Indoor scene reconstruction model processing method and device

Abstract

The embodiment of the invention discloses an indoor scene reconstruction model processing method and device, and the method comprises the steps: obtaining a surface mesh model obtained through Poisson reconstruction based on the point cloud of an indoor scene; determining a to-be-deformed target area and an associated patch of the target area according to the density values of the vertexes in the plurality of patches in the surface mesh model; and performing deformation processing on the target area according to the vertex information of the target area and the vertex position of the associated patch. According to the invention, on the premise of ensuring a complete watertight structure, 'bump' contraction can be realized.

Description

Technical field [0001] The present disclosure relates to the field of radar and indoor scene reconstruction technologies, in particular a processing method and apparatus for reconstruction model of indoor scene. Background technique [0002] When the application of the laser scanner is scanned, it is affected by the structure and material of the indoor scene itself. It is difficult to ensure that the scan is a complete point cloud model of the entire scene. There are two main situations: one is between the objects between the rooms. Block, for example, the scan of the bed is blocked to the other side of the bed and the ground, such as a higher cabinet, can block its own top and wall; the other is the transparency of the glass on the window, the scan is not Point cloud. [0003] Poisson Surface Reconstruction, PSR can get a water-densely model, that is, a non-porous model, but the surface grid model of the point cloud deletion area is from the lower depth of eight The fork tree no...

AI Technical Summary

Problems solved by technology

[0002] When using a laser scanner to scan and reconstruct the interior of a house, due to the influence of the structure and material of the indoor scene itself, it is difficult to ensure that the complete point cloud model of the entire scene can be obtained by scanning. There are two main situations: one is the distance between objects in the room Covering, for example, scanning on one side of the bed will cover the bed and floor on the other side, for example, a taller cabinet will cover its own top and wall; the other is the high transparency of the glass on the window, which cannot be scanned point cloud

[0003] Although Poisson Surface Reconstruction (PSR) using an incomplete point cloud model can obtain a watertight complete model, that is, a model without holes, the surface mesh model of the missing point cloud is composed of a lower depth eight Intuitively, the base function obtained by the fork tree node will be a smooth "bulge", which greatly affects the visual effect of the model.

[0004] In related technologies, when dealing with the "bulge" area formed by low-density point clouds, the patch in this area is directly deleted, but after the patch is deleted, the model is not completely watertight, and there will be damaged holes, which also greatly affect the vision. Effect

Images