Projective ground segmentation and complete object recovery for perceptual terrain reconstruction

Wei Song, Kyungeun Cho, Kyhyun Um, Cheesun Won, Sungdae Sim

Research output: Contribution to journalArticlepeer-review

Abstract

Terrain reconstruction and photorealistic visualization are required for the remote operation of mobile robots. We can generate textured terrain meshes by using 2D and 3D datasets obtained from multiple sensors. To detect the traversable regions of a terrain, we apply the Gibbs-Markov random field (MRF) model with a flood-fill algorithm to segment the ground and objects in the reconstructed terrain mesh and 2D images. We propose a height estimation method that recovers missing parts by finding object boundaries in 2D images and estimating the 3D coordinates of the boundaries. Our proposed methods were tested in an outdoor environment. The results show that ground data can be segmented effectively and that the unsensed parts of objects can be accurately recovered.

Original languageEnglish
Pages (from-to)985-990
Number of pages6
JournalInformation (Japan)
Volume17
Issue number3
StatePublished - Mar 2014

Keywords

  • Gibbs-Markov Random Field
  • Ground segmentation
  • Height estimation
  • Mobile robot
  • Multi-sensor integration
  • Terrain modeling

Fingerprint

Dive into the research topics of 'Projective ground segmentation and complete object recovery for perceptual terrain reconstruction'. Together they form a unique fingerprint.

Cite this