Abstract
Recently, research on the hardware system for generating point cloud data through 3D LiDAR scanning has improved, which has important applications in autonomous driving and 3D reconstruction. However, point cloud data may contain defects such as duplicate points, redundant points, and an unordered mass of points, which put higher demands on the performance of hardware systems for processing data. Simplifying and compressing point cloud data can improve recognition speed in subsequent processes. This paper studies a novel algorithm for identifying vehicles in the environment using 3D LiDAR to obtain point cloud data. The point cloud compression method based on the nearest neighbor point and boundary extraction from octree voxels center points is applied to the point cloud data, followed by the vehicle point cloud identification algorithm based on image mapping for vehicle recognition. The proposed algorithm is tested using the KITTI dataset, and the results show improved accuracy compared to other methods.
| Original language | English |
|---|---|
| Pages (from-to) | 1899-1910 |
| Number of pages | 12 |
| Journal | Tehnicki Vjesnik |
| Volume | 30 |
| Issue number | 6 |
| DOIs | |
| Publication status | Published - 25 Oct 2023 |
| Externally published | Yes |
Bibliographical note
This research was funded by Guangdong Provincial Department of Education New Generation Information Technology Key Special Field Fund project: Research on Key Technologies of Encoding and Decoding of Real-time Vehicular Lidar Point Cloud Sequences, grant number 2021ZDZX1123; and the innovation team of Big Data Analysis and Decision of Discrete Manufacturing project, grant number 2022KCXTD064.Keywords
- 3D LiDAR
- 3D reconstruction
- autonomous driving
- compression
- point cloud
- vehicle recognition