Authors: Erik Stenborg, Torsten Sattler and Lars Hammarstrand.
This repository consist of a set of tools to perform sequential visual localization of a moving camera platform (multi-camera rig) using odometry and a fixed map. The tool set includes 2 set of tools. The first is ĹocSLAM which is a hybrid SLAM system based on ORB_SLAM2 that computes a precise platform (body) trajectory in relation to a pre-built fixed map using images from a set of cameras, odometry, and fixed map matches (local features). The second is a set of Matlab scripts to perform coarse localization using odometry and coarse positioning information from either GPS or image retrieval, e.g. DenseVLAD. Please view the specific sub-repository for more details regarding each tool.
Erik Stenborg, Torsten Sattler and Lars Hammarstrand. Using Image Sequences for Long-Term Visual Localization, 2020 International Conference on 3D Vision (3DV), 2020, pp. 938-948, doi: 10.1109/3DV50981.2020.00104. PDF
Our tools are released under a GPLv3 license.
If you use any of the tools in an academic work, please cite:
@inproceedings{stenborg2020using,
title={Using Image Sequences for Long-Term Visual Localization},
author={Stenborg, Erik and Sattler, Torsten and Hammarstrand, Lars},
booktitle={2020 International Conference on 3D Vision (3DV)},
pages={938--948},
year={2020},
organization={IEEE}
}
We have tested the library in Ubuntu 16.04, and 20.04, but it should be easy to compile/run on other platforms.