Scene matching based EKF-SLAM visual navigation

For autonomous visual navigation of small UAV (SUAV), we proposed visual SLAM (Simultaneous Localization and Mapping) algorithms based on Extended Kalman Filtering (EKF) in unstructured natural environment. In this paper, a scene matching method with weighted Hausdorff distance was introduced firstly for waypoints accurate abstraction. On this foundation, the small UAV´s nonlinear state model was analyzed to establish nonlinear relationship model between the measurement and the waypoints, and then on to predict and estimate the state of the model, deal with data association and extend the state for new waypoints. Through the EKF-SLAM algorithm cycle prediction and estimation, our algorithm was realized to locate the small U AV accurately by visual navigation. Finally, by using waypoints abstract from scene matching navigation method, our simulation results show that the proposed algorithm could effectively reduce the estimation error of navigation system, simultaneously, provide theoretical supports for application of autonomous navigation.