View in EDS

A hybrid deep learning framework combining transformer and logistic regression models for automatic marine mucilage detection using sentinel-1 SAR data: A case study in Armutlu-Zeytinbağı, Marmara Sea.

Saved in:
Bibliographic Details
Title: A hybrid deep learning framework combining transformer and logistic regression models for automatic marine mucilage detection using sentinel-1 SAR data: A case study in Armutlu-Zeytinbağı, Marmara Sea.
Authors: Bakis, Enes, Acar, Emrullah, Yilmaz, Musa
Source: PLoS ONE; 9/25/2025, Vol. 20 Issue 9, p1-30, 30p
Subject Terms: MUCILAGE, DEEP learning, DETECTION algorithms, BODIES of water, REMOTE sensing, LOGISTIC regression analysis, IMAGE analysis, TRANSFORMER models
Geographic Terms: SEA of Marmara (Turkey)
Abstract: The identification of various objects and species found in nature is of great importance today. Active and passive imaging systems are in a beneficial position in this direction, both in terms of cost and convenience. Recently, mucilage events in our country pose a great risk for both marine life and human life. In this study, water areas in one of the regions affected by the mucilage event that occurred in May 2021 were chosen as the object. The region between Armutlu-Zeytinbağı in the Marmara Sea was chosen as the study area. 1300 samples were selected from the mucilage region and recorded with the help of GPS. After these selected samples were chosen as mucilage area for 17 May–22 May and as a clean area for 21 June-22 June (2600 samples in total), image analyses were made using time series with the help of Sentinel-1 satellite images. These image analyses were performed using Sentinel-1 band parameters (VV-VH). A unique data set was created by recording the numerical data showing the backscattering values of the VV-VH polarization band images. It is aimed to automatically detect the mucilage area by applying deep learning and machine learning to the obtained data set. It has been observed that the accuracies of our applied hybrid (Transformer Method + Logistic Regression), deep learning (RNN, CNN) and machine learning models (Decision Tree, Naive Bayes, SVM) are high (96%−100%). With the applied deep learning and machine learning methods, it is thought that regions can be detected more easily and intervened early in these regions. [ABSTRACT FROM AUTHOR]
Copyright of PLoS ONE is the property of Public Library of Science and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Complementary Index
Description
Abstract:The identification of various objects and species found in nature is of great importance today. Active and passive imaging systems are in a beneficial position in this direction, both in terms of cost and convenience. Recently, mucilage events in our country pose a great risk for both marine life and human life. In this study, water areas in one of the regions affected by the mucilage event that occurred in May 2021 were chosen as the object. The region between Armutlu-Zeytinbağı in the Marmara Sea was chosen as the study area. 1300 samples were selected from the mucilage region and recorded with the help of GPS. After these selected samples were chosen as mucilage area for 17 May–22 May and as a clean area for 21 June-22 June (2600 samples in total), image analyses were made using time series with the help of Sentinel-1 satellite images. These image analyses were performed using Sentinel-1 band parameters (VV-VH). A unique data set was created by recording the numerical data showing the backscattering values of the VV-VH polarization band images. It is aimed to automatically detect the mucilage area by applying deep learning and machine learning to the obtained data set. It has been observed that the accuracies of our applied hybrid (Transformer Method + Logistic Regression), deep learning (RNN, CNN) and machine learning models (Decision Tree, Naive Bayes, SVM) are high (96%−100%). With the applied deep learning and machine learning methods, it is thought that regions can be detected more easily and intervened early in these regions. [ABSTRACT FROM AUTHOR]
ISSN:19326203
DOI:10.1371/journal.pone.0330721