A 3D Lidar SLAM Algorithm Based on Graph Optimization in Indoor Scene

To improve the localization precision of the 3D lidar SLAM algorithm and solve the problem of z-axis offset, a method to improve the loop closure by adding edge constrains and ground constrains is proposed based on graph optimization in indoor scene. Firstly, MLESAC is used to filter the point cloud...

Full description

Saved in:
Bibliographic Details
Published in2022 6th Asian Conference on Artificial Intelligence Technology (ACAIT) pp. 1 - 5
Main Authors Chen, Bing, Li, Jun, Song, Tao, Zhao, Zhangzhen, Chen, Dan, Jia, Rui
Format Conference Proceeding
LanguageEnglish
Published IEEE 09.12.2022
Subjects
3D lidar point clouds
3D Normal Distribution Transform
Filtering algorithms
Graph optimization
Ground constrains
Laser radar
Location awareness
MLESAC
Point cloud compression
Semantics
Simultaneous localization and mapping
Three-dimensional displays
Online AccessGet full text
DOI10.1109/ACAIT56212.2022.10137963

Cover

More Information
Summary:To improve the localization precision of the 3D lidar SLAM algorithm and solve the problem of z-axis offset, a method to improve the loop closure by adding edge constrains and ground constrains is proposed based on graph optimization in indoor scene. Firstly, MLESAC is used to filter the point cloud data and remove outlier points. Then, the random sampling is used to downsample the point cloud data to improve the speed and accuracy of the point cloud registration. Based on 3D-NDT registration algorithm, the transformation relationship of the point cloud is obtained. Finally, edge and ground constrains are added to the optimization solver to obtain the optimized pose and point cloud map. Using Velodyne lidar scanned underground garage data shows that the proposed method can improve the loop closure of lidar SLAM and improve localization accuracy.
DOI:10.1109/ACAIT56212.2022.10137963