Skip to content

Latest commit

 

History

History
20 lines (16 loc) · 744 Bytes

planetarySLAM.md

File metadata and controls

20 lines (16 loc) · 744 Bytes

Visual SLAM-Based Robotic Mapping Method for Planetary Construction

Author: Hong

Year: 2021

Notes:

  • SLAM based method with stereo camera and deep learning based disparity map estimation
  • Adaptation of S-PTAM, which is a stereo BA based SLAM
  • Loop closure with BoW
  • S-PTAM enables sparse mapping => necessity of a dense mapping method
  • Self-supervised method for disparity map computation based on a SSIM loss
  • Disparity map is also used to assist localization: is used to perform geometric check for loop closure and for feature matching procedure
  • CNN run at 50Hz on a single GPU
  • AKAZE used to extract and match features

Commentaires:

Papier intéressant, mais pas de contribution majeure. S'en inspirer pour mes travaux