IEEE Xplore | arxiv | Video
This project contains the code for the IEEE IROS 2021 paper Towards Robust Monocular Visual Odometry for Flying Robots on Planetary Missions, M. Wudenka, M. G. Müller, N. Demmel, A. Wedler, R. Triebel, D. Cremers, W. Stürzl.
It is mainly build on basalt.
Main differences are:
- support of monocular setups for visual odometry (not mapping)
- new keyframe selection strategy based on negative entropy for non-inertial setups
- integration of landmarks "at infinity" to the optimization
- detection of features to track on more image pyramid levels
- computation of a scalar indicator for scale variance (stereo) or drift variance (monocular)
Clone the source code for the project and build it. For MacOS you should have Homebrew installed.
git clone --recursive https://github.com/DLR-RM/granite.git
cd granite
./scripts/install_deps.sh
Go to include/granite/utils/common_types.h and decide if the PixelType should be uint8_t or uint16_t. The PixelType determines how many bytes granite uses internally for image representation.
mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=RelWithDebInfo
make -j8
NOTE: It is possible to compile the code on Ubuntu 16.04, but you need to install cmake-3.10 or higher and gcc-7. See corresponding Dockerfile as an example.
- Camera, IMU and Mocap calibration. (TUM-VI, Euroc, UZH-FPV and Kalibr datasets)
- Visual odometry (no IMU). (MADMAX and KITTI dataset)
- Visual-inertial odometry and mapping. (TUM-VI and Euroc datasets)
- Simulation tools to test different components of the system.
The source code is provided under a MIT license. See the LICENSE file for details. Note also the different licenses of thirdparty submodules.
The original source code from Basalt was distributed under a BSD 3-clause license.