An image tracking algorithm using reduced sufficient statistics
Craig S. Agate, R.A. Iltis · 2002
An algorithm is presented for tracking an extended target from imaging array data, which is based on the reduced sufficient statistic (RSS) method of Kulhavy (1990). The method has wide application as it can be used in situations where the image does not have a closed form functional representation. This is accomplished by storing a finite set of templates which are averaged images of the target in various positions and orientations. The RSS method recursively calculates a set of sufficient statistics for a density which approximates the a posteriori density of the target parameter vector. While the parameter vector includes the orientation of the target as components, it only uses the orientation to find a good match between the received image and an expected image of the target given different positions and orientations. A sensor is individually capable of estimating the distance of the target, since the templates are functions of the target range. This manifests itself in the scale of the image due to the perspective transform in the camera model.