From underwater to drone: A novel multi-scale knowledge distillation approach for coral reef monitoring

被引:0
作者
Contini, Matteo [1 ,2 ]
Illien, Victor [1 ]
Barde, Julien [4 ]
Poulain, Sylvain
Bernard, Serge [3 ]
Joly, Alexis [2 ]
Bonhommeau, Sylvain [1 ]
机构
[1] IFREMER, Delegat Ocean Indien DOI, Rue Jean Bertho, F-97420 Le Port, La Reunion, France
[2] Univ Montpellier, INRIA, LIRMM, CNRS, F-34000 Montpellier, France
[3] Univ Montpellier, CNRS, LIRMM, F-34000 Montpellier, France
[4] Univ Montpellier, UMR Marbec, IRD, CNRS,Ifremer, F-34000 Montpellier, France
关键词
Coral reef monitoring; Computer vision; Knowledge distillation; Marine biodiversity; Multi-scale imaging; CLIMATE-CHANGE;
D O I
10.1016/j.ecoinf.2025.103149
中图分类号
Q14 [生态学(生物生态学)];
学科分类号
071012 ; 0713 ;
摘要
Drone-based remote sensing combined with AI-driven methodologies has shown great potential for accurate mapping and monitoring of coral reef ecosystems. This study presents a novel multi-scale approach to coral reef monitoring, integrating fine-scale underwater imagery with medium-scale aerial imagery. Underwater images are captured using an Autonomous Surface Vehicle (ASV), while aerial images are acquired with an aerial drone. A transformer-based deep-learning model is trained on underwater images to detect the presence of 31 classes covering various coral morphotypes, associated fauna, and habitats. For aerial analysis these predictions are refined (some classes are merged, others are retained, while some are removed) resulting in a final set of 12 ecological categories that serve as annotations for training a second model applied to aerial images. The transfer of information across scales is achieved through a weighted footprint method that accounts for partial overlaps between underwater image footprints and aerial image tiles. The results show that the multi-scale methodology successfully extends fine-scale classification to larger reef areas, achieving a high degree of accuracy in predicting coral morphotypes and associated habitats. The method showed a strong alignment between underwater-derived annotations and ground truth data, reflected by an AUC (Area Under the Curve) score of 0.9251. This shows that the integration of underwater and aerial imagery, supported by deep-learning models, can facilitate scalable and accurate reef assessments. This study combines multi-scale imaging and AI to provide scientific information on coral reef monitoring and conservation. Our approach leverages underwater and aerial imagery, aiming for the precision of fine-scale analysis while extending it to cover a broader reef area.
引用
收藏
页数:16
相关论文
共 46 条
[41]   Integrating artificial intelligence in biodiversity conservation: bridging classical and modern approaches [J].
Ullah, Fazal ;
Saqib, Saddam ;
Xiong, You-Cai .
BIODIVERSITY AND CONSERVATION, 2025, 34 (01) :45-65
[42]  
van Dam J.W., 2011, ECOLOGICAL IMPACT TO, P187, DOI DOI 10.2174/978160805121210187
[43]   Coastal benthic habitat mapping and monitoring by integrating aerial and water surface low-cost drones [J].
Ventura, Daniele ;
Grosso, Luca ;
Pensa, Davide ;
Casoli, Edoardo ;
Mancini, Gianluca ;
Valente, Tommaso ;
Scardi, Michele ;
Rakaj, Arnold .
FRONTIERS IN MARINE SCIENCE, 2023, 9
[44]   Knowledge Distillation and Student-Teacher Learning for Visual Intelligence: A Review and New Outlooks [J].
Wang, Lin ;
Yoon, Kuk-Jin .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2022, 44 (06) :3048-3068
[45]   'Structure-from-Motion' photogrammetry: A low-cost, effective tool for geoscience applications [J].
Westoby, M. J. ;
Brasington, J. ;
Glasser, N. F. ;
Hambrey, M. J. ;
Reynolds, J. M. .
GEOMORPHOLOGY, 2012, 179 :300-314
[46]  
Wolf T, 2020, PROCEEDINGS OF THE 2020 CONFERENCE ON EMPIRICAL METHODS IN NATURAL LANGUAGE PROCESSING: SYSTEM DEMONSTRATIONS, P38