Beach wrack monitoring using image dataset with artificial intelligence

Yaoru Pan

doi:10.21996/tmcd-6x04

Beach wrack monitoring using image dataset with artificial intelligence

Yaoru Pan

Biologisk Institut

Publikation: Afhandling › Ph.d.-afhandling

Abstract

Som en del af produktionen eksporteret fra Blue Carbon-økosystemer (BCEs), skylles løsrevne makrofytter op på land og deponeres på kysten. Ud over at være habitat og føde til
fauna på stranden, betragtes opskyl som en af de vigtigste kulstofpuljer, der eksporteres fra BCEs, og spiller således en væsentlig rolle i BCE's kulstofkredsløb. Opskyl kan dog være en
kilde til drivhusgasudledning, når det nedbrydes og kan være til gene for gæster på stranden. Derfor er det afgørende at undersøge skæbnen af opskyl for at kvantificere den økologiske
værdi. Sammenlignet med undersøgelser af kulstofbegravelse i BCEs er det fortsat en udfordring at kvantificere eksporten af materiale fra disse. På grund af begrænsede tids- og rumlige opløsninger kan satellitbilleder opnået ved hjælp af traditionelle
overvågningsmetoder via remote-sensing (RS) hverken opfylde kravet om fleksibel dataindsamling, og heller ikke opfange detaljerede oplysninger om opskyl på stranden. Derfor undersøgte denne afhandling først potentialet af ubemandede luftfartøjer (UAVs) og kamerafælde til at bestemme aflejringsmønstre af opskyl (manuskript 1 og 2). Maskinlæringsmetoder (ML) blev anvendt til at klassificere luftfotos (klassificeringsnøjagtighed > 75 %, manuskript 1) og deep learning blev anvendt (DL) til at identificere opskyl fra forskellige scenarier optaget af kamerafælder (nøjagtighed > 60 %, 187 billeder analyseret på 5 minutter, manuskript 2). Studierne viser at ved anvendelse af kunstig intelligens (AI) i analyse af billeddatasæt, kan UAVs og kamerafælde udvikles til praktiske
værktøjer til overvågning af opskyl samt til forvaltning af kyster. Endvidere viste et in-situ nedbrydningseksperiment, at massetab og ændring af totalt fosfor (P) i opskyl var signifikant påvirket af sammensætning af plantematerialet og aflejringsscenarie (Manuskript 3). I fremtidige studier kan information om biomasse af opskyl indsamles ved brug af funktionelle sensorer monteret på UAVs og kamerafælde. Ved at kombinere biomassen af opskyl med aflejrings- og nedbrydningsmønstre kan bidrag fra opskyl til kulstof- og næringsstofkredsløb i kystsystemer estimeres.

Originalsprog	Engelsk
Bevilgende institution	Syddansk Universitet
Vejledere/rådgivere	Holmer, Marianne, Hovedvejleder Flindt, Mogens, Vejleder
Dato for forsvar	11. mar. 2022
Udgiver	Syddansk Universitet. Det Naturvidenskabelige Fakultet
DOI	https://doi.org/10.21996/tmcd-6x04
Status	Udgivet - 15. mar. 2022

Note vedr. afhandling

Afhandlingen kan læses på SDUs bibliotek.

Dokumenter og links

10.21996/tmcd-6x04

SDU link

Tjek adgangen til materialet i Syddansk Universitetsbiblioteks søgemaskine

2 Tidsskriftartikel

Beach Wrack Dynamics Using a Camera Trap as the Real-Time Monitoring Tool
Pan, Y., Ayoub, N., Schneider-Kamp, P., Flindt, M. & Holmer, M., jan. 2022, I: Frontiers in Marine Science. 9, 10 s., 813516.
Publikation: Bidrag til tidsskrift › Tidsskriftartikel › Forskning › peer review

Åben adgang
Fil
177 Downloads (Pure)
Beach wrack mapping using unmanned aerial vehicles for coastal environmental management
Pan, Y., Flindt, M., Schneider-Kamp, P. & Holmer, M., 1. nov. 2021, I: Ocean and Coastal Management. 213, 9 s., 105843.
Publikation: Bidrag til tidsskrift › Tidsskriftartikel › Forskning › peer review

Åben adgang
Fil
156 Downloads (Pure)

Citationsformater

@misc{8bf8ce430bca4b9b9a565124a1211929,

title = "Beach wrack monitoring using image dataset with artificial intelligence",

abstract = "As part of production exported from Blue Carbon ecosystems (BCEs), detached macrophytes are washed up onshore and deposited as beach wrack. In addition to providing habitat and food for coastal fauna, beach wrack is considered as one of the largest parts of carbon exported from BCEs, serving a significant role in carbon cycle of BCEs. However, beach wrack becomes a source of greenhouse gas emission during decomposition and is a nuisance for local residence. Hence, exploring the fate of beach wrack is essential for quantifying its ecological value. Compared to the investigation of carbon burial in BCEs, it remains a challenge to track this exported part. Due to limited time- and spatial- resolutions, satellite image acquired by traditional monitoring method of satellite remote sensing (RS) can neither meet the requirement for flexible data collection, nor can it capture detailed information of land objects at beach-scale. Therefore, this thesis firstly explored the potential of Unmanned Aerial Vehicles (UAVs) and camera trap to investigate the deposition pattern of beach wrack (Manuscript 1 and 2). The capabilities of machine learning (ML) to classify aerial images (overall classification accuracy > 75 \%, Manuscript 1) and deep learning (DL) to efficiently identify beach wrack from various beach scenes captured in camera images (accuracy > 60 \%, 187 images analyzed in 5 minutes, Manuscript 2) indicated that with application of artificial intelligence (AI) in analyzing image dataset, UAVs and camera trap can be developed into practical tools for beach wrack monitoring as well as for coastal environmental management. Furthermore, the in-situ decomposition experiment showed that mass loss and change of total phosphorus (P) of beach wrack were significantly affected by wrack species and deposition scenario (Manuscript 3). In future work, biomass information of beach wrack can be collected with more functional sensors mounted on UAVs and camera trap. By combining beach wrack biomass with its deposition and decomposition patterns, the contribution of beach wrack to carbon and nutrient cycles in coastal systems can be estimated.",

