Three-dimensional reconstruction method and device based on surface elements

Abstract

The invention discloses a surface element-based three-dimensional reconstruction method and device, and the method comprises the steps: carrying out the edge point detection and super pixel extraction based on a surface element method in combination with a depth image and a binocular color image of a to-be-reconstructed scene, and introducing regularization constraint to obtain a dense three-dimensional map. According to the method, generation of edge'holes' in a reconstructed image can be obviously reduced, the plane reconstruction smoothness and the three-dimensional reconstruction robustness are improved, and the reconstruction accuracy and stability of surface element-based three-dimensional reconstruction at the edge are greatly improved.

Description

technical field [0001] The invention relates to the technical field of surface element-based three-dimensional reconstruction methods. Background technique [0002] At present, 3D reconstruction can be divided into surface element-based, voxel-based and point cloud-based 3D reconstruction. Among them, 3D reconstruction based on point cloud is not suitable for reconstruction of large scene environments, and in order to increase its observability, it usually needs to be used in conjunction with image textures, which has high application requirements; 3D reconstruction based on voxels has a high role in reconstructing large scenes. However, voxels are not easily deformed, and it is difficult to correct the modeling accuracy in real time, so the advantages of closed-loop correction cannot be well utilized; 3D reconstruction based on surface elements has the advantages of variability and small storage space, and can be applied to large scenes and Small environment 3D reconstruct...

AI Technical Summary

Problems solved by technology

Among them, 3D reconstruction based on point cloud is not suitable for reconstruction of large scene environments, and in order to increase its observability, it usually needs to be used in conjunction with image textures, which has high application requirements; 3D reconstruction based on voxels has a high role in reconstructing large scenes. However, voxels are not easily deformed, and it is difficult to correct the modeling accuracy in real time, so the advantages of closed-loop correction cannot be well utilized; 3D reconstruction based on surface elements has the advantages of variability and small storage space, and can be applied to large scenes and Small environment 3D reconstruction, but the accuracy of surface element-based 3D reconstruction depends on the extraction accuracy of pixels and depth maps, as well as the positioning accuracy of the positioning system, which is prone to low reconstruction accuracy and "holes" at the edge

Images