基于局部异常因子算法的三维声纳单帧重建研究

海洋环境的复杂性会使三维声纳获取到的点云中存在异常点，从而影响点云的后期处理和可视化。将获取到的声纳数据依次进行三维波束成像、最大值滤波和坐标系转换，进而得到原始点云。提出将局部异常因子(LOF)算法应用到点云去噪、剔除干扰的异常点，并用一种改进三角网生长方法进行单帧重建。改进三角网生长法简化了三角网生成中关键的“第三点”搜寻过程，通过实际数据处理分析得知：相比于传统阈值去噪方式，基于LOF算法的去噪方式能够在保持目标真实轮廓的前提下有效剔除非目标点；经过所提去噪方式的声纳数据能够得到更贴合实际目标轮廓的三维重建结果。

Research on the 3D Sonar Single-frame Reconstruction Based on Local Outlier Factor Algorithm

3D sonar always acquires the data point with outliers in the complex marine environment, which affects the post processing and visualization of point cloud. 3D beamforming imaging, maximum filtering and coordinate transformation are made for the sonar data to obtain the original point cloud. The local outlier factor (LOF) algorithm is used to eliminate the abnormal points, and a modified triangulation growing method is used for single-frame reconstruction. In the modified triangulation growing method, the crucial process of ‘the third point’ finding in triangular mesh generation is simplified. A denoising method based on LOF algorithm can be used to eliminate the points not belonging to the target under the premise in keeping the real contour of target compared with the traditional threshold denoising method. And the 3D reconstruction result of more realistic target contour can be achieved from the sonar data obtained by the proposed denoising method. Key