Multi-Sensor Kinematic and Attribute Tracking Using a Bayesian Belief Network Framework
Abstract
Situational awareness plays a major role in many military and civilian operations. Apart from the identity, type, location, and dynamics of targets of interest, the situational picture may also provide other target information, such as weapons state, fuel status, and intent. Many legacy systems incorporate an automatic tracking capability with identification, situational assessment, and decision making being left to the operators. The automation of many of these functions is the focus of much research and development. A necessary prerequisite for updating the state of a target is the correct association of measurements or other information to the track. The ability of Bayesian belief networks (BBNs) to model the uncertain relationships between continuous and discrete variables make them excellent candidates for incorporating both kinematic and attribute information in the association process. A BBN model for a single scan data association problem is presented and used to develop a global nearest neighbors solution using both kinematic and attribute information. Monte Carlo simulations demonstrate the benefit of using attribute information in the association process. Sixteen briefing charts summarize the presentation. (5 figures, 13 refs.)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2004
- Accession Number
- ADA428709
Entities
People
- Mark L. Krieg
Organizations
- Defence Science and Technology Group