The AZTI technology center, a leader in the integration of new technologies in the fishing industry, has developed an innovative Artificial Intelligence (AI) model to contribute to improving the management and sustainability of commercial fishing resources.
The AZTI team has developed a machine learning model that identifies the main pelagic species of the Bay of Bizkaia such as anchovies, sardines and greenfish based on their aggregation behavior, which allows us to know the specific composition of the fish schools detected using acoustic equipment. fishing. This technology advances the challenge of determining the composition of schools in situations of high diversity, allowing the study of specific changes in behavior in the presence of other species. Based on multidisciplinary campaigns such as JUVENA, these types of studies allow not only better management of the main species of pelagic fish, but also a better understanding of the integral functioning of the ecosystem, from plankton to apical predators such as birds and cetaceans.
The AI model was trained in a partially supervised manner, combining fully identified schools of fish with partially identified ones. The results are presented probabilistically, indicating the probability that a school of fish belongs to one species or another, which allows the model’s confidence in each prediction to be measured.
The results are promising: the application of AI in acoustic records from sonars and echo sounders from fishing vessels could be a useful and effective tool for improving fisheries management. Identification of fish schools would significantly reduce processing time and improve the accuracy of data used to monitor species distribution and abundance. But their application could go further, these models could be implemented in the fishing sector, contributing to the improvement of the effectiveness and sustainability of fishing activity, since a correct identification of the schools would reduce unwanted catches.
The results, published in the ICES Journal of Marine Science, show an accuracy of 63.5% in the classification of pelagic species in the supervised part and approximately 80% in the semi-supervised part.
The main author of the research is the marine scientist Aitor Lekanda, who is currently carrying out his thesis at AZTI under the supervision of researchers Guillermo Boyra and Maite Louzao. Lekanda highlights: «By automating the identification of species, we not only reduce the data processing time of scientific campaigns, but it also opens the doors to studying the aggregation behavior of pelagic species and to the development of new technologies for the improvement of the efficiency, selectivity and sustainability of the fishing sector.
Technological leadership
The development of this AI application reinforces AZTI’s technological leadership in the fishing sector. Its experience in managing massive data and predictions through machine learning, big data and AI has allowed it to offer innovative and sustainable solutions that benefit both the fishing sector and the marine environment.
A notable example is its participation in the European project SMARTFISH, an initiative that develops and promotes intelligent systems for the fishing sector of the European Union, reducing its ecological impact. In addition, AZTI coordinates a work package for size discrimination and species identification in purse seine fisheries, demonstrating its commitment to sustainability and efficiency in fishing.
These projects show how AZTI uses its expertise in advanced technologies to improve efficiency and sustainability in the fishing sector, establishing itself as a benchmark in the application of AI and big data in fishing.
#improve #fisheries #sustainability
