SEA-LAND SEGMENTATION MODELS IN DEEP LEARNING FROM REMOTE SENSING DATA DOI
R. Okhrimchuk, V. Demidov, Kateryna SLIUSAR

et al.

Visnyk of Taras Shevchenko National University of Kyiv Geology, Journal Year: 2024, Volume and Issue: 4 (107), P. 122 - 130

Published: Jan. 1, 2024

Background. Coastline changes can have a significant impact on coastal landscape, ecosystems and communities. Therefore, monitoring of such highly dynamic system as sea-land is an urgent task that be solved both by traditional methods using depth learning techniques to improve the efficiency processing class tasks. The object authors' research coastline along coast western part Crimean Peninsula, study which has become impossible due temporary occupation Peninsula since 2014. paper considers main indicators digitization. types satellite images well their combinations are compared for effective utilization shoreline mapping task. Many used recognize extract shorelines in images, generally divided into three groups: indexing, edge detection classification methods. Methods. Authors models efficiently its boundaries include ISODATA (Iterative Self-Organizing Data Analysis Technique), Maximum Likelihood Estimation (MLE), Random Forest (RF), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), U-Net, Segment Anything Model (SAM). Results. outlines were obtained basis PlanetScope ISODATA, MLE, RF, KNN, SVM, SAM performance compared. included development Python code automatically generate reports including information five evaluation metrics, accuracy (98.96), recall (99.45), precision (97.27), F1-score (98.34), IoU (96.74), facilitated different approaches Conclusions. comparative analysis highlights advantage U-Net model extraction from remotely sensed images. consistently provides most accurate detailed segmentation scenarios, demonstrating robustness accuracy.

Language: Английский

Using GIS tools to enhance the shape of coastline extracted from Sentinel-2 satellite images DOI

Emanuele Alcaras,

Ugo Falchi, Claudio Parente

et al.

Published: May 16, 2024

Sentinel-2 images are widely used for coastline extraction. One of the most widespread methods is Normalized Difference Water Index (NDWI), which permits distinguishing between water and non-water pixels. The result obtained, since it derives from satellite images, preserves shape pixels, often labeled as unrealistic. For this reason, Geographic Information Systems (GIS) tools in literature to simplify and/or smooth obtained order make similar reality possible. However, these operations not always optimal. In work, we analyze extracted concerning island Giglio (Italy), particular four coastlines compared: standard coastline, i.e. one directly NDWI; resulting smoothing application; simplification finally, both application. results show a higher efficiency compared simplification.

Language: Английский

Citations

1
Multi Network Algorithm for Coastal Line Segmentation in Remote Sensing Images DOI
Xuemei Li, Xing Wang, Huping Ye

et al.

IEEE Transactions on Geoscience and Remote Sensing, Journal Year: 2024, Volume and Issue: 62, P. 1 - 12

Published: Jan. 1, 2024

Language: Английский

Citations

1
Integrated water quality assessment of open water bodies using empirical equations and remote sensing techniques in Bangweulu Wetland lakes, Zambia DOI
Misheck Lesa Chundu, Kawawa Banda,

Henry M. Sichingabula

et al.

Journal of Great Lakes Research, Journal Year: 2024, Volume and Issue: unknown, P. 102451 - 102451

Published: Oct. 1, 2024

Language: Английский

Citations

1
Monitoring Coastal Evolution and Geomorphological Processes Using Time-Series Remote Sensing and Geospatial Analysis: Application Between Cape Serrat and Kef Abbed, Northern Tunisia DOI Creative Commons
Zeineb Kassouk, Emna Ayari, Benoı̂t Deffontaines

et al.

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(20), P. 3895 - 3895

Published: Oct. 19, 2024

The monitoring of coastal evolution (coastline and associated geomorphological features) caused by episodic persistent processes with climatic anthropic activities is required for management decisions. availability open access, remotely sensed data increasing spatial, temporal, spectral resolutions, promising in this context. coastline Northern Tunisia currently showing geomorphic process, such as erosion lateral sedimentation. This study aims to investigate the potential time-series optical data, namely Landsat (from 1985–2019) Google Earth® satellite imagery 2007 2023), analyze shoreline changes morphosedimentary between Cape Serrat Kef Abbed, Tunisia. Digital Shoreline Analysis System (DSAS) was used quantify multitemporal rates using two metrics: net movement (NSM) end-point rate (EPR). Erosion observed around tombolo near river mouths, exacerbated presence surrounding dams, where NSM up −8.31 m/year. Despite a total −15 m, seasonal dynamics revealed maximum winter (71% negative NSM) accretion spring (57% positive NSM). effects currents, winds, dams on dune were studied historical images Earth®. In period from 1994 2023, area marked face retreat removal more than 40% site, erosion. At finer spatial resolution according synergy field observations photointerpretation, four key shaping identified: wave/tide action, wind transport, pedogenesis, deposition. Given frequent areas, method facilitates maintenance updating databases, which are essential analyzing impacts sea level rise southern Mediterranean region. Furthermore, developed approach could be implemented range forecast scenarios simulate higher future sea-level enhanced climate change.

Language: Английский

Citations

1
Twin Satellites HY-1C/D Reveal the Local Details of Astronomical Tide Flooding into the Qiantang River, China DOI Creative Commons

Lina Cai,

Hengpan Zhang,

Xiaomin Ye

et al.

