An active target localization with monocular vision

A novel information-driven trajectory planning algorithm for visual localization based on a moving camera is proposed in this paper. The trajectories of camera have a great influence on localization accuracy and efficiency. The camera estimates the target´s position by means of extended Kalman filter and optimizes its motion directions by maximizing the mutual information between the position estimate and the motion of camera. Compared with camera mobility driven by information minimizing and random, the proposed algorithm requires fewer movements for position estimation. The information-driven maneuvers can lead to quick convergence to good position estimate. Simulation experiment veritifies the effectiveness of this method.