This code contains an algorithm to compute stereo visual odometry by using both point and line segment features.
Authors: Ruben Gomez-Ojeda and Javier Gonzalez-Jimenez
Related publication: Robust Stereo Visual Odometry through a Probabilistic Combination of Points and Line Segments
If you use PL-StVO in your research work, please cite:
@InProceedings{Gomez2015,
Title = {Robust Stereo Visual Odometry through a Probabilistic Combination of Points and Line Segments},
Author = {Gomez-Ojeda, Ruben and Gonzalez-Jimenez, Javier},
Booktitle = {Robotics and Automation (ICRA), 2016 IEEE International Conference on},
Year = {2016},
Publisher = {IEEE}
}
The provided code is published under the General Public License Version 3 (GPL v3). More information can be found in the "GPU_LICENSE.txt" also included in the repository.
Please do not hesitate to contact the authors if you have any further questions.
It can be easily found at http://opencv.org. In the case of line segments, we have modified the line_descriptor from the opencv_contrib repository, included in the 3rdparty folder.
Installation on Ubuntu:
sudo apt-get install libboost-dev
Installation on Ubuntu:
sudo apt-get install libyaml-dev
In case of using the provided representation class or the PointGrey Bumblebee2 Stereo Camera application. Download and install instructions can be found at: http://www.mrpt.org/
A CMakeLists.txt file is included to detect external dependencies and generate the project.
The project builds 2 different applications to evaluate and visualize it.
The first one is "imagesStVO", a customizable application where the user must introduce the inputs to the SVO algorithm, and then process the provided output.
The second one, called "bumblebeeSVO", is an application that computes stereo visual odometry between the successive frames readed by a PointGrey Bumblebee2 stereo camera, and shows a 3D visualization of the camera motion. It is built or not depending on the CMake variable "HAS_MRPT".