基于法线微分的3维声呐点云自适应简化方法

Adaptive Simplification Method of 3D Sonar Point Cloud Based on Normal Differentiation

在不损失原始点云数据质量的前提下,大幅约简点云数据量是减少存储空间、降低后期计算强度的重要预处理步骤。针对这一需求,提出了针对水下3维声呐点云数据的自适应简化方法。首先,定义法线微分算子来识别点云中几何尺度的骤变,从而实现原始点云中边界部分点云和主体部分点云的分割。其次,对于点云的边界部分,应用移动最小二乘法来对边界点进行优化,降低噪点的影响,并保持其曲面的几何一致性;基于体素栅格结构,在边界上使用八叉树进行降采样,并在此基础上实施局部最远点采样,在实现均匀简化的同时保证已简化点云的边界部分具有各向同性,有效保留边界部分点云的几何特征信息。再次,对于点云的主体部分,为保持简化后点云整体的各向同性,使用体素中心采样法来减少数据量。然后,通过高斯滤波平滑点云表面,最后,整合简化后的边界点云和主体点云,得到简化结果。实验结果表明,提出的简化方法计算成本低、处理速度快,在与现行典型算法保持一致简化率的情况下,对水下点云数据的简化速度提高了约32%。另外,通过表面密度对比与几何失真分析,证明了提出方法对水下3维点云边界点及整体分布的优化作用。综上,此方法能提高水下作业目标探测效率,得到保留重要几何特征信息并具有各向同性的水下任务目标点云简化结果。

Reducing the number of points without sacrificing the quality of the original point cloud data is an essential preprocessing step for downsizing storage space and lowering calculation intensity.This study proposes an adaptive simplification method for underwater 3D sonar point cloud data.The normal differential operator is determined to identify sudden changes in the geometric scale of the point cloud,enabling segmentation of the boundary and primary part of the original point cloud.For the boundary part of the point cloud,the moving least squares method is applied to optimize the boundary points,reduce noise influence,and maintain the geometric consistency of the surface.The octree is used for downsampling on the boundary based on the point cloud's voxel grid structure.Depending on this structure,local farthest point sampling is implemented,achieving uniform simplification while ensuring the isotropy of the boundary part of the simplified point cloud and effectively retaining the geometric feature information of the boundary part of the point cloud.For the main part of the point cloud,the voxel center sampling method is employed to reduce the data and maintain the overall isotropy of the simplified point cloud,and the point cloud surface is smoothed through Gaussian filtering.Finally,the simplified boundary and main body are integrated to obtain the final simplified results.Experiments showed that the proposed simplification method has a low computational cost and fast processing speed,improving the simplification speed of underwater point cloud data by approximately 32%while maintaining the same simplification rate as current typical algorithms.In addition,the optimization of the algorithm on the boundary points and overall distribution of the underwater 3D point cloud is demonstrated through surface density comparison and geometric distortion analysis.The method enhances the detection efficiency of underwater operational targets and produces simplified underwater mission target point clouds that retain critical geometric feature information and remain isotropic.