Organisiert von:
Vortrag
A CNN-based Approach for Reliable Concrete Pavement Damage Maps
Hellen Garita-Durán (Institute for Structural Analysis, TU Dresden, Dresden)
Tania Ávila-Esquivel (Lanamme, University of Costa Rica, San José)
Michael Kaliske (Institut für Statik und Dynamik der Tragwerke, TU Dresden, Dresden)
Freitag, 29. Mai 2026; 14:10 - 14:30 Uhr in Raum HSZ/0004
Kurzfassung:
The research compares traditional field and photogrammetric Pavement Condition Index (PCI) surveys using a statistical analysis, revealing systematic deviations. To address this, a convolutional neural network-based system for damage segmentation in concrete pavements is presented. Image-based metrics and practical implementation aspects are discussed to illustrate the potential and limitations of machine learning in structural asset management.
In der Sitzung:
MS20: Maschinelles Lernen im Bauwesen
Freitag, 29. Mai 2026; 13:30 - 15:30 Uhr in Raum HSZ/0004
Ganze Sitzung anzeigen
