Appearance learning by adaptive Kalman filters for FLIR tracking
Vijay Venkataraman, Guoliang Fan, Xin Fan, Joseph Havlicek · 2009 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops · 2009
This paper addresses the challenging issue of target tracking and appearance learning in Forward Looking Infrared (FLIR) sequences. Tracking and appearance learning are formulated as a joint state estimation problem with two parallel inference processes. Specifically, a new adaptive Kalman filter is proposed to learn histogram-based target appearances. A particle filter is used to estimate the target position and size, where the learned appearance plays an important role. Our appearance learning algorithm is compared against two existing methods and experiments on the AMCOM FLIR dataset validate its effectiveness.