SegMatch is a reliable loop-closure detection algorithm based on the matching of 3D segments. The SegMatch code is open-source (BSD License) and has been tested under Ubuntu 14.04 and ROS Indigo. Please remember that this is on-going research code which is subject to changes in the future. Several new features and demonstrations will be added soon.
Please consult our paper, video and wiki for the algorithm description and for instructions on running demonstrations. We recently uploaded a second video featuring SegMatch in a multi-robot configuration.
The following configuration was tested under Ubuntu 14.04 and ROS indigo. Please see the final note if you want to compile under ROS Kinetic.
First install the required system packages:
$ sudo apt-get install libopencv-dev python-wstool doxygen
Then use wstool for installing catkin dependencies:
$ cd ~/catkin_ws/src
$ wstool init
$ wstool merge segmatch/dependencies.rosinstall
$ wstool update
Finally build the laser_mapper package which will compile all SegMatch modules:
$ cd ~/catkin_ws
$ catkin build -DCMAKE_BUILD_TYPE=Release laser_mapper
See this link for installing catkin_tools. Building dependencies will require some time according to which new packages need to be built (eg. Building the pcl_catkin package can take up to two hours). Building pcl_catkin might fail if you do not have sufficient RAM. It can help to add -j2 to catkin build in order to limit the parallel jobs and reduce memory usage.
Consult the wiki for instructions on running the demonstrations.
Note: If you are using ROS Kinetic, you might want to run the following command in your catkin workspace prior to building the packages:
$ catkin config --merge-devel
We would be very grateful if you would contribute to the code base by reporting bugs, leaving comments and proposing new features through issues and pull requests. Please see the dedicated wiki page on this topic and feel free to get in touch at rdube(at)ethz.ch. Thank you!
Thank you for citing our SegMatch paper if you use any of this code:
@inproceedings{segmatch2017,
title={SegMatch: Segment based place recognition in 3D point clouds},
author={Dub{\'e}, Renaud and Dugas, Daniel and Stumm, Elena and Nieto, Juan and Siegwart, Roland and Cadena, Cesar},
booktitle={International Conference on Robotics and Automation (ICRA)},
pages={5266--5272},
year={2017},
organization={IEEE}
}