Image-enhanced estimation methods

The performance of an image-enhanced estimator is contrasted with that of the extended Kalman filter (EKF). A scenario in which a planar agile target moves with intermittent maneuvers is studied. The performance comparison clearly indicates that image enhanced estimation methods are worthy of consideration in applications involving agile targets. A dual-path estimation architecture in which one path infers the likelihood of maneuver from image data is considered. These maneuver likelihoods are used to adapt the filter gains to changing conditions. Although the image-based estimator employs what appears to be an orthodox algorithm, it is less susceptible to delays in detecting a maneuver. In this architecture, the image path uses observations of target shape to change the time constants in the range-bearing path. In effect, one path modulates the other, and the tracking system is able to locate the target and discern changes in its motion pattern, so that it follows target motion more accurately. The results illustrate both the potential and the limitation of image augmentation