Submission name:

Bassier & Vergauwen

Short description of the method:

The method is a general reconstruction pipeline with the following steps:

1. Region growing oversegmentation into smooth planar segments
2. Conditional Random field that merges similar segments (https://doi.org/10.5194/isprs-archives-XLII-2-W8-25-2017)
3. machine learning classification with Bagged Trees classifier {floor, ceiling, roof, wall, beam, other} based on local and contextual features (https://www.sciencedirect.com/science/article/pii/S2352710217305703?via%3Dihub)
4. Conditional Random field to cluster segments per wall (can be complex walls)(https://doi.org/10.5194/isprs-archives-XLII-2-W9-101-2019)
5. parametric partial wall reconstruction with RANSAC & TLS (straight, curved or polyline wall fit on cluster)
6. topology reconstruction between walls with connection evaluations (intersecting, orthogonal, blended and direct connections)

Reference:

Bassier, M., Vergauwen, M., 2020. Unsupervised reconstruction of Building Information Modeling wall objects from point cloud data. Automation in Construction, 120, 103338.

URL:

https://github.com/Saiga1105/Scan-to-BIM-Grasshopper

Submission date:

5 Jun. 2020

Last update:

5 Jun. 2020