Neural Radiance Fields(NeRF)has significant application prospects in 3D reconstruction and protection of cultural relics,but for large-scale 3D reconstruction of immovable cultural relics,the existing NeRF technology faces the challenges of high computational resource consumption and the inability of the model to locally repair.For this reason,this study proposes to adopt octree to improve the scene partitioning of the voxel sampling step of NeRF,to realize the parallel lightweight 3D reconstruction of NeRF,and to make its model support local targeted repair,which greatly improves the practicability of NeRF technology in the field of large-scale 3D reconstruction of immovable cultural relics.
关键词
文物保护/三维重建/神经辐射场/八叉树
Key words
cultural relics preservation/3D reconstruction/neural radiance Fields(NeRF)/octree
维普
万方数据