data. This paper presents an information theoretic approach for processing action selection which is predicated on predicting the amount of information flow each potential processing action is expected to generate. The aim is to select those exploitation algorithms (and, in general, the physical region and algorithm parameter settings) that will be most useful in refining the underlying estimate of the surveillance region state. We show by simulation on a model problem that the information theoretic method is able to outperform other methods of processing selection.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below