A visual integrated navigation for precise position estimation

In this article, we propose a visual integrated navigation method which is composed of a relative position estimation and an absolute position estimation. The aim is to estimate precise position and overcome accumulative position errors for long-distance and long-time visual navigation. The relative position estimation measures relative displacement and position by a simple optical flow method, but its accumulative position errors would increase with time. The absolute position estimation aids the relative position estimation for correcting these accumulative errors timely. The absolute position estimation employs a Kalman filter based on image entropy, which can continuously correct accumulative errors according to the actual and predicted image entropy. As a result, the proposed navigation method can supply precise position information all along the navigational path, which has been proven by some experiments in the paper.