Multi-resolution voxel meshing

Abstract

Various implementations disclosed herein generate a mesh representing the surfaces in a physical environment. The mesh is generated using multi-resolution voxels based on detected depth information, e.g., from a depth camera. The techniques may use multiple hash tables to store the multi-resolution voxel data. For example, the hash tables may store each voxel's 3D position and a truncated signed distance field (TSDF) value corresponding to each voxels' distance to a nearest surface. Each of the multiple hash tables may include data corresponding to a different level of resolution and those resolutions may depend upon distance / noise or other factors. For example, voxels close to a depth camera may have a finer resolution and smaller size compared to voxels that are further from the depth camera. Techniques disclosed herein may involve using a meshing algorithm that combines multi-resolution voxel information stored in multiple hash tables to generate a single mesh.

Description

technical field [0001] The present disclosure relates generally to generating three-dimensional geometric representations of a physical environment, and in particular to systems, methods, and devices for generating geometric representations based on depth information detected in a physical environment. Background technique [0002] The physical environment has been modeled (eg, reconstructed) by generating a three-dimensional (3D) mesh. These meshes represent 3D surface points and other surface features of floors, walls, and other objects of the physical environment. Such reconstructions may be generated based on images of the physical environment and depth measurements (eg, using RGB cameras and depth sensors). The reconstruction technique may use voxels to generate a mesh to provide the reconstruction. As used herein, a voxel refers to a volumetric pixel. Existing reconstruction techniques use fixed-sized voxels spaced in a regularly spaced grid in 3D space without gaps...

AI Technical Summary

Problems solved by technology

While such techniques use relatively large voxels, these techniques may not adequately represent detailed features of the physical environment

In contrast, when such techniques use relatively small voxels, these techniques may produce inaccurate reconstructions due to insufficient noise filtering and may require undesired resource usage, e.g. requiring processing that cannot be performed in real time or undesired or difficult to obtain memory storage levels

Therefore, existing reconstruction techniques may not provide sufficiently accurate and effective reconstructions of the physical environment

Images