【作者】 赵亮；

【导师】 程兰； 张飞虎；

【作者基本信息】 太原理工大学 ， 控制工程（专业学位）， 2025， 硕士

【摘要】 海洋环境勘探是我国建设海洋强国战略的重要技术支撑,而水下环境感知是进行海洋环境勘探的前提。多波束测深声纳(Multibeam Echo Sounder,MBES)是进行水下环境感知的常用传感器,但该设备只能得到垂直于前进方向的测深线数据,不能得到完整的水下地图,如完整的水下点云地图,因此无法提供水下的全局地图信息。鉴于此,本文根据多波束测深声纳采集的测深数据,基于船舶位姿估计方法建立水下局部点云地图,并在此基础上研究水下局部点云地图配准方法,建立水下全局点云地图。解决了由于海洋环境复杂以及单次数据采集受限无法获得完整水下点云地图的问题。主要工作如下:(1)针对根据多波束测深声纳采集的信息无法直接获取水下局部点云地图的问题,本文研究了基于船舶位姿估计的水下局部点云地图构建方法。该方法首先在多波束测深声纳采集的单帧点云数据基础上,利用误差状态卡尔曼滤波(ESKF)算法融合惯性测量单元(Inertial Measurement Unit,IMU)和全球导航卫星系统(Global Navigation Satellite System,GNSS)数据,估计船舶位姿信息。然后利用船舶位姿信息将单帧点云数据从局部坐标系转换到全局坐标系下,并进行拼接,进而得到水下局部点云地图。在公开数据集上的实验结果表明,所研究的方法在绝对位姿误差(APE)和相对位姿误差(RPE)指标下优于对比方法,验证了所研究方法在水下局部点云地图构建中的有效性。(2)针对单次数据采集建立的水下局部点云地图无法反映水下全局信息,且受设备限制以及环境干扰导致点云地图存在高异常值、低分辨率以及显著特征缺乏的问题,本文研究了鲁棒的水下局部点云地图配准方法。该方法在前期建立水下局部点云地图的基础上,首先使用提出的异常值剔除方法在保留点云地图细节特征的同时去除异常值。然后,该方法首次将四自由度的分支定界(Branch and Bound,BnB)算法应用在水下点云地图配准任务中,较好地匹配了水下局部点云地图,从而实现水下全局点云地图的构建。实验结果表明,所研究的方法在旋转和平移误差评价指标下与对比方法相比具有显著优势,同时运行时间满足实际项目需求。(3)为了更好地验证所研究的方法在真实场景中的实用性,本文在陕西咸阳醴泉湖进行了水下局部点云地图构建和水下局部点云地图配准实验。实验结果表明本文研究的方法在实际应用中可以构建连续一致的水下局部点云地图,同时可以鲁棒精确地实现水下局部点云地图到水下全局点云地图的构建。此外,针对水下场景建图需要繁琐命令行操作的问题,本文基于PyQt5设计并实现了一个水下环境感知系统。通过提供简洁直观的用户交互界面,实现了水下场景建图的核心功能,提高了建图的可操作性和可观性。更多还原

【Abstract】 Marine environmental exploration serves as a crucial technological pillar for China’s strategy of building a strong maritime nation,and underwater environmental perception is the prerequisite for conducting marine environmental exploration.The Multibeam Echo Sounder(MBES)is a commonly used sensor for underwater environmental perception.How-ever,this device can only obtain bathymetric line data perpendicular to the direction of move-ment and cannot produce a complete underwater map,such as a full underwater point cloud map.As a result,it is unable to provide comprehensive global map information of the under-water environment.In light of this,this thesis utilizes the bathymetric data collected by the Multibeam Echo Sounder to construct a local underwater point cloud map based on vessel pose estimation methods.Furthermore,it investigates registration methods for local underwater point cloud maps to establish a global underwater point cloud map.This resolves the issue of being unable to obtain a complete underwater point cloud map due to the complexity of the marine environment and the limitations of single data acquisitions.The main work is as follows:(1)To address the issue that the information collected by the Multibeam Echo Sounder can-not directly provide a local underwater point cloud map,this thesis investigates a method for constructing local underwater point cloud maps based on vessel pose estimation.This method builds on the single-frame point cloud data collected by the Multibeam Echo Sounder and uti-lizes the Error State Kalman Filter(ESKF)algorithm to fuse data from the Inertial Measurement Unit(IMU)and the Global Navigation Satellite System(GNSS),thereby estimating the vessel’s pose information.Finally,the vessel’s pose information is used to transform the single-frame point cloud data from the local to the global coordinate system,followed by stitching,to obtain the local underwater point cloud map.Experimental results on public datasets demonstrate that the proposed method outperforms comparative methods in terms of Absolute Pose Error(APE)and Relative Pose Error(RPE)metrics,validating the effectiveness of the proposed approach in constructing local underwater point cloud maps.(2)To address the issue that local underwater point cloud maps constructed from single data acquisitions cannot reflect global underwater information,and to tackle problems such as high outliers,low resolution,and a lack of distinctive features caused by equipment limitations and environmental interference,this thesis investigates robust registration methods for local underwater point cloud maps.Building on the previously established local underwater point cloud maps,this method first utilizes the outlier removal technique proposed in this thesis to eliminate outliers while preserving the detailed features of the point cloud maps.Then,for the first time,this method applies a four-degree-of-freedom Branch and Bound(BnB)algorithm to the task of underwater point cloud map registration,effectively aligning local underwater point cloud maps and thereby enabling the construction of a global underwater point cloud map.Experimental results demonstrate that the proposed method exhibits significant advantages over comparative methods in terms of rotation and translation error metrics,while its runtime meets the requirements of practical applications.(3)To better validate the practicality of the proposed method in real-world scenarios,this thesis conducted experiments on the construction and registration of local underwater point cloud maps at Liquan Lake in Xianyang,Shaanxi Province.The experimental results indicate that the method proposed in this thesis can construct continuous and consistent local underwater point cloud maps in practical applications,while also achieving robust and accurate registration from local to global underwater point cloud maps.Additionally,to address the issue of cumber-some command-line operations required for underwater scene mapping,this thesis designed and implemented an underwater environmental perception system based on PyQt5.By providing a simple and intuitive user interface,the system implements core functionalities for underwater scene mapping,enhancing both its operability and observability.更多还原