Point cloud completion using extrusions

Oliver Kroemer, Heni Ben Amor, Marco Ewerton, Jan Peters

Research output: Chapter in Book/Report/Conference proceedingConference contribution

20 Scopus citations

Abstract

In this paper, we propose modelling objects using extrusion-based representations, which can be used to complete partial point clouds. These extrusion-based representations are particularly well-suited for modelling basic household objects that robots will often need to manipulate. In order to efficiently complete a partial point cloud, we first detect planar reflection symmetries. These symmetries are then used to determine initial candidates for extruded shapes in the point clouds. These candidate solutions are then used to locally search for a suitable set of parameters to complete the point cloud. The proposed method was tested on real data of household objects and it successfully detected the extruded shapes of the objects. By using the extrusion-based representation, the system could accurately capture various details of the objects' shapes.

Original languageEnglish (US)
Title of host publication2012 12th IEEE-RAS International Conference on Humanoid Robots, Humanoids 2012
Pages680-685
Number of pages6
DOIs
StatePublished - 2012
Externally publishedYes
Event2012 12th IEEE-RAS International Conference on Humanoid Robots, Humanoids 2012 - Osaka, Japan
Duration: Nov 29 2012Dec 1 2012

Publication series

NameIEEE-RAS International Conference on Humanoid Robots

Other

Other2012 12th IEEE-RAS International Conference on Humanoid Robots, Humanoids 2012
Country/TerritoryJapan
CityOsaka
Period11/29/1212/1/12

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Vision and Pattern Recognition
  • Hardware and Architecture
  • Human-Computer Interaction
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Point cloud completion using extrusions'. Together they form a unique fingerprint.

Cite this