New global area estimates for coral reefs from high-resolution mapping

Lyons, Mitchell B.; Murray, Nicholas J.; Kennedy, Emma V.; Kovacs, Eva M.; Castro-Sanguino, Carolina; Phinn, Stuart R.; Acevedo, Rodney Borrego; Alvarez, Alexandra Ordoñez; Say, Chantel; Tudman, Paul; Markey, Kathryn; Roe, Meredith; Canto, Robert F.; Fox, Helen E.; Bambic, Brianna; Lieb, Zoë; Asner, Gregory P.; Martin, Paulina M.; Knapp, David E.; Li, Jiwei; Skone, Matthew; Goldenberg, Eldan; Larsen, Kirk; Roelfsema, Chris M.

doi:10.1016/j.crsus.2024.100015

Cited by 7 publications

(7 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The aleatoric component describes how much a strategy's results depend on random processes in the environment, while the epistemic component describes how much risk could be eliminated through better knowledge of the environment [107], [108]. 2 Actions which reveal information on the hidden state of the POMDP reduce epistemic risk. Conversely, aleatoric risk cannot be avoided in OLPs, as it is constant for a fixed hidden state.…”

Section: Risk Quantification Decomposition and Minimizationmentioning

confidence: 99%

“…Typically, the vocabulary is domain-specific while the cell size and the scalar hyperparameters are tuned for a specific mission. 2 All robots on the same mission use the same hyperparameters. Finally, each robot's topic model uses a stochastic process to develop an ever-evolving set of topics online based on its own visual observations.…”

Section: Online Stm-based Semantic Mappingmentioning

confidence: 99%

“…This is particularly problematic at shallow depths (<100m), long target ranges (>1m), and in turbid waters. 2 At shallow depths, light from the surface (e.g., sunlight) is scattered in all directions by tiny particulate in the water; the spectrum of this scattered light consist of wavelengths that were not absorbed higher up in the water column, and thus tends to be blue or blue-green. The greater the range to the target, the more intervening particulate there is to backscatter this light into the camera; the resulting "haze" in the image makes it more difficult to see the target, especially when it saturates the camera's dynamic range.…”

Section: Underwater Image Formationmentioning

confidence: 99%

“…Sites targeted for exploration are generally remote, unfamiliar, and/or dynamic, qualities which make novel or unexpected discoveries more likely. For example, unusual topographical map features may inspire a search for hydrothermal vents in unexpected parts of the deep sea [1], 1 or multispectral satellite imagery may suggest the presence of coral reefs along remote coasts [2], motivating higher-resolution in situ exploration to characterize possible new species and reef compositions [3]. Meanwhile, the routine monitoring of underwater assets or ecosystems is often motivated by the potential discovery of emergent threats, such as invasive species, diseases, or biofouling, which often demand immediate follow-up investigations or interventions.…”

Section: Chapter 1 Introductionmentioning

confidence: 99%

“…An exploration mission begins with only an educated guess of what might be found; one cannot be certain about what the most important priorities will be until after the mission has started and new discoveries have been made. 2 In the practices of underwater exploration, humans and robots have traditionally occupied distinct roles. Human explorers, equipped with intuition and experience, recognize immediately when new and interesting phenomena are discovered and react by directing their search towards them.…”

Section: Chapter 1 Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Enabling human-multi-robot collaborative visual exploration in underwater environments

Jamieson

View full text Add to dashboard Cite

This thesis presents novel approaches to vision-based autonomous exploration in underwater environments using human-multi-robot systems, enabling robots to adapt to evolving mission priorities learned via a human supervisor’s responses to images collected in situ. The robots model the spatial distribution of various habitats and terrain types in the environment using semantic classes learned online, and send image queries to the supervisor to learn which of these classes are associated with the highest concentration of targets of interest. The robots do not require prior examples of these targets, and learn these concentration parameters online. This approach is suitable for exploration in unfamiliar environments where unexpected phenomena are frequently discovered, such as coral reefs. A novel risk-based online learning algorithm identifies the concentration parameters using the fewest possible number of queries, enabling the robots to adapt quickly and reducing the operational burden on the supervisor. I introduce four primary contributions to address prevalent challenges in underwater exploration. Firstly, a multi-robot semantic representation matching algorithm enables interrobot sharing of semantic maps, generating consistent global maps with 20-60% higher quality scores than those produced by other methods. Next, we present DeepSeeColor, a novel real-time algorithm for correcting underwater image color distortions, which achieves up to 60 Hz processing speeds, thereby enabling improved semantic mapping and target recognition accuracy online. Thirdly, an efficient risk-based online learning algorithm ensures effective communication between robots and human supervisors, and, while remaining computationally tractable, overcomes the myopia which would cause previous algorithms to underestimate a query’s value. Lastly, we propose a new reward model and planning algorithm tailored for autonomous exploration, together enabling a 25-75% increase in the number of targets of interest located when compared to baseline surveys. These experiments were conducted with simulated robots exploring real coral reef maps and with real, ecologically meaningful targets of interest. Collectively, these contributions overcome key barriers to vision-based autonomous underwater exploration, and enhance the capability of autonomous underwater vehicles to adapt to new and evolving mission objectives in situ. Beyond marine exploration, these contributions have value in broader applications, such as space exploration, ecosystem monitoring, and other online learning problems.

show abstract

Section: Risk Quantification Decomposition and Minimizationmentioning

confidence: 99%

Section: Online Stm-based Semantic Mappingmentioning

confidence: 99%

Section: Underwater Image Formationmentioning

confidence: 99%