Research Statement

My research focuses on graph-based frameworks for automated Scan-to-BIM, specifically developing the Architectural Opening Object Graph Model (AOOGM) to reconstruct doors and windows from point cloud data.

AOOGM captures essential architectural components—panels, frames, load/motion-bearing elements—and their spatial relationships, enabling interpretable detection despite occlusion and noise. Complementary work explores point cloud embeddings for BIM object retrieval, bridging reality capture with digital BIM libraries to accelerate AECO digitalization.

Research Areas

Scan-to-BIM

Automated generation of Building Information Models from 3D scans

Point Cloud Data

Processing, analysis, and AECO applications of point cloud data

Architectural Openings

Computer vision for identifying maneuvering/passable spaces in built environments

Computer Vision

Machine learning applications for BIM and point cloud processing

Involved Projects

Scan-to-BIM Automation System for Built Assets Digitalization in Hong Kong

Keywords: 3D point cloud; deep learning; BIM; CIM; open BIM

Architectural openings detection from 3D point clouds: A bi-objective optimization approach addressing local hollow semantic graph and global design regularity.

Keywords: Architectural opening; LiDAR point cloud; bi-objective optimization; graph learning; supervoxel clustering

Research Impact

AOOGM and PCD-BIM embeddings enable scalable Scan-to-BIM, transforming reality capture into actionable BIM models. Applications include accessibility optimization, fire safety routing, natural ventilation analysis, and digital preservation of heritage structures.