MARLIT, artificial intelligence against marine waste

Marine macro-waste threatens the conservation of marine ecosystems worldwide. Ocean gyres have the highest density of floating debris – circular stream systems that spin and trap waste – but the polluting waste is abundant in coastal waters and semi-enclosed oceans such as the Mediterranean Sea.

MARLIT, an algorithm-based open access web app designed with deep learning techniques, will allow the detection and measurement of floating plastic in the sea with a reliability above 80%, according to a study published in the journal Environmental pollution and conducted by experts of the Faculty of Biology and the Biodiversity Research Institute of the University of Barcelona (IRBio).

This method is the result of the analysis through artificial intelligence methods of more than 3,800 aerial images of the Mediterranean coast of Catalonia, and will allow researchers to make progress in assessing the presence, the density and circulation of plastic pollutants in the oceans and seas. all over the world. Participants in the study, published in the journal Environmental Pollution, include experts from the UB and IRBio Joint Research Group on Marine Vertebrates, and the UB Research Group on Biostatistics and Bioinformatics (GRBIO), united in the Barcelona Bioinformatics Platform (BIB).

Litter that floats and pollutes the ocean

Historically, direct observation (ships, planes, etc.) has been the basis for the common approach to assessing the impact of marine macro-waste (FMML). However, the size of the ocean and the size of the data make it difficult for the researchers to proceed with the research studies.

“Automated photography methods combined with analytical algorithms are more effective protocols for the control and analysis of such pollutants”, notes Odei Garcia-Garin, first author of the article and member of the CRG on Marine Mammals Large, directed by Professor Alex Aguilar.

“But, -he continues-, there is a remote automated awareness of these materials at an early stage. There are several objects in the ocean (waves, wind, clouds, etc.) that harden the detection of floating debris to automated with aerial imagery of the sea surface.This is why only a few studies have attempted to work on algorithms to be relevant to this new research context “.

The experts designed a new algorithm to automate the measurement of floating plastics in the sea through aerial photography using the deep learning techniques, an automated learning method with artificial neuronal networks capable of learning and learning. take it to higher levels.

“The large number of sea surface images obtained by drones and aircraft in the monitoring of marine litter missions – also in experimental studies with known floats – has allowed us to develop and test a new algorithm. which reaches 80% of error in the margin of floating marine macro-waste awareness ”, notes Garcia-Garin, a member of the UB and IRBio’s Department of Evolutionary Biology, Ecology and Environmental Sciences.

The seas were preserved with deep learning techniques

The new algorithm was implemented to MARLIT, an open access web app described in the article and available to all managers and professionals in analyzing the detection and measurement of marine macro-waste with images from the adhar. In particular, this is a testament to the concept of the Shiny R package, a methodological innovation with a strong interest in accelerating marine macro-waste monitoring methods.

MARLIT enables stand-alone image analysis, as well as divided into several sections – adhering to user instructions -, identifying the presence of debris in each specific area and estimating condensed by the image metadata (height, resolution). In the future, it is planned to switch the app to a remote sensor (for example, drone) to automate the remote sensing process.

At European level, the EU Marine Strategy Framework Directive highlights the use of FMML monitoring methods to carry out continuous assessment of the state of the marine environment. “Therefore, automating monitoring processes and the use of applications such as MARLIT would help member states implement the guidelines”, concludes the authors of the study.

###