Research Experience — Automated SLAM-Based Room Labeling

Fall 2024

During this research project, I developed an automated pipeline to assign room labels to 360-degree construction imagery without requiring manual tagging. By combining visual-inertial SLAM with simplified floorplans derived from BIM, my work enabled full spatial referencing of thousands of panoramic frames collected on active construction sites. I designed the SLAM alignment, room segmentation, and georeferencing algorithms while also integrating PCA-based point cloud processing and trajectory slicing. The system eliminates the tedious process of manually tagging images and significantly accelerates progress tracking, inspection workflows, and safety monitoring.

Contributions

• Automated room detection in 360° construction site imagery by integrating SLAM trajectories with BIM floor plans.
• Achieved 87% labeling accuracy through 3D reconstruction and PCA-based alignment.
• Co-authored a paper published at ACCCBE 2025.

Publication

Automated Room-Level Labelling of 360 Degree Video Frames Using SLAM

ACCCBE 2025

View DOI |Download PDF

Media