Panella, Fabio (2023) Automating inspection of tunnels with photogrammetry and deep learning. PhD thesis, University College London, UK.
Abstract
Asset Management of large underground transportation infrastructure requires frequent and detailed inspections to assess its overall structural conditions and to focus available funds where required. At the time of writing, the common approach to perform visual inspections is heavily manual, therefore slow, expensive, and highly subjective. This research evaluates the applicability of an automated pipeline to perform visual inspections of underground infrastructure for asset management purposes. It also analyses the benefits of using lightweight and low-cost hardware versus high-end technology. The aim is to increase the automation in performing such task to overcome the main drawbacks of the traditional regime. It replaces subjectivity, approximation and limited repeatability of the manual inspection with objectivity and consistent accuracy. Moreover, it reduces the overall end-to-end time required for the inspection and the associated costs. This might translate to more frequent inspections per given budget, resulting in increased service life of the infrastructure. Shorter inspections have social benefits as well. In fact, local communities can rely on a safe transportation with minimum levels of disservice. At last, but not least, it drastically improves health and safety conditions for the inspection engineers who need to spend less time in this hazardous environment. The proposed pipeline combines photogrammetric techniques for photo-realistic 3D reconstructions alongside with machine learning-based defect detection algorithms. This approach allows to detect and map visible defects on the tunnel’s lining in local coordinate system and provides the asset manager with a clear overview of the critical areas over all infrastructure. The outcomes of the research show that the accuracy of the proposed pipeline largely outperforms human results, both in three-dimensional mapping and defect detection performance, pushing the benefit-cost ratio strongly in favour of the automated approach. Such outcomes will impact the way construction industry approaches visual inspections and shift towards automated strategies.
| Item Type: | Thesis (Doctoral) |
|---|---|
| Index terms: | mapping, machine learning, tunnel, social benefit, health and safety, manager, visual inspection, accuracy, strategy, asset management, subjectivity, hardware, transportation infrastructure, automation, construction industry, service life, reconstruction, pipeline, inspection, deep learning, underground infrastructure, engineer |
| Subjects: | practitioner, human factors and perception, professional practice, theoretical framing, spatial and geospatial analysis, automation and robotics, health safety and environment, building construction, asset management, quality assurance, profession, computer hardware, artificial intelligence, management, professional development, infrastructure and transport systems, industry analysis |
| Topics: | Engineering Principles, Health and Safety, Quality Management, Research Practice, Information Management, Business Strategy, Roles and Professions, Digital Applications |
| Descriptive scope: | 3 PCT |
N.B. Descriptive scope is a count of how many of the five facets of empirical research are indicated by the words used in title, abstract and keywords. It is not intended as a judgement on the research; merely a count of the kind of word we would expect to indicate Phenomenon, Concepts, Theoretical framing, Empirical techniques, Analytical techniques. If all five are present, then a code of “5 PCTEA” will indicate this. If you feel the coding for this record is questionable, we welcome discussion around the terms we matched or the way we categorized them. The facet you would expect may not be coded, or a facet may be coded inappropriately. This can also bear on a larger question, of which facets should be treated as defining in construction management research. Please get in touch, and we will look at it. More details here