| 摘 要: 点云数据的高维、稀疏和非结构化特性使得对其进行大规模智能分析和处理充满挑战。近年来，基于深度学习的点云处理和分析方法学习了点云的深层特征表示，展现出较强的鲁棒性和泛化能力。但现有方法面临点云数据规模增长和结构复杂性提高的问题，需要探索更高效、鲁棒的点云分析算法。为此，提出了一种稀疏结构感知的三维点云表示学习方法。该方法面向不同的数据集，学习到具有丰富语义的结构点，并在点云形状分类、语义分割等任务中进行了实验论证与分析，验证了其高效性、稳定性。 |
| 关键词: 深度学习 结构分析 三维点云 稀疏结构表示 对比学习 |
|
中图分类号:
文献标识码:
|
|
| Unsupervised Structural Point Learning for 3D Point Clouds Driven by Contrastive Learning |
|
zhangxiangyi, doufeng, sunxinglu, wanghao
|
Dalian Neusoft University of Information
|
| Abstract: The high-dimensional, sparse and unstructured characteristics of point cloud data make it challenging to conduct large-scale intelligent analysis and processing. In recent years, point cloud processing and analysis methods based on deep learning have learned deep feature representations of point clouds, showing strong robustness and generalization ability. However, the existing methods face the problem of the growth of point cloud data scale and the increase of structural complexity, so it is necessary to explore more efficient and robust point cloud analysis algorithms. Therefore, a sparse structure aware 3D point cloud representation learning method is proposed. This method is oriented to different data sets, learns structural points with rich semantics, and conducts experimental demonstration and analysis in tasks such as point cloud shape classification and semantic segmentation, which verifies its efficiency and stability. |
| Keywords: deep learning structure analysis 3D point cloud sparse structure representation contrastive learning |
