AUTHOR=Tang Fangzhou , Zhang Shuting , Zhu Bocheng , Sun Junren 

TITLE=Outdoor large-scene 3D point cloud reconstruction based on transformer

JOURNAL=Frontiers in Physics

VOLUME=Volume 12 - 2024

YEAR=2024

URL=https://www.frontiersin.org/journals/physics/articles/10.3389/fphy.2024.1474797

DOI=10.3389/fphy.2024.1474797

ISSN=2296-424X

ABSTRACT=3D point clouds collected by low-channel Light Detection and Ranging (LiDAR) are relatively sparse, while high-channel LiDAR is costly for application. To address this issue, an outdoor large scenes point cloud reconstruction (LSPCR) technique based on Transformer is proposed. LSPCR first projects the original sparse 3D point cloud onto a 2D range image, then enhances the resolution in the vertical direction of the 2D range image, and finally converts the highresolution range image back into a 3D point cloud to obtain the reconstructed point cloud data. Experiments on the real-world KITTI dataset show that LSPCR achieves an average accuracy improvement of over 60% compared to non-deep learning algorithms, and it achieves better performance compared to latest deep learning algorithms. Therefore, LSPCR can serve as an effective solution for sparse point cloud reconstruction, effectively addressing the challenge of obtaining high-resolution LiDAR point clouds.