Author: | Luo, Xianghuan |
Title: | Imaging and diagnosis of urban subsurface hazards by ground penetrating radar |
Advisors: | Lai, Wallace (LSGI) Ding, Xiao-li (LSGI) |
Degree: | Ph.D. |
Year: | 2020 |
Subject: | Ground penetrating radar Hong Kong Polytechnic University -- Dissertations |
Department: | Department of Land Surveying and Geo-Informatics |
Pages: | xiv, 131 pages : color illustrations |
Language: | English |
Abstract: | Due to the increasing city population, the demand for water supplies, water discharge, energy and telecommunication, a city's underground is getting more complex (more utility types) and congested (higher density). However, without proper care, they are unavoidably ageing and create different sorts of problems, for example, ground subsidence caused by void formation initiated after a water leak from a pipe. Their conditions require an approach of proper imaging and health diagnosis for decision-making of rehabilitation or replacement scheme, as if the diagnostic process in a hospital would be conducted for human beings before any decision of surgery and medication. Without excavation, ground penetrating radar (GPR) is the most suitable means for imaging the ground truth in terms of (1) its 2D and 3D image representations close to ground truth, (2) efficiency of data capture, (3) fine resolution up to centimetre scale, and (4) its good match of depth ranging with the utilities' depths which are in the scale of meters. For such purpose, GPR depth/time slice (C-scan) at user-selected depths is the best representation. But to date, the process of constructing 3D C-scans is still subjective and mainly relies upon the operator's knowledge and experience. Biases from human judgements or survey setting are inevitable, and therefore doubts arise on the imaging and diagnostic purpose. The main achievement of this research is to establish a first of its kind, systematic and bias-free workflow for urban's underground diagnosis with GPR C-scans and pattern recognitions of B-scan radargrams based on forward and inverse methods. Also, the thesis contributed to the development of methods integrated with image processing algorithms for extracting spatial and temporal features (i.e. hazards) from GPR C-scans. The process imitates the decision-making process normally made by skilled professionals but in a semi-automatic and more robust fashion, especially when even the most skilled professionals would fall short of the ability to handle huge volume of GPR data. This thesis contributed to the research and engineering/surveying community in the following four aspects imminently. Firstly, an object-oriented workflow for generalizing reliable C-scans was developed, with empirical thresholds/ranges of three identified and crucial parameters (profile spacing, slice thickness and interpolation algorithms). The workflow was validated after comparing C-scan images and reality through ground-truth of twenty-five carefully designed experiments conducted on concrete members, underground utilities/infrastructures and an archaeological site. Secondly, the GPR responses of air voids, the most important kind of hazard, were quantitatively analysed with five attachments (materials, size, thickness, etc.) with numerical simulations and two experiments, one in the laboratory and one in the field. It was concluded that the ratios of void spread to the GPR footprints result in various GPR response patterns, and voids were only visible when the said ratio is larger than the factor of two to three. Thirdly, a workflow integrating the pyramid pattern recognition was developed to identify air voids from GPR data semi-automatically. The void recognition workflow was validated with a series of laboratory tests and site experiments. The true positive rate (i.e. identified voids do exist) of two validation-cases was sixty-five percent, and there was no omission (true negative and false negative) with the developed workflow. Lastly, with reliable C-scans and patterns of underground voids in the GPR data, the fourth aim achieved in this study was the development of a workflow to identify temporal changes from time-lapse GPR data with image change detection techniques. The workflow used K-means clustering to classify pixels into changed or unchanged. These four imminent contributions provide a gateway towards reliable and consistent imaging and diagnosis, and a basis of time-lapsed comparison with a well-established pattern recognition database. In short, this research establishes a health diagnostic approach for the urban underground. Human subjective interventions are reduced as much as possible in this developed approach. It is believed to trigger a start for establishing a full coverage health-record for underground utilities and associated hazards with GPR. |
Rights: | All rights reserved |
Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/10592