Automated Indoor Image Localization to Support a Post-Event Building Assessment

Area of Science:

• Engineering
• Computer Vision
• Geospatial Analysis

Background:

• Image data is crucial for post-event building assessments after natural hazards.
• Global Positioning System (GPS) struggles with indoor localization due to signal interruption.
• Lack of spatial context in images hinders analysis and organization.

Purpose of the Study:

• To develop a methodology for localizing images within buildings and linking them to structural drawings.
• To overcome limitations of GPS in indoor environments for image data.
• To enhance the spatial context and usability of post-event building assessment images.

Main Methods:

• Utilizing visual odometry to reconstruct a 3D point cloud model from image streams.
• Employing structure-from-motion to create local 3D textured models of building components.
• Linking parallel image sets using time synchronization and projecting point clouds onto structural drawings.

Main Results:

• Successfully localized images and overlaid them onto structural drawings, providing essential spatial context.
• Enabled 3D reconstruction of building components and damage for detailed geospatial assessments.
• Demonstrated the technique's effectiveness through a simulated post-event building assessment.

Conclusions:

• The developed methodology significantly improves the spatial context and utility of image data for post-event building assessments.
• This approach addresses the challenge of indoor image localization where GPS is unreliable.
• The technique facilitates more comprehensive and accurate analysis of building conditions after natural hazards.