Enhancing species video detection capabilities at the obsea observatory through the integration of emuas cameras within the aneris project framework

High-resolution images captured by EMUAS cameras, equipped with a 4K sensor and set to 1440p resolution for the OBSEA deployment, are analysed using the YOLO (You Only Look Once) object detection algorithm, trained with labelled datasets from OBSEA. The cameras operate at 20 frames per second (fps)...

Descripción completa

Detalles Bibliográficos
Autores: Prat i Bayarri, Oriol|||0009-0009-9975-9343, Baños Castelló, Pol|||0000-0003-0780-3255, Carandell Widmer, Matias|||0000-0003-0559-4453, Martínez, Enoc, Mihai Toma, Daniel, Rambech, Alexander, Kaba, Christopher, Alcocer, Alex, Río Fernández, Joaquín del|||0000-0002-6191-2201
Tipo de recurso: artículo
Fecha de publicación:2025
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/452096
Acceso en línea:https://hdl.handle.net/2117/452096
https://dx.doi.org/10.5821/iwp.2025.24.13985
Access Level:acceso abierto
Palabra clave:Oceanography -- Research
Oceanography -- Equipment and supplies
Digital cameras
High-resolution cameras
EMUAS
Biofouling
UV-C
AI algorithms
YOLO
Object detection
Oceanografia -- Investigació
Oceanografia -- Aparells i instruments
Càmeres fotogràfiques digitals
Àrees temàtiques de la UPC::Enginyeria civil::Geologia::Oceanografia
Àrees temàtiques de la UPC::Enginyeria electrònica::Instrumentació i mesura
Àrees temàtiques de la UPC::So, imatge i multimèdia::Dispositius de so, imatge i multimèdia
Descripción
Sumario:High-resolution images captured by EMUAS cameras, equipped with a 4K sensor and set to 1440p resolution for the OBSEA deployment, are analysed using the YOLO (You Only Look Once) object detection algorithm, trained with labelled datasets from OBSEA. The cameras operate at 20 frames per second (fps) with H.264+ encoding and a maximum bitrate of 16384. The machine learning model used efficiently identifies and classifies up to 24 marine species. By leveraging convolutional neural networks, the system provides accurate and real-time species recognition, supporting biodiversity assessments and facilitating data-driven marine conservation efforts.