Feature-aided global nearest pattern matching with non-Gaussian feature measurement errors

System-level discrimination performance for missile defense relies on how well data can be associated between participating sensors. Under the existing architecture, there may be a handover of tracks between two sensors in which tracks formed by one sensor are passed to another sensor to improve knowledge of the targets. The global nearest pattern matching (GNPM) problem is a mathematical programming formulation that has proven to be successful at correctly correlating tracks based solely on kinematic data from two sensors, while simultaneously removing inter-sensor bias and accounting for false tracks and missed detections. Despite this success, there is continued interest to improve correlation performance by exploiting feature data collected on targets. This paper addresses this issue by extending the GNPM formulation to account for feature observations whose measurement errors follow an arbitrary distribution. This is accomplished by augmenting the GNPM likelihood function to include a term representing the incremental likelihood of track-to-track assignments based solely on feature observations. Computational results are presented to illustrate the success of this approach.