Advanced Minehunting Sonar Data Fusion

We propose developing a data fusion system which takes a set of contact lists from a multi-sonar multi-look MCM system and generates a unified, fused, and geo-registered contact list. We will design the system to begin with a list of contacts with geographic locations and a set of detection features. The first step in our system will be to preprocess the contact locations for sensor registration and gridlock. At the same time, we propose a feedback loop: if the gridlock finds that two lists ought to have a contact in common but do not, we can go back to the raw sonar data and look for evidence of the contact. Further, we will maintain a history of gridlock solutions as functions of platform kinematic state and use this to generate gridlock even in the absence of significant contact list overlap. Once the contact locations have been locked to a common grid, we apply Metrons custom-developed eXtended Maximum A Posteriori (XMAP) Bayesian contact association algorithm which utilizes all contact features. Once contacts are associated we will perform Bayesian data fusion feature-by-feature. This fusion step also constitutes geo-registration as we also fuse location estimates from all available contact lists.