摘要
三维声呐点云数据的滤波效果直接影响点云重建的精度。针对测深数据滤波方法在三维声呐点云数据适用性不足的现状,开展了基于超体素划分的三维声呐点云数据滤波方法研究。利用超体素聚类划分方法构建点云块趋势面,给出顾及三维方向偏差的检测数据构建策略,并对三种滤波效果进行了定量分析。计算结果显示,Dixon滤波和Grubbs滤波的总误差分别为2.22%和3.04%,总误差较小且接近,Dixon滤波可以更好地保留水下结构物、底层地面点等位置的特征信息,Grubbs滤波对于近地噪点的过滤效果优于Dixon滤波。3σ滤波总误差为5.84%,较前两种滤波总误差较大,在水下结构物、近地点等区域易出现过滤波和欠滤波的问题,滤波效果较差。
The filtering effect of 3D sonar point cloud data directly impacts the accuracy of point cloud reconstruction.Considering the insufficient applicability of depth data filtering methods for 3D sonar point clouds,this study explores a filtering method for 3D sonar point cloud data based on supervoxel segmentation.A point cloud block trend surface is constructed using a supervoxel clustering segmentation method,and a strategy for detecting data that accounts for deviations in the three-dimensional direction is proposed.A quantitative analysis of the filtering effects of three methods is conducted.The results show that the total errors of Dixon filtering and Grubbs filtering are 2.22%and 3.04%,respectively,both of which are small and close in value.Dixon filtering better preserves the characteristic information of features such as underwater structures and bottom ground points,while Grubbs filtering is more effective at filtering near-ground noise points.The total error of 3σfiltering is 5.84%,which is larger than that of the previous two methods,leading to issues of over-filtering and under-filtering in areas such as underwater structures and near-ground points,resulting in poor filtering performance.
作者
甘淏柽
张帆
贺正军
孙爱国
熊荣军
吴云龙
GAN Haocheng;ZHANG Fan;HE Zhengjun;SUN Aiguo;XIONG Rongjun;WU Yunlong(School of Geography and Information Engineering,China University of Geosciences,Wuhan 430074,China;Key Laboratory of Geological Survey and Evaluation of Ministry of Education,China University of Geosciences,Wuhan 430074,China;Changjiang Waterway Institute of Planning,Design&Research,Wuhan 430040,China)
出处
《城市勘测》
2025年第1期51-58,共8页
Urban Geotechnical Investigation & Surveying
基金
国家自然科学基金(42274111)。
关键词
三维声呐点云数据
超体素
滤波方法比较
3D sonar point cloud data
supervoxel
comparison of filtering methods
