(Invited) A Novel Self-Localization Method for USV Based on GPS and Visual-Map

Abstract

With the vigorous development of computer technology, information technology and intelligent transportation, their combination in the maritime field is becoming more and more extensive. At present, unmanned surface vehicle (USV) has been widely applied in the water transportation environment, especially for dangerous and time-consuming tasks. However, the intellectualization of USV is still far from actually being “unmanned”. Among them, USV uses its own sensors to perceive the surrounding environment and obtain high-precision position estimation is a prerequisite for its intelligence. In response to the demand, a novel two-step approach for USV self-localization is proposed in this paper. In the first step, the coarse-to-fine strategy is used to obtain the nearest node. Firstly, coarse image node is got by matching the current GPS information and the DGPS information of the visual-map. Then, the fine image node is obtained by the multi-vector fusion algorithm. In the second step, the pose of the current image is gained with respect to the nearest node by the epipolar geometry constraint, so as to achieve high-precision self-localization for USV. This method only needs an ordinary camera and GPS module in positioning, which can reduce the cost of high-precision positioning to the greatest extent. Pool experiments show that this method can achieve centimeter-level positioning accuracy.