Filter algorithm for visual tracking of maneuvering target

A filter algorithm is presented in this paper for visual tracking of a maneuvering target. Emphasis is given to find a solution for the degradation in relative position and orientation estimation, which is incurred by the measurement noise in the image coordinates of feature points. Superior to previous approaches that were limited to the assumption that the target motion is slow and smooth, this algorithm is implementable for a maneuvering target that acts in an unknown manner. First, by analyzing the effect of noise in 2-D images on the position and orientation estimation, linear measure equations based on the sequence of motion parameters are given. Then, two filter schemes are introduced respectively. The first filter uses maneuver detection technique, in which optimized detectors for fast and slow maneuver are deduced respectively, and limited memory filtering is adopted to update the filter. The second filter uses numerical differentiation technique, in which a fading factor is adaptively estimated to restrain the divergence caused by truncation errors of estimate model of numerical differentiation. Finally, generalized pseudo Bayes algorithm is employed to combine the two filters for a higher tracking precision. Simulation and experiment results illustrate the capacity of this algorithm.