视觉与激光联合的三维网格重建

Joint mesh reconstruction from visual and LiDAR

针对激光点云密度大、精度高,但构网结果噪声多且存在冗余,具有丰富语义和纹理信息的多视影像由于遮挡导致重建点云精度低,分布不均且出现空洞的问题,提出一种基于改进加权图割重构的联合建模方法,根据点云空间位置关系赋值可视信息,通过点云间距及深度阈值筛选四面体网格顶点,最后判断顶点属性更改权重,提升了对重建曲面细节的刻画及点云稀疏处的完整性重建。实验表明,所提方法可以减少激光点云构网冗余、填补视觉点云漏洞,实现对目标物体精细完整建模。

The primary methods for 3D mesh reconstruction include LiDAR point cloud mesh reconstruction and multi-view images reconstruction.LiDAR point clouds offer high density and accuracy,but the resulting mesh construction often contains significant noise and redundancy.Multi-view images provide rich semantic and texture information,but suffer from low reconstruction point cloud accuracy,unevenly distribution and the presence of voids due to occlusion.To address this issue,this paper proposes a joint modeling method based on improved weighted graph cut reconstruction.The visual information is assigned based on the spatial relationships of point cloud positions,tetrahedral mesh vertices are filtered based on point could distances and depth thresholds.Finally,weights are modified to enhance the portrayal of surface details in the reconstructed surface and maintain the integrity in sparse areas of the point cloud.The experimental results demonstrate that this method successfully reduces redundancy in the LiDAR point cloud mesh construction,fills gaps in the visual point cloud,and achieves details and complete modeling of the target object.