Rapid monitoring of infrastructural health using remote sensing

Dr Lucas Hogan. Lecturer, Dr Max Stephens, Lecturer, Civil and Environmental Engineering, Dr Chris Seal, Centre for eResearch

Home

•

Project

•

Rapid monitoring of infrastructural health using remote sensing

Aotearoa, New Zealand, and the Ring of Fire

Located on the Pacific ‘Ring of Fire’, Aotearoa is a seismically active country, experiencing on the order of 15,000 earthquakes every year. While only a small fraction of these is large enough to be felt, this still leads to around 100 earthquakes of this magnitude every year. Historically, several magnitude 6 earthquakes can be expected to affect Aotearoa, New Zealand every year. (https://www.gns.cri.nz/Home/Learning/Science-Topics/Earthquakes/New-Zealand-Earthquakes).

As a result, New Zealand structural design codes have adapted to ensure that buildings are resilient and are designed to remain standing, even when damaged by a seismic event. One of the challenges with assessing the level of this damage after an earthquake is the number of structures which need to be assessed and the significant time required to turn field observations into meaningful data to allow engineers to confirm if a structure is still safe for use.. In more remote parts of the country, this poses an ongoing challenge that use of remote sensing platforms, such as remotely operated drones capturing photographic images of the infrastructure, has a potential to mitigate. These can then be combined with techniques such as photogrammetry to create 3D digital representations.

Photogrammetry

Photogrammetry is defined as the science and technology of making measurements using photographic images and has found use across the electromagnetic spectrum. With appropriate calibration, accurate measurements of surface features can be made and there are a range of software tools that can be used to facilitate this process.

Structural damage assessment

Research in the Faculty of Engineering is leveraging photogrammetry and remote sensing platforms, specifically cameras mounted on drones, to enable rapid monitoring of damage to infrastructure in response to this challenge. Images captured by these drones can be compiled to form a 3D model of the structure complete with evidence of surface damage that is indicative of the state of health of the structure assessed. This requires the use of software and GPU-enabled compute to create the 3D model which is often beyond the capabilities of a typical desktop. Once 3D reconstruction is completed, these models can then be exported and traditional image processing techniques can be utilized to identify cracks and other damage.

Open source software and GPU enabled compute

Through working with the researchers involved, the Centre for eResearch (CeR) identified our Nectar research cloud as being the most suitable for the work and provisioned an allocation of resources accordingly. Given the nature of the analysis being carried out GPU acceleration was identified as being a fundamental requirement. CeR’s Nectar research cloud maintains a number of GPUs that can be requested and booked through a user-friendly web interface, developed at The University of Auckland by CeR.

In keeping with the mantra of the research programme open source software was identified as a desired pathway for the research team. The open-source photogrammetry programme, Meshroom, provided by AliceVision (https://alicevision.org/) fits this ideology well, although the codebase to run on a virtual platform with appropriate GPU has not been tested in many cases.

CeR solutions team provided expertise, to enable the use of Meshroom on the virtual platform using GPUs, Identifying issues with a Windows based operating system, the solution involved migrating to Linux system with better OpenGL GPU support . CeR carried out the migration and ensured that the installation of Meshroom was working as expected using test data, before handing over the virtual machine for use as a research platform.