Digital twinning of bridges from point cloud data by deep learning and parametric models

M. S. Mafipour, S. Vilgertshofer, A. Borrmann

Publikation: Beitrag in Buch/Bericht/KonferenzbandKonferenzbeitragBegutachtung

4 Zitate (Scopus)

Abstract

The Digital Twin (DT) of a bridge is a geometric-semantic model that supports and facilitates the operation and maintenance process of the structure. For existing structures, the semantically enriched 3D model of the DT is typically created by processing point cloud data (PCD). Semantic segmentation and parametric modeling are two essential but laborious steps in the digital twinning of bridges. This paper contributes to automating these steps by applying deep learning and metaheuristic algorithms. Semantic features of points are extracted, and a deep learning model is trained. Subsequently, the segmented parts are parametrically modeled by applying a metaheuristic algorithm for model fitting. The presented results show that the DT of bridges can be created with a mean intersection over union (mIoU) of 88.45% and mean accuracy (mAcc) of 95.62% in semantic segmentation, as well as a mean absolute error (MAE) of 4 cm/m in parametric modeling.

OriginalspracheEnglisch
TiteleWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022
Redakteure/-innenEilif Hjelseth, Sujesh F. Sujan, Raimar J. Scherer
Herausgeber (Verlag)CRC Press/Balkema
Seiten543-550
Seitenumfang8
ISBN (Print)9781032406732
DOIs
PublikationsstatusVeröffentlicht - 2023
Veranstaltung14th European Conference on Product and Process Modelling, ECPPM 2022 - Trondheim, Norwegen
Dauer: 14 Sept. 202216 Sept. 2022

Publikationsreihe

NameeWork and eBusiness in Architecture, Engineering and Construction - Proceedings of the 14th European Conference on Product and Process Modelling, ECPPM 2022

Konferenz

Konferenz14th European Conference on Product and Process Modelling, ECPPM 2022
Land/GebietNorwegen
OrtTrondheim
Zeitraum14/09/2216/09/22

Fingerprint

Untersuchen Sie die Forschungsthemen von „Digital twinning of bridges from point cloud data by deep learning and parametric models“. Zusammen bilden sie einen einzigartigen Fingerprint.

Dieses zitieren