Paper Link: https://ieeexplore.ieee.org/document/9690581

Title: Stereo Orientation Prior for UAV Robust and Accurate Visual Odometry

Abstruct: This paper presents a novel outlier rejection approach for feature-based visual odometry. The proposed approach is based on an empirical observation that shows that some 2D-3D correspondences with very low reprojection error can cause a high error in pose estimation. This work exploits such observations for odometry when a stereo camera is available. We argue that an explicit pose error measure is desired over that of implicit reprojection -- whenever the former is possible -- to classify correspondences into inliers vs. outliers for robust long-term odometry. To explicitly measure the plausible pose error, we derive bounds on the individual pose parameters with the help of the known orientation of stereo cameras. In this process, we formulate our bounds using the sum-of-square polynomials, which allow us to test whether a given correspondence satisfies any solution within the expected bounds. If the correspondence does not satisfy bounds for any parameter, it is considered to be an outlier. We implemented and tested the proposed method for the unmanned aerial vehicle (UAV) indoor navigation. The experiments from both benchmark evaluations (EuRoC and KITTI) and UAV onboard tests indicate that the inlier group refined by the proposed method significantly improves odometry estimation compared to the traditional outlier rejection methods. In fact, the proposed method performs as accurately as inertial measurement unit (IMU) aided methods in the state of the art.

EuRoC MAV Dataset, Intel Realsense T265, KITTI(ros melodic)

We have tested in the following environment:

Ubuntu 16.04 + ROS Kinetic (recommend)

Ubuntu 18.04 + ROS melodic (parameters in config file need to be refined)

4.1 Clone the repository to the catkin work space eg. /catkin_ws/src

4.2 Install sopvo:

./install_sopvo_XXXX.sh

4.3 Compile

cd ~/catkin_ws
catkin_make

5.1 EuRoC MAV Dataset Download the dataset(say MH_05_difficult) into the bag folder:

roscd sopvo/bag/
wget http://robotics.ethz.ch/~asl-datasets/ijrr_euroc_mav_dataset/machine_hall/MH_05_difficult/MH_05_difficult.bag

Edit the corresponding bag name in sopvo_euroc.launch file:

<node pkg="rosbag" type="play" name="rosbag" args="$(find sopvo)/bag/MH_05_difficult.bag"/>

run the following launch files:

roslaunch sopvo sopvo_euroc.launch

5.2 Intel Realsense T265 (live version): Make sure you have installed realsense-ros properly. To run the live demo:

roslaunch sopvo sopvo_t265_live.launch

5.3 KITTI: Please download the odometry dataset(gray image), then use kitti_img2rosbag_tool to convert the sequense into rosbag. There are different image resolutions of the KITTI datasets, please modify the config files for the testing. In folder ./config we provide two examples: kitti_1226.yaml and kitti_1241.yaml.

KITTI benchmark results and demo video:

Please install our Gazebo simulation tool from: https://github.com/rduan036/E2ES.git

roslaunch sopvo sopvo_gazebo_sim.launch

SOPVO + object tracking (https://github.com/arclab-hku/ICRA2021tracking):

UAV platform: lattepanda + D435i

Ran Duan, ArcLab at PolyU, HK, China [email protected] [email protected]

I would like to thank Dr. Shengyang Chen(Dept.ME,PolyU) for his assistant in coding.

R. Duan, D. P. Paudel, C. Fu and P. Lu, "Stereo Orientation Prior for UAV Robust and Accurate Visual Odometry," in IEEE/ASME Transactions on Mechatronics, doi: 10.1109/TMECH.2022.3140923.

Latex:

@ARTICLE{9690581, author={Duan, Ran and Paudel, Danda Pani and Fu, Changhong and Lu, Peng}, journal={IEEE/ASME Transactions on Mechatronics}, title={Stereo Orientation Prior for UAV Robust and Accurate Visual Odometry}, year={2022}, volume={27}, number={5}, pages={3440-3450}, doi={10.1109/TMECH.2022.3140923}}