SCIENCE@home

Towards a Graph-Based Approach for Mesh Healing for Blocky Objects with Self-Similarities

Matthew Jones,Nicole Hippolite,R. Patel,Sebastian N. Peters,Priyankit Singh,Chia-Yen Chen,B. Wünsche

Published 2018 in Image and Vision Computing New Zealand

ABSTRACT

3D digital models are used in a wide range of applications such as computer games, virtual reality, engineering and urban design. A common method to create 3D models is to create a point cloud using laser scanners, structured lighting sensors, or image-based modelling techniques, and from that construct a 3D mesh. The resulting meshes often exhibit unsatisfactory and erroneous mesh regions, e.g. due to inaccessible parts, reflective surfaces, and occlusion. In this paper we present a new approach for correcting erroneous mesh regions in blocky objects, such as buildings, using self-similarities. Our system uses a novel graph-based search technique where each node is a plane fitted to the model using RANSAC and edges are made between adjacent planes. This graph is used to find defects in a mesh by analysing the planes and neighbourhoods of all nodes in the point cloud. The system recognizes a hole in the mesh and fills it by copying matching patches from the same mesh. Regions of the mesh that must be replaced are found by a voting system where dissimilar nodes with a similar neighbourhood vote against each other. Our solution produces visually accurate models of blocky objects with holes and/or missing corners and edges. The approach currently only works for simple models and can only “heal” individual or connected faces, rather than complex holes. However, in contrast to existing solutions it works with noisy point clouds and does not require the construction of a polygonal modal.

PUBLICATION RECORD

CITATION MAP

EXTRACTION MAP

CLAIMS

  • No claims are published for this paper.

CONCEPTS

  • No concepts are published for this paper.

REFERENCES

Showing 1-24 of 24 references · Page 1 of 1

CITED BY

  • No citing papers are available for this paper.

Showing 0-0 of 0 citing papers · Page 1 of 1