Description
The problem of how fish receive and collectively respond to stimuli presents a challenge for the research community. The project aims at creating an automated approach to detect fish behavior using sonar data. This work proposes the adaptation of computer vision techniques to acoustic sensing. The project will use sonar recordings of herrings being exposed to a series of stressors, including killer whale playbacks, vessel noise and predator models mimicking attacks. The results of this project will be addressed to assess the impact of noise regulation on herring population, both from an ecological and a commercial point of view. The donor partner, Institute of Marine Research in Norway, is an international research center with expertise in marine life, computer science, mathematics, and maritime protection among others. The project promoter, Carlos III University, has a solid background on object detection from different spectra imagery as well as expertise in human activity detection and multisensory systems. The main beneficiaries of the project will be the scientific community working in marine life.
Summary of project results
Despite the growing interest of researchers from diverse disciplines –biomechanics, physics, robotics, ecology, evolutionary biology or collective behaviour- to ascertain the mechanisms that underlie the capability of aggregated individuals to perform coordinate collective flash events, our understanding of these collective processes remains impaired by the methodological approaches currently available to quantify large-scale dynamic reactions exhibited by large groups. This project has been focused on accurately measuring such behaviours in different fish species, atlantic herring (clupea harengus) and atlantic mackerel (scomber scombrus). The results have a direct impact on ensuring the welfare of fish. In this line, the results are of great importance to a couple of Norwegian projects namely Collpen and Redslip, one of their final goals being to ensure that captured fish that have to be released to the sea when the fishing quota is reached are still in good condition that ensures their survival in the wild environment. This fish condition can be measured through their behaviour, detecting when there are abnormalities in the normal swimming patterns. This project allowed processing whole sequences to ensure the reliability of the results and although so far the algorithm has been tested with pre-recorded videos, partners hope that in the future the will be able to integrate it within a real-time monitoring system. Real experiments were conducted in Austevoll, one of Europe’s largest and most advanced research facilities on fish research. The station provided two setups: tanks that allowed controlled experiments and offshore net-pens for testing under real conditions; also, during the experiments, lots of data were collected for further analysis, conducted in the IMR facilities in Bergen. Using acoustics partners identified swimming dynamics, collective reactions and the speed of the propagating waves of evasion induced by a mobile predator model. Results show that large schools make structural and behavioural adjustments in response to perceived risk in a way that improves collective information transfer, and thus responsiveness, during predator attacks. Several papers are under preparation to be submitted to international scientific journals, and partners expect results to be beneficial for fishery industries tin order to increase the welfare of fish in accordance to the international legislation.
Summary of bilateral results
The multidisciplinary team, composed by biologists, engineers and computer scientists has proven to be crucial at several levels since merging knowledge from different research fields allowed to study and measure a series of phenomena that up to now were just evaluated by hand given that there were no accurate automated solutions such as the one partners are proposing. Apart from the productive research carried out during the funded period, the grant has enabled opening new lines of collaboration between both institutions. Opportunities for new projects are being explored to keep the partnership active as well as opening new research lines applying state-of-the-art techniques from machine learning to the sustainability and welfare of fish. The Since the welfare of fish is a hot topic in countries such as Norway where the fisheries constitute one of the pillars of the economy, the host institution is very interested on keeping the collaboration active. In the same way, the Spanish grantee has achieved some degree of expertise in the field that enabled opening new research lines in his home lab (RoboticsLab at University Carlos III of Madrid). Therefore, both groups are seeking means of collaboration by EU funding mechanisms such as H2020.