Coverage estimation of benthic habitat features by semantic segmentation of underwater imagery from South-eastern Baltic reefs using deep learning models

Underwater imagery (UI) is an important and sometimes the only tool for mapping hard-bottom habitats. With the development of new camera systems, from hand-held or simple “drop-down” cameras to ROV/AUV-mounted video systems, video data collection has increased considerably. However, the processing a...

Full description

Saved in:
Bibliographic Details
Main Authors: Andrius Šiaulys, Evaldas Vaičiukynas, Saulė Medelytė, Kazimieras Buškus
Format: Article
Language:English
Published: Institute of Oceanology of the Polish Academy of Sciences 2024-04-01
Series:Oceanologia
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S0078323423000933
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849305090551185408
author Andrius Šiaulys
Evaldas Vaičiukynas
Saulė Medelytė
Kazimieras Buškus
author_facet Andrius Šiaulys
Evaldas Vaičiukynas
Saulė Medelytė
Kazimieras Buškus
author_sort Andrius Šiaulys
collection DOAJ
description Underwater imagery (UI) is an important and sometimes the only tool for mapping hard-bottom habitats. With the development of new camera systems, from hand-held or simple “drop-down” cameras to ROV/AUV-mounted video systems, video data collection has increased considerably. However, the processing and analysing of vast amounts of imagery can become very labour-intensive, thus making it ineffective both time-wise and financially. This task could be simplified if the processes or their intermediate steps could be done automatically. Luckily, the rise of AI applications for automatic image analysis tasks in the last decade has empowered researchers with robust and effective tools. In this study, two ways to make UI analysis more efficient were tested with eight dominant visual features of the Southeastern Baltic reefs: 1) the simplification of video processing and expert annotation efforts by skipping the video mosaicking step and reducing the number of frames analysed; 2) the application of semantic segmentation of UI using deep learning models. The results showed that the annotation of individual frames provides similar results compared to 2D mosaics; moreover, the reduction of frames by 2–3 times resulted in only minor differences from the baseline. Semantic segmentation using the PSPNet model as the deep learning architecture was extensively evaluated, applying three variants of validation. The accuracy of segmentation, as measured by the intersection-over-union, was mediocre; however, estimates of visual coverage percentages were fair: the difference between the expert annotations and model-predicted segmentation was less than 6–8%, which could be considered an encouraging result.
format Article
id doaj-art-0dd90b4cf92a4239a673223d0b8d509f
institution Kabale University
issn 0078-3234
language English
publishDate 2024-04-01
publisher Institute of Oceanology of the Polish Academy of Sciences
record_format Article
series Oceanologia
spelling doaj-art-0dd90b4cf92a4239a673223d0b8d509f2025-08-20T03:55:33ZengInstitute of Oceanology of the Polish Academy of SciencesOceanologia0078-32342024-04-0166228629810.1016/j.oceano.2023.12.004Coverage estimation of benthic habitat features by semantic segmentation of underwater imagery from South-eastern Baltic reefs using deep learning modelsAndrius Šiaulys0Evaldas Vaičiukynas1Saulė Medelytė2Kazimieras Buškus3Marine Research Institute, Klaipeda University, Klaipeda, Lithuania; Corresponding author at: Marine Research Institute, Klaipeda University, Klaipeda, Lithuania.Faculty of Informatics, Kaunas University of Technology, Kaunas, LithuaniaMarine Research Institute, Klaipeda University, Klaipeda, LithuaniaFaculty of Informatics, Kaunas University of Technology, Kaunas, LithuaniaUnderwater imagery (UI) is an important and sometimes the only tool for mapping hard-bottom habitats. With the development of new camera systems, from hand-held or simple “drop-down” cameras to ROV/AUV-mounted video systems, video data collection has increased considerably. However, the processing and analysing of vast amounts of imagery can become very labour-intensive, thus making it ineffective both time-wise and financially. This task could be simplified if the processes or their intermediate steps could be done automatically. Luckily, the rise of AI applications for automatic image analysis tasks in the last decade has empowered researchers with robust and effective tools. In this study, two ways to make UI analysis more efficient were tested with eight dominant visual features of the Southeastern Baltic reefs: 1) the simplification of video processing and expert annotation efforts by skipping the video mosaicking step and reducing the number of frames analysed; 2) the application of semantic segmentation of UI using deep learning models. The results showed that the annotation of individual frames provides similar results compared to 2D mosaics; moreover, the reduction of frames by 2–3 times resulted in only minor differences from the baseline. Semantic segmentation using the PSPNet model as the deep learning architecture was extensively evaluated, applying three variants of validation. The accuracy of segmentation, as measured by the intersection-over-union, was mediocre; however, estimates of visual coverage percentages were fair: the difference between the expert annotations and model-predicted segmentation was less than 6–8%, which could be considered an encouraging result.http://www.sciencedirect.com/science/article/pii/S0078323423000933Underwater videoEpibenthosMosaickingSegmentationComputer vision, PSPNetResNet
spellingShingle Andrius Šiaulys
Evaldas Vaičiukynas
Saulė Medelytė
Kazimieras Buškus
Coverage estimation of benthic habitat features by semantic segmentation of underwater imagery from South-eastern Baltic reefs using deep learning models
Oceanologia
Underwater video
Epibenthos
Mosaicking
Segmentation
Computer vision, PSPNet
ResNet
title Coverage estimation of benthic habitat features by semantic segmentation of underwater imagery from South-eastern Baltic reefs using deep learning models
title_full Coverage estimation of benthic habitat features by semantic segmentation of underwater imagery from South-eastern Baltic reefs using deep learning models
title_fullStr Coverage estimation of benthic habitat features by semantic segmentation of underwater imagery from South-eastern Baltic reefs using deep learning models
title_full_unstemmed Coverage estimation of benthic habitat features by semantic segmentation of underwater imagery from South-eastern Baltic reefs using deep learning models
title_short Coverage estimation of benthic habitat features by semantic segmentation of underwater imagery from South-eastern Baltic reefs using deep learning models
title_sort coverage estimation of benthic habitat features by semantic segmentation of underwater imagery from south eastern baltic reefs using deep learning models
topic Underwater video
Epibenthos
Mosaicking
Segmentation
Computer vision, PSPNet
ResNet
url http://www.sciencedirect.com/science/article/pii/S0078323423000933
work_keys_str_mv AT andriussiaulys coverageestimationofbenthichabitatfeaturesbysemanticsegmentationofunderwaterimageryfromsoutheasternbalticreefsusingdeeplearningmodels
AT evaldasvaiciukynas coverageestimationofbenthichabitatfeaturesbysemanticsegmentationofunderwaterimageryfromsoutheasternbalticreefsusingdeeplearningmodels
AT saulemedelyte coverageestimationofbenthichabitatfeaturesbysemanticsegmentationofunderwaterimageryfromsoutheasternbalticreefsusingdeeplearningmodels
AT kazimierasbuskus coverageestimationofbenthichabitatfeaturesbysemanticsegmentationofunderwaterimageryfromsoutheasternbalticreefsusingdeeplearningmodels