Performance analysis for ground-based target orientation estimation: FLIR/LADAR sensor fusion

Recognizing 3-D objects from imaging sensors has received considerable attention in the last few years. Target recognition inherently depends on target pose estimation, because target signatures vary greatly with pose even for a single target/sensor combination. This paper addresses pose estimation for ground-based targets viewed with a combination of active and passive imagers, specifically a laser radar (LADAR) range imager and a forward-looking infrared (FLIR) thermal imager. The objective is to develop Cramer-Rao type bounds for the mean-squared error which explicitly reveal the roles of sensor and scenario parameters, and permit quantitative assessment of the benefits of sensor fusion. These analytical results are compared with simulation results obtained using the Hilbert-Schmidt norm as the performance measure.