United States Department of Transportation
i
Low-Cost 3-D LIDAR Development for Transportation
-
2023-07-01
-
[PDF-4.03 MB]
Details:
-
Creators:
-
Corporate Creators:
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Publication/ Report Number:
-
Resource Type:
-
Geographical Coverage:
-
Edition:Final Report, August 2020 -June 2023
-
Corporate Publisher:
-
Abstract:Mobile light detection and ranging (LIDAR) technology offers a significant opportunity to increase transportation safety and efficiency. However, most commercial systems are prohibitively expensive for usage. Building on past efforts, two relatively inexpensive three-dimensional single point LIDAR systems were generated. The third-generation system demonstrated efficiency, ease of use, and options for various resolution levels under the cost of $500. The fourth-generation system enhanced the system by removing leading edge bias while creating a standalone system that outpaces the prior generation by 12% in acquisition rate with a resolution four times as dense. Subsequent use of this system with an adaptive active fusion network for scene depth estimation provides upsampling and use of a convolutional spatial propagation network (CSPN) helps refine the full depth map. The performance of proposed network was evaluated and compared with the state-of-the-art methods on the NYUv2 dataset, and the experiment demonstrated its outperformance in reconstruction accuracy, reliance on the number of laser scans, and robustness. Subsequently employing a novel Iterative Statistical Outliers Removal (ISOR) method removes noise from the system and the upsampling methods. Finally, through the combination of upsampling and employing the ISOR method, a lower accuracy point cloud from the fourth generation single-point 3-D LIDAR system can be enhanced to create an accurate 3-D point cloud bordering on real time. This would make the complete system applicable for all transportation-based environments.
-
Format:
-
Funding:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
