Carleton University
An Amodal Segmentation Pipeline for Critical Infrastructure Asset Imaging
Abstract
dc:description.abstractAs automated image-based critical infrastructure inspection sees wider adoption, there is a need to alleviate issues and occlusion affecting the quality of road-based infrastructure imagery. We have addressed this need by developing an image acquisition pipeline to provide artifact- and occlusion-free images of a specified infrastructure asset at a given location using a geotagged image dataset. We have developed deep learning models to reliably detect problematic imaging artifacts typical of non-specific image sources and compared their performance with promptable zero-shot models. We have developed a general method to detect the presence of obstructions occluding an asset using state-of-the-art amodal foundational models. Robust performance is observed, resulting in problem-free imaging in 68.5% of cases (up from 43.0%), while requiring an average of 1.72 images per asset. The proposed pipeline will be of interest to disaster response teams, utilities, and other critical infrastructure asset managers.
Degree
thesis:*- Name thesis:degree_name
- Master of Applied Science (M.App.Sc.)
- Level thesis:degree_level
- Master's
- Discipline thesis:degree_discipline
- Data Science and Analytics
- Grantor dc:publisher
- Carleton University
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Dupuis, Jonathan
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- Copyright © 2024 the author(s). Theses may be used for non-commercial research, educational, or related academic purposes only. Such uses include personal study, distribution to students, research and scholarship. Theses may only be shared by linking to the Carleton University Institutional Repository and no part may be copied without proper attribution to the author; no part may be used for commercial purposes directly or indirectly via a for-profit platform; no adaptation or derivative works are permitted without consent from the copyright owner.
- Language dc:language.iso
- en