Research And Implementation Of A Multi-faceted Joint Sensing System For USV

In the complex and changing marine environment,with the development of sensor technology and the improvement of equipment intelligence,unmanned surface vehicles are playing an increasingly important role.Real-time and accurate environment perception is a prerequisite for unmanned surface vehicles to complete autonomous navigation.To address the problem of joint situational awareness of unmanned surface vehicles with multiple information in unknown marine environments,this thesis takes the “Lanxin” USV of Dalian Maritime University as the research object and experimental platform and researches the joint sensing technology based on the navigation radar and 3D Li DAR onboard the unmanned vehicle.The main research contents are as follows.1.For a class of unmanned surface vehicles in a highly dynamic and complex navigation environment navigation radar targets accurate tracking problem,considering the high dynamic characteristics of the unmanned surface vehicle make the observation noise of navigation radar is the case of change,a navigation radar target tracking method based on improved adaptive Kalman filter for the unmanned surface vehicle is proposed.The method first generates radar images by the boat navigation radar and uses image processing techniques to extract the target information in the images.Then the unmanned surface vehicle attitude sensor is used to collect real-time bow angle rate information,and the observation noise covariance is updated adaptively according to the attitude change,so that the observation noise covariance in the Kalman filter algorithm is closer to the actual value,thus improving the tracking accuracy of the target.2.For the problem of point cloud map generation in an unknown water surface environment,this thesis proposes a 3D point cloud map construction method for the water surface environment.The method first preprocesses the acquired real-time point cloud information,then takes into account the point cloud characteristics in the water surface environment,and uses a series of methods such as the region grow method to segment and filter the invalid point clouds on the water surface,to improve the reliability of the constructed point cloud map of the water surface environment.The LOAM algorithm with better real-time accuracy is used to build the environment map,which provides a guarantee for the unmanned surface vehicle to explore unknown areas in the offshore environment.3.To address the problem that unmanned surface vehicles use a single sensor to obtain limited sensory information,in order to obtain accurate and all-around environmental sensing information,this thesis proposes an image fusion method based on boat-based navigation radar and 3D Li DAR by combining the advantages of navigation radar and Li DAR.The method converts 3D point cloud information into a 2D raster map through the rasterization process,and some 3D point cloud attributes are retained in the rasterization process.In the raster attribute determination method,the height information retained by each raster in the raster map is judged to assist in determining the passable area.Image processing techniques are used to normalize the multi-scale images,and a specific pixel fusion scheme is used to fuse the navigation radar images with the raster map.The final fused image is obtained,making the obstacle information of the surrounding environment acquired by the unmanned surface vehicle more accurate and comprehensive.4.In order to verify the feasibility and effectiveness of the above method,this thesis has been verified through real ship experiments at sea.Using the “Lanxin” unmanned surface vehicle as to the experimental platform,the real ship experiments were conducted in the sea within three kilometers of Linghai Campus of Dalian Maritime University to complete the analysis and verification of the above method.