Doctor of Philosophy
In recent years, the frequency and intensity of natural disasters have increased significantly due to global climate change. Remote sensing (RS) combined with cutting-edge artificial intelligence (AI) techniques, such as deep learning (DL), has been proven effective in rapidly acquiring ground information in disaster-related applications. However, the lack of annotated data limits the usability of DL in disaster scenarios. This thesis investigates the utilization of the latest DL algorithms based on limited labels for RS-based disaster tasks using high-resolution optical images, aiming to enhance the applicability of DL in real-world scenarios.
Firstly, given the advancements of semi-supervised learning (SSL) algorithms that leverage a mass of unlabeled data, a consistency regularization (CR)-based SSL framework is developed for RS image semantic segmentation. Encouragingly, based on five datasets with diverse tasks, e.g., road extraction, building detection, and land cover classification, the proposed SSL method using only 5% labeled data achieves a relative accuracy ratio over 89% compared to the fully supervised learning method using 100% labeled samples.
Secondly, to extract floodwater rapidly and accurately in urban areas with dense shadows, a modified fully convolutional network model is designed and integrated with a novel SSL framework incorporating CR, RandMix, and test-time augmentation techniques. Experiments on aerial images of the 2013 Calgary flood demonstrate that the presented method achieves an impressive F1-score of 96.34% for flood mapping, utilizing only 4.47% of the total labeled data.
Thirdly, to meet the urgent need for timely and accurate building damage assessment, a novel SSL framework is proposed, combining multitask semantic segmentation with a perturbed dual mean teachers’ scheme. Experiments on three datasets indicate that even with a small fraction of labeled samples (e.g., 5%), the proposed method can generate meaningful results for building damage assessment, offering a potential solution for timely disaster response in emergencies.
Lastly, this thesis presents an operational workflow for rapid urban flood mapping, including a novel weak training data generation strategy and an end-to-end weakly supervised learning (WSL) framework with structure constraints and cross self-distillation. The proposed workflow better balances timeliness and accuracy in flood mapping, exhibiting promising operability in response to urban floodings.
Summary for Lay Audience
During the past few decades, due to the growing severity of global climate change, the cost-effective and accurate extraction of ground information from high-resolution remote sensing (RS) data has become increasingly vital in various disaster-related applications. Unlike traditional machine learning techniques relying on manual feature design based on specialized knowledge, data-driven deep learning (DL) methods excel at automatically learning features and patterns from large labeling datasets. However, studies have shown that the heterogeneity of multi-source, multi-temporal, and multi-modal RS data presents challenges in effectively generalizing DL models to different locations. Additionally, RS images contain more complex scenes and finer details than natural images, making creating high-quality pixel-wise annotated RS data an extraordinarily time-consuming and expensive process. Therefore, obtaining labeled data has become a major impeding factor for applying DL in time-sensitive disaster response tasks.
To enhance the applicability of DL algorithms in disaster scenarios with limited labels, this thesis primarily focuses on exploring the innovative application of semi-supervised and weakly supervised learning methods in disaster-related RS tasks. Specifically, semi-supervised learning (SSL) uncovers hidden patterns from large amounts of unlabeled data, and weakly supervised learning (WSL) concentrates on training with weak supervision: 1) incomplete supervision, where only a subset of training data is sparsely labeled; 2) inexact supervision, where the given labels are only coarse labeled; and 3) inaccurate supervision, where the given labels may be incorrect. The results of this study indicate that utilizing SSL and WSL can improve the model's performance based on limited labeled data, thereby enhancing disaster response efforts. These advanced techniques contribute to better disaster management and strengthen our readiness to tackle natural disasters exacerbated by climate change by making the most of available data resources and broadening the scope of applicability. This research offers promising insights into optimizing the use of RS data and empowering decision-makers to respond effectively to future disaster events.
He, Yongjun, "Advancing Intelligent Interpretation of Remote Sensing Imagery for Disaster-related Applications Using Deep Learning with Limited Labels" (2023). Electronic Thesis and Dissertation Repository. 9703.
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License
Available for download on Saturday, September 06, 2025