author = "Yaoru Pan",

year = "2022",

month = mar,

day = "15",

doi = "10.21996/tmcd-6x04",

language = "English",

publisher = "Syddansk Universitet. Det Naturvidenskabelige Fakultet",

address = "Denmark",

school = "SDU",

}

TY - GEN

T1 - Beach wrack monitoring using image dataset with artificial intelligence

AU - Pan, Yaoru

PY - 2022/3/15

Y1 - 2022/3/15

N2 - As part of production exported from Blue Carbon ecosystems (BCEs), detached macrophytes are washed up onshore and deposited as beach wrack. In addition to providing habitat and food for coastal fauna, beach wrack is considered as one of the largest parts of carbon exported from BCEs, serving a significant role in carbon cycle of BCEs. However, beach wrack becomes a source of greenhouse gas emission during decomposition and is a nuisance for local residence. Hence, exploring the fate of beach wrack is essential for quantifying its ecological value. Compared to the investigation of carbon burial in BCEs, it remains a challenge to track this exported part. Due to limited time- and spatial- resolutions, satellite image acquired by traditional monitoring method of satellite remote sensing (RS) can neither meet the requirement for flexible data collection, nor can it capture detailed information of land objects at beach-scale. Therefore, this thesis firstly explored the potential of Unmanned Aerial Vehicles (UAVs) and camera trap to investigate the deposition pattern of beach wrack (Manuscript 1 and 2). The capabilities of machine learning (ML) to classify aerial images (overall classification accuracy > 75 %, Manuscript 1) and deep learning (DL) to efficiently identify beach wrack from various beach scenes captured in camera images (accuracy > 60 %, 187 images analyzed in 5 minutes, Manuscript 2) indicated that with application of artificial intelligence (AI) in analyzing image dataset, UAVs and camera trap can be developed into practical tools for beach wrack monitoring as well as for coastal environmental management. Furthermore, the in-situ decomposition experiment showed that mass loss and change of total phosphorus (P) of beach wrack were significantly affected by wrack species and deposition scenario (Manuscript 3). In future work, biomass information of beach wrack can be collected with more functional sensors mounted on UAVs and camera trap. By combining beach wrack biomass with its deposition and decomposition patterns, the contribution of beach wrack to carbon and nutrient cycles in coastal systems can be estimated.

AB - As part of production exported from Blue Carbon ecosystems (BCEs), detached macrophytes are washed up onshore and deposited as beach wrack. In addition to providing habitat and food for coastal fauna, beach wrack is considered as one of the largest parts of carbon exported from BCEs, serving a significant role in carbon cycle of BCEs. However, beach wrack becomes a source of greenhouse gas emission during decomposition and is a nuisance for local residence. Hence, exploring the fate of beach wrack is essential for quantifying its ecological value. Compared to the investigation of carbon burial in BCEs, it remains a challenge to track this exported part. Due to limited time- and spatial- resolutions, satellite image acquired by traditional monitoring method of satellite remote sensing (RS) can neither meet the requirement for flexible data collection, nor can it capture detailed information of land objects at beach-scale. Therefore, this thesis firstly explored the potential of Unmanned Aerial Vehicles (UAVs) and camera trap to investigate the deposition pattern of beach wrack (Manuscript 1 and 2). The capabilities of machine learning (ML) to classify aerial images (overall classification accuracy > 75 %, Manuscript 1) and deep learning (DL) to efficiently identify beach wrack from various beach scenes captured in camera images (accuracy > 60 %, 187 images analyzed in 5 minutes, Manuscript 2) indicated that with application of artificial intelligence (AI) in analyzing image dataset, UAVs and camera trap can be developed into practical tools for beach wrack monitoring as well as for coastal environmental management. Furthermore, the in-situ decomposition experiment showed that mass loss and change of total phosphorus (P) of beach wrack were significantly affected by wrack species and deposition scenario (Manuscript 3). In future work, biomass information of beach wrack can be collected with more functional sensors mounted on UAVs and camera trap. By combining beach wrack biomass with its deposition and decomposition patterns, the contribution of beach wrack to carbon and nutrient cycles in coastal systems can be estimated.

U2 - 10.21996/tmcd-6x04

DO - 10.21996/tmcd-6x04

M3 - Ph.D. thesis

PB - Syddansk Universitet. Det Naturvidenskabelige Fakultet

ER -

Beach wrack monitoring using image dataset with artificial intelligence

Abstract

Note vedr. afhandling

Dokumenter og links

SDU link

Fingeraftryk

Relaterede publikationer

Beach Wrack Dynamics Using a Camera Trap as the Real-Time Monitoring Tool

Beach wrack mapping using unmanned aerial vehicles for coastal environmental management

Citationsformater