Although oceans offer more than 60% of ecosystem services to humans, we still ignore most of its biodiversity and its natural resources. Consequences of this ignorance are broad and problematic in the current context of global changes.
One of the greatest challenges is to assess environmental health, which often requires large amount of data extremely difficult and expensive to acquire because of the technical difficulties to work in such a large and complex environment. Thus, the BUBOT project, through an interdisciplinary approach, proposes the development of new and innovative tools for marine biodiversity observation and their usage to evaluate anthropic impact on marine reef environment. This environmental assessment is coupled to human densities and activities data as well as an anthropology study on the practices, culture and knowledge on fishing resources.
Furthermore, with the intention of thorough testing and to demonstrate the usefulness of this approach, a real challenging assessment of the coastal marine environmental health, using fish assemblages as proxies, is proposed in the world second marine biodiversity hot spot, the Mozambican channel, along an anthropogenic impact gradient. The BUBOT project proposes to study 3 main hotspots of this region: Mayotte, the Scattered Islands and The Mozambique coast.
WP1: New innovative tools for marine biodiversity observation
The BUBOT project aims at merging expertise from biologists, roboticists and computer scientists, to conceive new tools for exploration, observation and monitoring of the marine environment, that do not require heavy logistics nor expensive investment while being efficient to survey large areas repeatedly.
It proposes to develop innovative tools in the three following scientific and technological domains:
Reliable Autonomous Underwater Vehicle (AUV)
Nowadays, available underwater robots and technics do not answer the needs of marine biologists for efficient and autonomous environment assessment. They need small autonomous robots without communication and energy supply cables that are able to perform complex missions. In that context, the robot must become fully autonomous and reliable.
As submarine environment is by nature complex and dynamic, we propose to develop a specific methodology enabling the robot to cope with the high probability of problem occurence (internal hardware and/or software failures, perception uncertainties, adverse environmental conditions or unforeseen events).
Vision-based system
The aim of this work package is to work on the vision system of the robot, both on the hardware and software approaches. We aim at building a fully functional modular underwater stereovision system, allowing the robot to complete the three following functionnalies:
- Detect the surrounding environment, and especially localize obstacles.
- Record a video data for offline processing and fishes identification.
- Provide an estimated displacement vector to the localization algorithm.
The difficulty is to consider in only one vision systems, fully integrated with the planning and control systems of the robot, both the specific caracteristic of the underwater environment (turbidity, specific lighting conditions and reflections, optical deformations, etc.) as well as integrate the difference of the working conditions of these 3 tasks: working distances, on/off line computing, real time execution period..
Automatic reef fish species identification
New observation methods, mainly based on video recording (either still camera or robotic or human video transect), allow biodiversity measuring with high temporal frequency and wide spatial coverage, which is one of the challenges of marine ecology to assess the increasing impacts of global changes.
One of the deadlocks is then the automatic analysis of these videos. For example for fish identification, it is urgent to develop automatic algorithms for locating individuals, recognizing species and measuring their sizes in order to effectively handle large volumes of stereo-videos.
We thus develop a CNN-based (machine learning) software prototype for automatic localization and identification of reef fish species in underwater HD monocular videos, and we built a very precious annotated image database which contains hundreds of fish species. The goal is to obtain a robust prototype allowing fishes identification with a high success rate in very various acquisition conditions.
We also plan to improve the performance of our approaches using the reconstruction of depth information from stereo videos, instead of monocular data.
WP2 : Marine biodiversity changes in relation to depth and anthropic perturbation
Robotic transect based survey (ecology, fish and habitat)
Still stereo camera based survey (mobile species, temporal variation and current impact)
Functional trait measurement (functional fish trait and behavior)
WP3 : Fishing practices, knowledge and perception of fishers on their environment
In the BUBOT project, the reliable autonomous underwater vehicle is an innovative tool to help biologists to increase the frequencies of observation and monitoring of the marine environment. But since few decades, fishers are also considered by scientists as a really reliable sources of information. They are almost all the time at sea, during the day or during the night and at all seasons, whereas scientific campaigns are principally during the day and during the better season. Fishers, by their experiential ecological knowledge of the marine environment, can inform biologists and give them some clue to formulate new hypothesis and design their research. Moreover, if we want to improve the ecosystem health it is essential to invest fishers in the management. Scientists, managers and fishers share a common objective: the sustainability of the fish resources.
Fishing practices in Mayotte
 |  |  |
 |  |  |
The aim of WP3 is to assess the diversity of fishers’ ecological knowledge, their attachment to the marine environment and their sense of belonging to the fishers community. Previous works based on the concept of “place attachment” often showed that the sense of community, the ecosystem dependence and feelings emerging from the interaction to natural environment were in favor of proenvironmental behaviour. These elements define the identity of the person as a fisher and as a person attached to the fishing community. We analyse how fishers became fishers, how they have constructed their link to the marine environment from their childhood to today and how and by who they learnt to be a fisher. On their ecological knowledge, we adopt an ecosystemic approach by asking them several questions about eight species of reef fish selected by the biologists of the project and by asking them their observation of the fish stock evolution on the last decades.