indoorSLAM: Robust RGB-D SLAM based on Plane-Line-Point features for indoor scenes

We offer a master thesis/guided research topic in the field of computer vision and robotics.

In this project, we aim to develop a robust RGB-D visual odometry system for indoor environments based on the assumption of Manhattan World (MW)[2]. For indoor scenes, structured information, including parallel and orthogonal planes and lines, can be used to build global constraints. In this way, it is possible to estimate the camera pose accurately without using global bundle adjustment, which can not only improve the accuracy and robustness of the system but also reduce computation cost.

In reality, furniture and other objects are not always parallel to the wall. So using one MW for the whole environment is often not correct for realistic environments. In the project, we would like to develop a mixture manhattan world model to represent structured objects and track the models over frames.

REFERENCE

• [1]Kim et al. Low-Drift Visual Odometry in Structured Environments by Decoupling Rotational and Translational Motion. ICRA2018
• [2]Kim et al. Visual Odometry with Drift-Free Rotation Estimation Using Indoor Scene Regularities. BMVC2017
• [3]Straub et al. The Manhattan Frame Model – Manhattan World Inference in the Space of Surface Normals. TPAMI2018
• [4]Li et al. Leveraging Structural Regularity of Atlanta World for Monocular SLAM. ICRA2019

If you are interested in any of these topics, please contact us via e-mail.

Advisor: Federico Tombari
Supervision by: Yanyan Li