Speaker
Mr
Keitaro Taniguchi
(National Institute of Technology, Tokuyama College)
Description
The efficiency and sophistication of maintenance are expected to improve through the use of 3D models that organize damage conditions and locations. For bridges, research has focused on using 3D models constructed with technologies like SfM/MVS, though challenges such as high data collection costs and time requirements persist. This study proposes a methodology for automating 3D model construction by extracting dimensions from blueprint images using a Transformer-based object detection model and OCR. Applied to actual bridge drawings, the method successfully constructed accurate 3D models. Further development of this approach into a practical application could enable accessible, user-friendly 3D model construction.
Primary author
Mr
Keitaro Taniguchi
(National Institute of Technology, Tokuyama College)
Co-author
Dr
Tatsuro Yamane
(National Institute of Technology, Tokuyama College)
Peer reviewing
Paper
Paper files:
