Assessment of a Bayesian Approach to Recognising Relocatable Targets

摘要

This paper considers an automatic target recognition concept in which a long-range targeting sensor is used to aid a radar seeker-equipped weapon operating in an area containing high-value relocatable targets. The weapon seeker is designed to engage the high-value targets, while minimizing collateral damage. Previous work proposed a Bayesian approach that enables the weapon seeker to exploit the targeting information before making its final decision. The approach matches the scenes in the seeker domain with those from the targeting sensor, while taking into account uncertainty and data latency. The proposed solution utilizes a Bayesian technique known as particle filtering, and had previously only been applied to a synthetic example. This paper summarizes the approach, and presents results from an assessment using scenarios derived from an airborne data set containing short-range Doppler beam sharpened imagery. The issue of differing resolutions for the two sensors is addressed by super-resolution techniques, which are also assessed.