Digital Twins of Building Extraction from Dual-Channel Airborne LiDAR Point Clouds

EasyChair Preprint 10246

Abstract

The new dual-channel airborne LiDAR system can acquire dense point clouds of roofs and facades at the same time, RIEGL VQ-1560i has the most advanced dual-channel LiDAR system with unique and innovative bi-directional scanning angles, which provides better building instance extraction than traditional Scanner for more complete and precise data. This abstract presents the first point cloud building extraction for urban digital twins using dual-channel airborne LiDAR data. The main challenges of this lidar data are the large number of points, complex data structure, and multiple classes of objects. We propose a preprocessing-free architectural extraction method. It consists of three steps, namely point cloud slicing, projection, and constraint-based extraction of labels. Point cloud slices consist of top-down merging of elevation and 3D semantic segmentation to reorganize point cloud scenes into interrelated point groups. This greatly reduces the processing difficulty and computational burden of complex structures while removing multiple classes of non-building points. Second, we project the point group into an image to further reduce computational complexity while improving processing efficiency. Finally, we label the building with its up-down relationship and remap it as a 3D building. Experimental results show that the proposed method achieves an average recall rate of 95.36% and an average F1 score of 93.59%. For digital twin instance segmentation, the quality of the two public test scenarios reaches 92.86% and 98.31%, respectively.

Keyphrases: Dual-channel airborne LiDAR, building instance extraction, point clouds

@booklet{EasyChair:10246,
  author    = {Yiping Chen and Huifang Feng and Jonathan Li},
  title     = {Digital Twins of Building Extraction from Dual-Channel Airborne LiDAR Point Clouds},
  howpublished = {EasyChair Preprint 10246},
  year      = {EasyChair, 2023}}