Building Damage Assessment From Post-Hurricane Imageries Using Unsupervised Domain Adaptation With Enhanced Feature Discrimination

Chen, Lin; Li, Yundong; Liu, Yi; Wang, Xiang; Geng, Shuo

doi:10.1109/tgrs.2021.3054869

Cited by 15 publications

(8 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…They first completed pre-training on the labeled source domain, aligning the features in the source and target domains by maximum mean discrepancy (MMD), and later transferred to two new types of hurricane data to complete the classification. They [10] later proposed a new generative adversarial network to align the source and target domains in an unsupervised approach to achieve better results on both transferred tasks.…”

Section: Related Workmentioning

confidence: 99%

“…Some scholars take the difficulty of obtaining large amounts of labeled data of disaster into consideration, and disaster types and datasets require different labeling methods. They adopted a semi-supervised or unsupervised approach, using only small amounts of labeled or/and unlabeled data for research [6,7,8,9,10].…”

Section: Introductionmentioning

confidence: 99%

“…First, existing semi-supervised learning requires a certain amount of labeled data and does not fully address the problem of the complex labeling of disaster datasets [6,7,8]. Second, current unsupervised learning generally relies on pretraining on labeled data [9,10], which has not good enough for portability. Third, the interpretability of deep learning is poor, so it cannot be reliably used in our daily life.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Self-Supervised Learning for Building Damage Assessment from Large-scale xBD Satellite Imagery Benchmark Datasets

Xia¹,

Li²,

Bai³

et al. 2022

Preprint

View full text Add to dashboard Cite

In the field of post-disaster assessment, for timely and accurate rescue and localization after a disaster, people need to know the location of damaged buildings. In deep learning, some scholars have proposed methods to make automatic and highly accurate building damage assessments by remote sensing images, which are proved to be more efficient than assessment by domain experts. However, due to the lack of a large amount of labeled data, these kinds of tasks can suffer from being able to do an accurate assessment, as the efficiency of deep learning models relies highly on labeled data. Although existing semi-supervised and unsupervised studies have made breakthroughs in this area, none of them has completely solved this problem. Therefore, we propose adopting a self-supervised comparative learning approach to address the task without the requirement of labeled data. We constructed a novel asymmetric twin network architecture and tested its performance on the xBD dataset. Experiment results of our model show the improvement compared to baseline and commonly used methods. We also demonstrated the potential of self-supervised methods for building damage recognition awareness.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Self-Supervised Learning for Building Damage Assessment from Large-scale xBD Satellite Imagery Benchmark Datasets

Xia¹,

Li²,

Bai³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…In general, CCFs manifest when multiple components of a system fail due to a shared underlying cause [17,18]. To illustrate, consider a scenario where an extreme weather event, such as a hurricane or flood, causes widespread power outages across a city [6,19,20]. In this case, the loss of electricity could lead to the simultaneous failure of multiple EMS components, such as communication networks, resulting in a collective system failure.…”

Section: Introductionmentioning

confidence: 99%

Reliability and Detectability of Emergency Management Systems in Smart Cities under Common Cause Failures

Jesus,

Portugal,

Costa

et al. 2024

Sensors

View full text Add to dashboard Cite

Urban areas are undergoing significant changes with the rise of smart cities, with technology transforming how cities develop through enhanced connectivity and data-driven services. However, these advancements also bring new challenges, especially in dealing with urban emergencies that can disrupt city life and infrastructure. The emergency management systems have become crucial elements for enabling cities to better handle urban emergencies, although ensuring the reliability and detectability of such system remains critical. This article introduces a new method to perform reliability and detectability assessments. By using Fault Tree Markov chain models, this article evaluates their performance under extreme conditions, providing valuable insights for designing and operating urban emergency systems. These analyses fill a gap in the existing research, offering a comprehensive understanding of emergency management systems functionality in complex urban settings.

show abstract

“…In recent decades, various natural disasters are frequently occurring all over the world. Earthquakes, floods, hurricanes, and other extremely destructive natural disasters not only cause massive property losses but also threaten the safety of human lives 1 – 3 . In addition to improving detection and early warning capability before a disaster occurs, it is critical to obtain disaster information quickly after a disaster occurs 4 .…”

Section: Introductionmentioning

confidence: 99%

Post-disaster building damage assessment based on improved U-Net

Deng

Wang

2022

Sci Rep

View full text Add to dashboard Cite

When a severe natural disaster occurs, the extraction of post-disaster building damage information is one of the methods to quickly obtain disaster information. The increasingly mature high-resolution remote sensing technology provides a solid foundation for obtaining information about building damage. To address the issues with inaccurate building positioning in existing building damage assessment methods, as well as poor classification due to similar minor and major damage characteristics in building damage classification. Based on U-Net, we designed a two-stage building damage assessment network. The first stage is an independent U-Net focused on building segmentation, followed by a Siamese U-Net focused on building damage classification. The Extra Skip Connection and Asymmetric Convolution Block were used for enhancing the network's ability to segment buildings on different scales; Shuffle Attention directed the network's attention to the correlation of buildings before and after the disaster. The xBD dataset was used for training and testing in the study, and the overall performance was evaluated using a balanced F-score (F1). The improved network had an F1 of 0.8741 for localization and F1 of 0.7536 for classification. When compared to other methods, it achieved better overall performance for building damage assessment and was able to generalize to multiple disasters.

show abstract

Building Damage Assessment From Post-Hurricane Imageries Using Unsupervised Domain Adaptation With Enhanced Feature Discrimination

Cited by 15 publications

References 39 publications

Self-Supervised Learning for Building Damage Assessment from Large-scale xBD Satellite Imagery Benchmark Datasets

Self-Supervised Learning for Building Damage Assessment from Large-scale xBD Satellite Imagery Benchmark Datasets

Reliability and Detectability of Emergency Management Systems in Smart Cities under Common Cause Failures

Post-disaster building damage assessment based on improved U-Net

Contact Info

Product

Resources

About