Point matching and pose optimization are two important processes in LiDAR odometry. The former one is prone to noise and initial pose estimation, while the latter often falls into local minima due to improper matched points. In this paper, we propose a new point pair similarity method in the combination of the normal vector, the smallest eigenvalue of the spatial covariance matrix, and the KL divergence of local intensity values. In pose optimization step, we use both the proposed point pair similarity and planarity as the weight. Experimental results show that it may guarantee higher accuracy with our method and is able to run at 27 FPS on a common PC.
IGICP: Intensity and Geometry Enhanced LiDAR Odometry
IEEE Transactions on Intelligent Vehicles ; 9 , 1 ; 541-554
2024-01-01
5447872 byte
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
DeLiO: Decoupled LiDAR Odometry
IEEE | 2019
|Towards Image-Grade LiDAR Odometry and Mapping
IEEE | 2022
|Piecewise Linear De-skewing for LiDAR Inertial Odometry
IEEE | 2021
|