Computer Vision–Based Estimation of Flood Depth in Flooded-Vehicle Images

Park, Somin; Baek, Francis; Sohn, Jiu; Kim, Hyoungkwan

doi:10.1061/(asce)cp.1943-5487.0000956

Cited by 31 publications

(8 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A survey of window and door heights in the test area may also be considered to obtain the most reliable ground truth heights for the comparison of model performance. In addition to the forecast flood water depth, water level estimations can also be extracted from geotagged social media images or surveillance camera data, based on people (Feng et al, 2020), vehicles (Park et al, 2021), or stop signs (Kharazi and Behzadan, 2021) in the flood water. This real-time information source has the potential to optimize the current forecastbased building flood risk mapping.…”

Section: Discussionmentioning

confidence: 99%

Determination of building flood risk maps from LiDAR mobile mapping data

Feng,

Xiao,

Brenner

et al. 2022

Preprint

View full text Add to dashboard Cite

With increasing urbanization, flooding is a major challenge for many cities today. Based on forecast precipitation, topography, and pipe networks, flood simulations can provide early warnings for areas and buildings at risk of flooding. Basement windows, doors, and underground garage entrances are common places where floodwater can flow into a building. Some buildings have been prepared or designed considering the threat of flooding, but others have not. Therefore, knowing the heights of these facade openings helps to identify places that are more susceptible to water ingress. However, such data is not yet readily available in most cities. Traditional surveying of the desired targets may be used, but this is a very time-consuming and laborious process. Instead, mobile mapping using LiDAR (light detection and ranging) is an efficient tool to obtain a large amount of high-density 3D measurement data. To use this method, it is required to extract the desired facade openings from the data in a fully automatic manner.This research presents a new process for the extraction of windows and doors from LiDAR mobile mapping data. Deep learning object detection models are trained to identify these objects. Usually, this requires to provide large amounts of manual annotations. In this paper, we mitigate this problem by leveraging a rule-based method. In a first step, the rule-based method is used to generate pseudo-labels. A semi-supervised learning strategy is then applied with three different levels of supervision. The results show that using only automatically generated pseudo-labels, the learning-based model outperforms the rule-based approach by 14.6% in terms of F 1 -score. After five hours of human supervision, it is possible to improve the model by another 6.2%.By comparing the detected facade openings' heights with the predicted water levels from a flood simulation model, a map can be produced which assigns per-building flood risk levels. Thus, our research provides a new geographic information layer for fine-grained urban emergency response. This information can be combined with flood forecasting to provide a more targeted disaster prevention guide for the city's infrastructure and residential buildings. To the best of our knowledge, this work is the first attempt to achieve such a large scale, fine-grained building flood risk mapping.

show abstract

Section: Discussionmentioning

confidence: 99%

Determination of building flood risk maps from LiDAR mobile mapping data

Feng,

Xiao,

Brenner

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Simple solutions for reflection removal are based on the analysis of multiple input images, including pairs of images that are taken from different orientations, or with different polarizations (Sarel and Irani, 2004;Kong et al, 2013). Table 1) A commonly used metric for describing the discrepancy in flood depth estimation is mean absolute error (MAE) (Chaudhary et al, 2019;Cohen et al, 2019;Park et al, 2021;Alizadeh and Behzadan, 2020;. As shown in Equation ( 1), the MAE for pole length estimation (𝑀𝐴𝐸 ! )…”

Section: Techniques To Overcome Challenging Casesmentioning

confidence: 99%

“…obtained an MAE of 4 in. by detecting submerged objects in social media images, and comparing them with their predefined sizes Park et al (2021). reported an MAE of 2.5 in.…”

mentioning

confidence: 99%

Crowdsourced-based deep convolutional networks for urban flood depth mapping

Bahareh¹,

Amir²

2022

Proceedings of the 2022 European Conference on Computing in Construction

View full text Add to dashboard Cite

Successful flood recovery and evacuation require access to reliable flood depth information. Most existing flood mapping tools do not provide real-time flood maps of inundated streets in and around residential areas. In this paper, a deep convolutional network is used to determine flood depth with high spatial resolution by analyzing crowdsourced images of submerged traffic signs. Testing the model on photos from a recent flood in the U.S. and Canada yields a mean absolute error of 6.978 in., which is on par with previous studies, thus demonstrating the applicability of this approach to lowcost, accurate, and real-time flood risk mapping.

show abstract

“…Arguably, this cross-domain perception gap can be addressed with CycleGAN by transferring the visual information in the two involved domains into an identical one. In civil engineering, researchers have explored GANs' capability to generate plausible new samples as a promising data augmentation technique for improving performance in structural defect detection (Gao et al, 2019;Maeda et al, 2020), construction resource management (Bang et al, 2020), and flood depth estimation (S. Park et al, 2021). However, CycleGAN has rarely been applied for style transfer in civil engineering contexts, with only a few exceptions (Hong et al, 2020;Pouyanfar et al, 2019).…”

Section: Cross-domain Style Transfermentioning

confidence: 99%

Indoor camera pose estimation via style‐transfer 3D models

Chen

Liu

et al. 2021

Computer aided Civil Eng

View full text Add to dashboard Cite

Many vision-based indoor localization methods require tedious and comprehensive pre-mapping of built environments. This research proposes a mapping-free approach to estimating indoor camera poses based on a 3D style-transferred building information model (BIM) and photogrammetry technique. To address the cross-domain gap between virtual 3D models and real-life photographs, a CycleGAN model was developed to transform BIM renderings into photorealistic images. A photogrammetry-based algorithm was developed to estimate camera pose using the visual and spatial information extracted from the styletransferred BIM. The experiments demonstrated the efficacy of CycleGAN in bridging the cross-domain gap, which significantly improved performance in terms of image retrieval and feature correspondence detection. With the 3D coordinates retrieved from BIM, the proposed method can achieve near real-time camera pose estimation with an accuracy of 1.38 m and 10.1 • in indoor environments.

show abstract

Computer Vision–Based Estimation of Flood Depth in Flooded-Vehicle Images

Cited by 31 publications

References 41 publications

Determination of building flood risk maps from LiDAR mobile mapping data

Determination of building flood risk maps from LiDAR mobile mapping data

Crowdsourced-based deep convolutional networks for urban flood depth mapping

Indoor camera pose estimation via style‐transfer 3D models

Contact Info

Product

Resources

About