| Although binocular stereo vision is actively used in the field of intelligent transportation at present, it rarely used in nautical field. Officer’s perception for the Ship surroundings directly affects the safety of the ship, especially at night and in the conditions of poor visibility. It is more necessary to get target information by effective technology.Regard to deficiency of Radar and AIS for maritime dim targets detecting, the paper proposes binocular stereo vision, which includes image acquisition, camera calibration, feature extraction, stereo matching and 3D restructuring, applied to locating maritime targets and take it as supplement for Radar and AIS to assist safe navigation of the ship. The paper makes some theoretical and experimental researches on camera calibration, stereo matching and 3D restructuring. Regard to shipborne videos, the method of maritime targets binocular vision positioning is proposed. Firstly, combined with features of marine environment and existing calibration methods, the corresponding method of shipborne camera calibration is designed. Intrinsic parameters of the camera is calibrated by plane template method, and extrinsic parameters of the camera is calibrated by solving essential matrix. Secondly, stereo matching algorithm is designed by analyzing characteristics of maritime targets in the image. The matching algorithm widely reducing searching area of targets matching by epipolar constraint, ensures the high real-time performance with good precision. Finally, regard to the swaying problem during sailing, this paper propose the algorithm for correcting target position based on posture measurement by analyzing imaging model of binocular stereo vision. Using posture information collected by Micro-Electro-Mechanical inertial sensor to correct target position, which compensating the positioning error caused by swaying.This paper designs the intrinsic parameter calibration equipment and camera posture simulation platform, and does some experiments including the camera calibrate, target stereo matching and position correction based on posture information. Verified the method proposed in this paper by doing experiment on YuKun, a teaching practice ship of Dalian Maritime University. |