An Effective Voxelization Method for LiDAR-based 3D Object Detection

Rongxuan Wang, Chao Yang, Weida Wang, Changle Xiang, Ying Li*

*此作品的通讯作者

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

LiDAR-based 3D object detection is crucial for intelligent vehicles to perceive the environment. Voxelization is an important method that can convert unstructured point clouds into structured tensors. However, for LiDARs with different resolutions, what voxel size to use remains to be studied. Therefore, this paper proposes a voxelization method for LiDAR-based 3D object detection. First, the point cloud distribution characteristics of LiDARs with different resolutions are analyzed. Then, a voxelization method adapted to LiDARs with varying resolutions for 3D object detection is proposed. Finally, models based on SECOND and Voxel R-CNN are built and the proposed voxelization method is added to more effectively utilize the point cloud information. Experiments on datasets show that the proposed voxelization method can improve the average precision of the 3D object detection models.

源语言英语
主期刊名2022 6th CAA International Conference on Vehicular Control and Intelligence, CVCI 2022
出版商Institute of Electrical and Electronics Engineers Inc.
ISBN(电子版)9781665453745
DOI
出版状态已出版 - 2022
活动6th CAA International Conference on Vehicular Control and Intelligence, CVCI 2022 - Nanjing, 中国
期限: 28 10月 202230 10月 2022

出版系列

姓名2022 6th CAA International Conference on Vehicular Control and Intelligence, CVCI 2022

会议

会议6th CAA International Conference on Vehicular Control and Intelligence, CVCI 2022
国家/地区中国
Nanjing
时期28/10/2230/10/22

指纹

探究 'An Effective Voxelization Method for LiDAR-based 3D Object Detection' 的科研主题。它们共同构成独一无二的指纹。

引用此

Wang, R., Yang, C., Wang, W., Xiang, C., & Li, Y. (2022). An Effective Voxelization Method for LiDAR-based 3D Object Detection. 在 2022 6th CAA International Conference on Vehicular Control and Intelligence, CVCI 2022 (2022 6th CAA International Conference on Vehicular Control and Intelligence, CVCI 2022). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/CVCI56766.2022.9965035