Our Papers

Explore our published papers at RATE@UM.

2026

3
Semantic BIM enrichment for firefighting assets: Fire-ART dataset and panoramic image-based 3D reconstruction
ISPRS Journal of Photogrammetry and Remote Sensing JournalSelected

Semantic BIM enrichment for firefighting assets: Fire-ART dataset and panoramic image-based 3D reconstruction

Wen, Y., Qiao, Y., Lam, C. C., Brilakis, I., Lee, S., and Wong, M. O.*

Inventory management of firefighting assets is crucial for emergency preparedness, risk assessment, and on-site fire response. However, conventional methods are inefficient due to limited capabilities in automated asset recognition and reconstruction. To address the challenge, this research introduces the Fire-ART dataset and develops a panoramic image-based reconstruction approach for semantic enrichment of firefighting assets into BIM models. The Fire-ART dataset covers 15 fundamental assets, comprising 2,626 images and 6,627 instances, making it an extensive and publicly accessible dataset for asset recognition. In addition, the reconstruction approach integrates modified cube-map conversion and radius-based spherical camera projection to enhance recognition and localization accuracy. Through validations with three real-world case studies, the proposed approach achieves an average F1-score of 83.3% and an average localization error of 0.37 m, respectively. The Fire-ART dataset and the reconstruction approach offer valuable resources and robust technical solutions to enhance the accurate digital management of fire safety equipment.

BIM and GIS-integrated multi-stakeholder fire risk assessment model for heritage buildings
International Journal of Disaster Risk Reduction Journal

BIM and GIS-integrated multi-stakeholder fire risk assessment model for heritage buildings

Qiao, Y., Lam, C. C.*, Zhong, H., Pan, M., Zhang, T., and Wong, M. O.*

Heritage buildings are vulnerable to fires due to combustible materials, outdated suppression systems, and conservation constraints. Systematic fire risk assessment is therefore essential to appraise building conditions, identify dangers early, and enable proactive mitigation. However, current practice suffers from insufficient risk identification, limited multi-stakeholder involvement, and low integration with digital technologies. This study aims to develop RATER, a multi-stakeholder fire risk assessment model integrating AHP-derived expert weights with data-driven analysis for heritage buildings. RATER identifies 17 fire risk factors across fire hazards, heritage and firefighting equipment, surrounding environment, and fire safety management categories. To quantify risk weights, a pairwise comparison survey involving 53 professionals from fire services, heritage management, conservation, and safety engineering domains is conducted. The analytic hierarchy process (AHP) is applied to derive weighted priorities of fire risk factors and reveal consensus and divergence in expert judgments. The study further conducts professional-centric assessment and data-driven analysis using building information modeling (BIM) and geographic information systems (GIS) to display quantitative geo-spatial risk distribution. Through this dual assessment, the model is validated on three UNESCO heritage buildings in Macao. The results indicate that ignition sources, combustible materials, and integrated fire service installations are the top-ranked risk factors. Moreover, professional-centric and BIM-GIS data-driven results present high consistency in risk quantification, which reinforces the validity of the findings and offers complementary insights. RATER provides a structured fire risk model supported by multi-stakeholder insights and digital technology integration. Practically, it enables managers and policymakers to identify high-risk areas and support proactive mitigation planning.

Toward Interactive Crowd Dynamics in Virtual Built Environment: A BIM-driven and Macro–Micro Integrated Crowd Simulation Framework
Developments in the Built EnvironmentJournal

Toward Interactive Crowd Dynamics in Virtual Built Environment: A BIM-driven and Macro–Micro Integrated Crowd Simulation Framework

Liang, H., Sun, J., Wong, S. C., Liu, X., Lam, C. C., and Wong, M. O.*

Interactive crowd dynamics are crucial for achieving high-fidelity pedestrian behavior in virtual built environments. This study presents IntCrowd, a novel framework for enhancing interactive crowd simulation within immersive virtual environments. The framework comprises a BIM-driven navigational graph, a macro—micro integrated crowd simulation model comprising multidirectional potential fields and modified social force model, and a human—agent information fusion system. Simulation experiments show IntCrowd is more efficient than agent-based models, achieving an 85.6% reduction in additional computational time for generating more agents. A human-in-the-loop experiment conducted on a CAVE platform further demonstrates that IntCrowd can produce realistic pedestrian behavior closely aligned with real-world characteristics on the pedestrian fundamental diagram. The experiment also indicates IntCrowd significantly increases pedestrians’ perceived presence. The findings underscore the importance of high-fidelity and interactive crowd simulation for experiencing virtual built environments. Moreover, IntCrowd offers a more reliable testbed for studying pedestrian behavior in virtual built environments.

2024

1
Semantic 3D reconstruction-oriented image dataset for building component segmentation
Automation in ConstructionJournalSelected

Semantic 3D reconstruction-oriented image dataset for building component segmentation

Wong, M.O., Ying, H.*, Yin, M.*, Yi, X., Xiao, L., He, C., & Tang, L.

In image-driven 3D building reconstruction, instance segmentation is fundamental to pixel-wise building component detection, which can be fused with 3D data like point clouds and meshes via camera projection for semantic reconstruction. While deep learning-based segmentation has obtained promising results, it relies heavily on large-scale datasets for training. Unfortunately, existing large-scale image datasets often include irrelevant objects that obstruct building components, making them unsuitable for 3D building reconstruction. This paper addresses this gap by introducing a large-scale building image dataset to facilitate building component segmentation for 3D reconstruction. The dataset comprises 3378 images captured from both interiors and exteriors of 36 university buildings, annotated with 49,380 object instances across 11 classes. Rigorous quality control measures were employed during data collection and annotation. Evaluation of five typical deep learning-based instance segmentation models demonstrates the dataset's suitability for training and its value as a benchmark dataset for building component segmentation.