The project delivers two pilot case bridges: (1) a metal truss bridge on Gaagweg street, Schipluiden, the Netherlands, (2) a Constructie bridge on Weaslandseweg street, Delft, The Netherlands. All of them were scanned by a terrestrial laser scanner. Unfortunately, due to Covid-19, data acquisition using an UAV-laser scanning to capture a bridge structure could not be organized.
The project proposed two segmentation methods called cell-based and voxel-based region growing methods for point cloud segmenting. The methods can be applied for various types of structures for example bridges and buildings. Moreover, the statistical model has been implemented to support to automatically select parameters for example cell size and voxel size. Particularly, the statistical model can automatically select and adjust thresholds for the segmentation method regarding to the quality of data set and structure size.
The project proposed a new framework to extract point clouds describing surfaces of bridge components. The framework deployed spatial point clouds and contextual knowledge of bridge components to extract bridge structural elements in a consecutive order: superstructure and substructure. For each structural element, the proposed method consists of two main steps called coarse extraction and fine filtering. In the coarse extraction, candidate points of a structure were extracted while the fine filtering aimed to obtain the final points of the surface. As such, the framework would handle a massive data points of the bridge easily but also gave high accuracy of an extraction. The proposed framework has been tested on two box or slab bridge, and all bridge elements were successfully extracted without any manual intervention. Moreover, the framework was also extended for extracting the building structures successfully.
Additionally, the project has developed three methods called point-to-point, point-to-cell and cell-to-cell to measure the deformation of structural members. The method was used to measure the vertical clearance of the bridge and deformation of an individual structure. The method was also extended to measure other structures like beams and columns of a building. Additionally, the project also developed machine learning-based methods to determine the scaling/spalling of a concrete structure.
The project proposed a methodology to generate 3D models of bridge components. The method was successfully reconstructed the 3D bridge pier model by using a sweeping technique based on the cross-sections along a principal direction of the pier.
Results of this project were presented in 8 top conferences (one conference was postponed to 2021 due to Covid-19) in fields of remote sensing, building information modelling (BIM), bridge and building structural engineering. Moreover, one journal manuscript is in review and two others are in preparation. All publications are shared in a TU Delft repository for open access.
During the project, the researcher (Dr. Linh Truong-Hong) was invited to deliver a talk Fugro Company in the Netherlands and at the workshop “Point Cloud – 3D Technology” held at Ho Chi Minh City University of Technology (HCMUT), Viet Nam. Additionally, the researchers have also organized the short course on “Laser scanning for structural and environmental engineering” at HCMUT, Viet Nam, which was attracted more than 20 students and professional surveyors. Results of the project was also presented at NCG (Netherlands Geodetic Commission) symposium 2019 and 2020.
The researcher also presents the project at Research day of Geoscience and Remote sensing, and Geoscience Engineering, in which participants are from academia and professional. Moreover, during the project, the researcher was also served as a guest lecturer for MSc class at Dept. of Geoscience and Remote Sensing and supervised BSc and MSc thesis.
As Covid-19, the researcher cannot organize a half-day workshop, the researcher contacted to potential companies to present the results of the project and to establish collaboration. A list of the companies includes Leap3D (www.leap3D) and Asset Insight (
https://assetinsight.nl/(se abrirá en una nueva ventana)) and SSIFT (
https://ssift.de/(se abrirá en una nueva ventana)) in testing the algorithms and translating knowledge of laser scanning for bridge surveying, bridge inspection, and bridge reconstruction, VMT (
https://vmts.ch/(se abrirá en una nueva ventana)) for using laser scanning in building inspection, Acernis (
https://www.acernis.ch/(se abrirá en una nueva ventana)) in generating a digital twin for power transmission tower, and BlackWolf (
https://blackwolfinspection.com/(se abrirá en una nueva ventana)) for measuring the deformation of oil tanks. Moreover, the researcher also discussed with Fugro (www.fugro.com/nl) in collaborating on processing point clouds for the railway.
