B2DC: Balanced Point Cloud Data Reduction Based on Biplanar 2D Curvature

Citations

SCOPUS

0

초록

As the use of 3D point cloud data increases in applications such as autonomous driving, augmented reality, and robotics, there is a growing need for technologies to process this data in real-time while minimizing storage requirements. Since the efficiency of subsequent processing is highly dependent on the size of these datasets, there is growing research on point cloud reduction that effectively reduces points while preserving essential features. However, traditional methods often lack of considering feature loss, while 3D curvature-based and deep learning-based methods, though more accurate, typically involve high computational costs that challenge real-time processing. To address these challenges, we propose binplanar 2D curvature (B2DC), a novel point cloud reduction algorithm which projects 3D data onto two 2D planes. Unlike prior works, B2DC balances feature preservation and computational efficiency by leveraging 2D curvature to retain significant features while effectively reducing complexity. Both qualitative and quantitative results demonstrate that B2DC is about 30% faster compared to the 3D curvature-based approach while achieving over 90% feature preservation accuracy. © 2024 IEEE.

키워드

2D Projection; Curvature; Kdtree; Point Cloud Reduction; Voxel-Grid
제목
B2DC: Balanced Point Cloud Data Reduction Based on Biplanar 2D Curvature
저자
Park, Jaeseong; Kim, Woongsoo; Ryu, Junhyeong; Paek, Jeongyeup
DOI
10.1109/ICTC62082.2024.10827285
발행일
2024-10
유형
Conference paper
저널명
International Conference on ICT Convergence
권
2024
페이지
452 ~ 455