Unscented Kalman filter for visual curve tracking

Visual contour tracking in complex background is a difficult task. The measurement model is often nonlinear due to clutter in images. Traditional visual tracker based on Kalman filter employs simple linear measurement model, and often collapses in tracking process. The paper presents a new contour tracker based on Unscented Kalman filter that is superior to extended Kalman filter both in theory and in many practical situations. The new tracker employs more accurate nonlinear measurement model, without the computation of Jacobian matrix. Thanks to a set of appropriately chosen sample points, multiple hypothesises can be maintained and the best observation can be made during one time step. The algorithm can obtain more exact estimate of the state of the system, and has the same order of complexity as that of Extend Kalman Filter. The experiments show that the new algorithm is superior to those based on Kalman filter.