Flood extent mapping in SAR images using semi-supervised approach

Floods pose a significant threat to both human populations and critical infrastructure. They are caused by excessive precipitation, snowmelt, or infrastructure failures. Precise mapping of flood levels is essential for directing emergency response, allocating resources as efficiently as possible, an...

Full description

Saved in:
Bibliographic Details
Main Authors: Girisha S, Savitha G, Sughosh P
Format: Article
Language:English
Published: Elsevier 2025-06-01
Series:Results in Engineering
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S259012302501374X
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850149413141348352
author Girisha S
Savitha G
Sughosh P
author_facet Girisha S
Savitha G
Sughosh P
author_sort Girisha S
collection DOAJ
description Floods pose a significant threat to both human populations and critical infrastructure. They are caused by excessive precipitation, snowmelt, or infrastructure failures. Precise mapping of flood levels is essential for directing emergency response, allocating resources as efficiently as possible, and estimating the degree of damage. Accurate semantic segmentation of flood-zones from SAR images is necessary for disaster management as it reduces human error and facilitates fast decision-making. However, the complexity of SAR images and the scarcity of annotated datasets make it difficult to develop efficient computer vision algorithms for this use. The paper employs consistency regularization and pseudo-label generation to overcome these issues and presents a novel hybrid semi-supervised semantic segmentation method. In particular, the study presents a student-teacher paradigm in which robust feature learning is ensured by training the teacher model through consistency regularization and self-supervised learning. An attention module-based tailored student model is developed to precisely identify flood-zones. Auxiliary decoders are also added throughout the training phase to reinforce consistency regularization and help the model capture extensive hierarchical characteristics. The suggested method was assessed with publicly accessible datasets, and the outcomes show notable gains in both qualitative and quantitative measures. The findings highlight the method's effectiveness in improving flood-zone segmentation accuracy, leading to more efficient disaster management.
format Article
id doaj-art-bbc97f7b5cfd451a9e0723d546efff69
institution OA Journals
issn 2590-1230
language English
publishDate 2025-06-01
publisher Elsevier
record_format Article
series Results in Engineering
spelling doaj-art-bbc97f7b5cfd451a9e0723d546efff692025-08-20T02:26:56ZengElsevierResults in Engineering2590-12302025-06-012610530410.1016/j.rineng.2025.105304Flood extent mapping in SAR images using semi-supervised approachGirisha S0Savitha G1Sughosh P2Department of Data Science and Computer Applications, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, 576104, Karnataka, IndiaDepartment of Data Science and Computer Applications, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, 576104, Karnataka, IndiaDepartment of Civil Engineering, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, 576104, Karnataka, India; Corresponding author.Floods pose a significant threat to both human populations and critical infrastructure. They are caused by excessive precipitation, snowmelt, or infrastructure failures. Precise mapping of flood levels is essential for directing emergency response, allocating resources as efficiently as possible, and estimating the degree of damage. Accurate semantic segmentation of flood-zones from SAR images is necessary for disaster management as it reduces human error and facilitates fast decision-making. However, the complexity of SAR images and the scarcity of annotated datasets make it difficult to develop efficient computer vision algorithms for this use. The paper employs consistency regularization and pseudo-label generation to overcome these issues and presents a novel hybrid semi-supervised semantic segmentation method. In particular, the study presents a student-teacher paradigm in which robust feature learning is ensured by training the teacher model through consistency regularization and self-supervised learning. An attention module-based tailored student model is developed to precisely identify flood-zones. Auxiliary decoders are also added throughout the training phase to reinforce consistency regularization and help the model capture extensive hierarchical characteristics. The suggested method was assessed with publicly accessible datasets, and the outcomes show notable gains in both qualitative and quantitative measures. The findings highlight the method's effectiveness in improving flood-zone segmentation accuracy, leading to more efficient disaster management.http://www.sciencedirect.com/science/article/pii/S259012302501374XSARSemantic segmentationDeep learningSemi-supervised learningConsistency regularizationPseudo-label generation
spellingShingle Girisha S
Savitha G
Sughosh P
Flood extent mapping in SAR images using semi-supervised approach
Results in Engineering
SAR
Semantic segmentation
Deep learning
Semi-supervised learning
Consistency regularization
Pseudo-label generation
title Flood extent mapping in SAR images using semi-supervised approach
title_full Flood extent mapping in SAR images using semi-supervised approach
title_fullStr Flood extent mapping in SAR images using semi-supervised approach
title_full_unstemmed Flood extent mapping in SAR images using semi-supervised approach
title_short Flood extent mapping in SAR images using semi-supervised approach
title_sort flood extent mapping in sar images using semi supervised approach
topic SAR
Semantic segmentation
Deep learning
Semi-supervised learning
Consistency regularization
Pseudo-label generation
url http://www.sciencedirect.com/science/article/pii/S259012302501374X
work_keys_str_mv AT girishas floodextentmappinginsarimagesusingsemisupervisedapproach
AT savithag floodextentmappinginsarimagesusingsemisupervisedapproach
AT sughoshp floodextentmappinginsarimagesusingsemisupervisedapproach