An AI-based approach to create spatial inventory of safety-related architectural features for school buildings

Abstract

An understanding of the spatial information for safety-related architectural features inside a school building is crucial for effective responses to emergency situations such as active shooter or fire incidences. However, efficiently creating high-quality spatial inventory for a wide variety of safety-related assets inside a large school building is a challenge for building owners. This study addresses this challenge by developing an AI-based approach for rapid generation of spatial inventory of safety-related architectural features within school buildings from point cloud data. A method for creating cross-modality data from laser scans is proposed to not only address the limitations in individual 2D and 3D datasets but also enable the transfer of the segmentation and generalization performance from the state-of-the-art Segment Anything model to point cloud segmentation. Additionally, our proposed method also demonstrates great potential to address the challenge of extracting glass façade frames from point cloud datasets.

Keywords

• School buildings
• Safety-related architectural features
• Spatial inventory
• Object detection
• Point cloud segmentation
• Glass façade extraction