A fuzzy-controlled Kalman filter applied to stereo-visual tracking schemes

Abstract

In this paper, the authors propose a fuzzy-controlled Kalman filtering scheme applied to stereo visual tracking. Two control levels have been designed: first, a fuzzy methodology allows the filter to fine tune to actual conditions by estimating the plant noise covariance matrix in every time instant. Second, a fuzzy control stage based on a fuzzy feedback system is used to reinitiate the filter in those cases in which a track is lost. The principal coordinates paradigm has been used for the-filter, but the methodology here proposed carries over easily to other more popular Kalman schemes. (C) 2002 Elsevier Science B.V. All rights reserved.