Enhanced Image Tracking: Analysis of Two Acceleration Models in Tracking Multiple Hot-Spot Images

摘要

Two extended Kalman filter algorithms that estimate target position, velocity, and acceleration, as well as atmospheric jitter are developed for use within a laser weapon system. Digital signal processing techniques are employed on data obtained from a forward looking infrared (FLIR) sensor in order to identify the underlying shape of multiple hot-spot targets. No a priori information is assumed about such images. The identified shape is used in the measurement model portion of the extended Kalman filters in order to estimate target position offsets from the center of the sensor field of view. The two dynamics models incorporated within the filters are 1) a first order Gauss-Markov acceleration model and 2) a constant turn-rate acceleration model. Performance of these two filters are compared in tracking scenarios involving constant G and constant roll-rate maneuvers. Extensive consideration is given to simulating realistic multiple hot-spot images on the FLIR image plane. Performance of a previously developed adaptive filter is shown to be seriously degraded when faced with multiple hot-spot images since it assumes a priori information about the target image.