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(9), P. 1507 - 1507

Published: April 24, 2024

This article extracts the Qiantang River tidal bore, analyzing water environment characteristics in front of line bore and behind it. The Index (QRI) was established using HY-1C, HY-1D, Gao Fen-1 wide field-of-view (GF-1 WFV) satellite data to precisely determine location details bore. Comparative analyses changes on two sides were conducted. results indicate following: (1) QRI enhances visibility lines, accentuating their contrast with surrounding river water, resulting a more vivid character. proves be an effective extraction method, potential applicability similar lines different regions. (2) Observable roughness occur at location, smoother surface textures observed compared those There is discernible increase suspended sediment concentration (SSC) as passes through. (3) study reveals mechanism change induced by emphasizing its significance promoting vertical body exchange well scouring bottom sediments. effect increases SSC roughness.

Language: Английский

Citations

0
Analysis of Changes in Aquaculture Ponds in Liaodong Bay from 1985 to 2020 DOI
Dongyang Zhao, Yiping Li, Wei Li

et al.

Published: Jan. 1, 2024

Language: Английский

Citations

0
Event-driven nearshore and shoreline coastline detection on SpiNNaker neuromorphic hardware DOI Creative Commons
Mazdak Fatahi, Pierre Boulet, Giulia D’Angelo

et al.

Neuromorphic Computing and Engineering, Journal Year: 2024, Volume and Issue: 4(3), P. 034012 - 034012

Published: Sept. 1, 2024

Abstract Coastline detection is vital for coastal management, involving frequent observation and assessment to understand dynamics inform decisions on environmental protection. Continuous streaming of high-resolution images demands robust data processing storage solutions manage large datasets efficiently, posing challenges that require innovative real-time analysis meaningful insights extraction. This work leverages low-latency event-based vision sensors coupled with neuromorphic hardware in an attempt decrease a two-fold challenge, reducing the computational burden ∼0.375 mW whilst obtaining coastline map as little 20 ms. The proposed Spiking Neural Network runs SpiNNaker platform using total 18 040 neurons reaching 98.33% accuracy. model has been characterised evaluated by computing accuracy Intersection over Union scores ground truth real-world dataset across different time windows. system’s robustness was further assessed evaluating its ability avoid non-coastline profiles funny shapes, achieving success rate 97.3%.

Language: Английский

Citations

0
Regional-Scale Image Segmentation of Sandy Beaches in Southeastern Australia DOI Creative Commons
Suk Yee Yong, Julian O’Grady, Rebecca Gregory

et al.

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(18), P. 3534 - 3534

Published: Sept. 23, 2024

Beaches play a crucial role in recreation and ecosystem habitats, are central to Australia’s national identity. Precise mapping of beach locations is essential for coastal vulnerability risk assessments. While point over 11,000 beaches documented from citizen science projects, the full spatial extent outlines many Australian remain unmapped. This study leverages deep learning (DL), specifically convolutional neural networks, binary image segmentation map along coast Southeastern Australia. It focuses on Victoria New South Wales coasts, each approximately 2000 2500 km length. Our methodology includes training evaluating model using state-specific datasets, followed by applying trained predict outlines, size, shape, morphology both regions. The results demonstrate model’s ability generate accurate rapid predictions, although it faces challenges such as misclassifying cliffs sensitivity fine details. Overall, this research presents significant advancement integrating DL with science, providing scalable solution efforts comprehensive support sustainable management conservation across Open access datasets models provided further around

Language: Английский

Citations

0
SEA-LAND SEGMENTATION MODELS IN DEEP LEARNING FROM REMOTE SENSING DATA DOI
R. Okhrimchuk, V. Demidov, Kateryna SLIUSAR

et al.

Visnyk of Taras Shevchenko National University of Kyiv Geology, Journal Year: 2024, Volume and Issue: 4 (107), P. 122 - 130

Published: Jan. 1, 2024

Background. Coastline changes can have a significant impact on coastal landscape, ecosystems and communities. Therefore, monitoring of such highly dynamic system as sea-land is an urgent task that be solved both by traditional methods using depth learning techniques to improve the efficiency processing class tasks. The object authors' research coastline along coast western part Crimean Peninsula, study which has become impossible due temporary occupation Peninsula since 2014. paper considers main indicators digitization. types satellite images well their combinations are compared for effective utilization shoreline mapping task. Many used recognize extract shorelines in images, generally divided into three groups: indexing, edge detection classification methods. Methods. Authors models efficiently its boundaries include ISODATA (Iterative Self-Organizing Data Analysis Technique), Maximum Likelihood Estimation (MLE), Random Forest (RF), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), U-Net, Segment Anything Model (SAM). Results. outlines were obtained basis PlanetScope ISODATA, MLE, RF, KNN, SVM, SAM performance compared. included development Python code automatically generate reports including information five evaluation metrics, accuracy (98.96), recall (99.45), precision (97.27), F1-score (98.34), IoU (96.74), facilitated different approaches Conclusions. comparative analysis highlights advantage U-Net model extraction from remotely sensed images. consistently provides most accurate detailed segmentation scenarios, demonstrating robustness accuracy.

Language: Английский

Citations

0