Team member:
- Cuong Ha
- Warakorn Jetlohasiri
We adapted Optical flow to track and match keypoints in Visual Odometry (replace feature matching) by extending Visual Odometry pipeline.
Two Sparse Optical Flow methods were used:
- Lucas-Kanade method: OpenCV implementations.
- Usenko's KLT-based optical flow.
Optical flow were use to track and match keypoints for both frame to frame and stereo image pairs.
The extending code are mainly included in (include/visnav/optical_flow.h, include/visnav/optical_flow_utils.h, include/visnav/of_grid.h, src/optical_flow_odometry.cpp).
optical_flow.h: contains Usenko's optical flow method implementations.optical_flow_utils.h: contains find matches landmarks, localization, add landmarks function for optical flow odometry.of_grid: grid-based keypoints adding to create new flows.optical_flow_odometry: optical flow visual odometry executor.
Evaluation were processed using ATE, RPE metrics with TUM RGB-D benchmark tools. Trajectory were evalutated with RPG trajectory evaluation.
This code is a part of the practical course "Vision-based Navigation" (IN2106) taught at the Technical University of Munich.
It was originally developed for the winter term 2018. The latest iteration is winter term 2021/2022.
The authors are Vladyslav Usenko, Nikolaus Demmel, David Schubert and Zhakshylyk Nurlanov.
The code for this practical course is provided under a BSD 3-clause license. See the LICENSE file for details.
Parts of the code (include/tracks.h, include/union_find.h) are adapted from OpenMVG and distributed under an MPL 2.0 licence.
Parts of the code (include/local_parameterization_se3.hpp, src/test_ceres_se3.cpp) are adapted from Sophus and distributed under an MIT license.
Note also the different licenses of thirdparty submodules.
You can find setup instructions here.