Point Cloud to BIM: A Review of Automated Scan-to-BIM Methods

Scan-to-BIM automates the conversion of laser scanning data into semantically rich building models. This review systematically categorizes existing methods into geometry-driven, data-driven and hybrid paradigms, compares their performance on public benchmarks, and discusses open challenges including occlusion handling, parameter sharing and modelling standards.

综述范围

本文系统梳理 Scan-to-BIM 领域的研究进展，将现有方法划分为几何驱动、数据驱动与混合范式三类。

主要结论

• 数据驱动方法在构件语义分割任务上已取得较高精度，但在遮挡与噪声场景下的鲁棒性仍不足；
• 几何驱动方法可解释性强，但对复杂构件的适应性有限；
• 混合范式是当前最具工程可行性的路径。

开放问题

遮挡处理、跨项目参数共享、建模标准与精度评价体系的统一，是该领域待解决的关键问题。

@article{M.Sun20242024-poi,
  title   = {Point Cloud to BIM: A Review of Automated Scan-to-BIM Methods},
  author  = {M. Sun and T. Luo and W. Li},
  journal = {ISPRS Journal of Photogrammetry and Remote Sensing},
  year    = {2024},
  doi     = {10.1016/j.isprsjprs.2024.100006},
}