Scan Denoising and Normal Distribution Transform for Accurate Radar Odometry and Positioning
Rongxi Zhang, Yuanhui Zhang, Duo Fu, Kang Liu
Abstract
This letter proposes an accurate scanning radar odometry and a radar positioning algorithm based on radar maps. Based on the measurement principle of mechanical scanning Frequency-Modulated Continuous Wave (FMCW) radar, we propose a Scan Denoising (SD) method and an improved Normal Distributions Transform (NDT) algorithm for our radar odometry and positioning. The SD method has a good filtering effect on radar noise, which is robust under different environmental conditions. We improved the registration accuracy of radar point clouds by adjusting outlier ratios of NDT cells instead of weighting. The proposed approach has been tested on three publicly available radar datasets: Oxford Radar RobotCar Dataset, MulRan Dataset, and RADIATE Dataset. All the results show good accuracy of our radar odometry and map-based positioning algorithm under various working conditions.
BibTeX
@inproceedings{ral2023_scandenoisingand,
title = {Scan Denoising and Normal Distribution Transform for Accurate Radar Odometry and Positioning},
author = {Rongxi Zhang and Yuanhui Zhang and Duo Fu and Kang Liu},
booktitle = {RA-L 2023},
year = {2023}
}