SMART SYSTEM FOR FISH FEEDING CONTROL (SICA) IN OFFSHORE SEA CAGES

Abstract

Improving the efficiency of the feeding process remains one of the major challenges for the aquaculture sector. The lack of control over the fish feeding process, the considerable associated costs and the related environmental impacts are one of the main obstacles to overcome. The Marine Technological Centre (CTN) deployed its cost-effective Smart System for Feeding Control (SICA) in several offshore sea cage farms to validate this real time monitoring technology. The results obtained during these performances in salmon, Gilthead Seabream and Seabass cultures demonstrate that the SICA technology is able to detect different behaviours of fish during feeding process, anticipating human decisions and thus optimizing the feeding process. The potential contribution of SICA technology to reduce feed waste and improve the efficiency of fish production have been demonstrated.

Introduction

The control of the feeding process in aquaculture farms has traditionally been carried out by means of qualified stuff and observations of behavioural cues that indicate fish appetite while feeding (Li et al., 2020). Currently, although computer vision technology and camera-based systems are increasingly in use, they are limited by their dependency on illumination, water conditions, that define visibility, and their inability to monitor the entire population of fish in a net pen. Therefore, alternatives based on acoustic observation of fish behaviour have been examined.

According to different studies (Maniva, 1976; Samueloff, 2000), during the feeding process, fish make different sounds due to their own movement in the water or the splash on the surface to catch the food. In this context, SICA system (Smart System for Feeding Control), uses passive acoustics to distinguish the different sounds produced during the feeding process, without masking problems. For this, the device is based on decision-making through machine learning, a particular approach to Artificial Intelligence (AI). This makes possible to differentiate the moment when fish stop eating and thus indicate when the supply should be stopped.

The aim of this study was to validate the use of SICA technology in offshore sea cages in real production scenarios. First approaches were performed at high commercial interest species in Europe: Gilthead Seabream and Seabass (Dicentrarchus labrax) in Mediterranean fish farms and Salmon (Salmo salar) in Norwegian fish farm.

Material and methodology

SICA prototype is composed of two modules: Data Logger (DL) and Control Unit (CU).  First module is deployed in the sea cage by using a mooring system, this DL acquires the signal through a hydrophone and transmits it to the CU trough wireless connection. Once the signal is received by the second module it is stored, then the acoustic records are processed and analysed automatically.

Information related to fish behaviour was collected during feeding process conducted according to industry standard. Appetite and feed management was registered by CTN researchers and fish farm technicians by means of underwater video camera registers and water surface fish observations.

Using acquired data by the SICA, the information collected during feeding process and a machine learning process, validation test is performed.

Results and discussion

In contrast to the traditional methodology undertaken with underwater video cameras the SICA system was found to be more efficient in detecting unusual fish behaviour since it was able to make earlier predictions during the feeding process. The accuracy in low feed intake detection was of over 84% in Seabream and Seabass cultures. In salmon cultures we find an accuracy 95% due to differences in feeding methodology and technology, an example of salmon feeding acoustic recordings are shown in Figure 1.

The SICA technology results a cost-effective and non-invasive solution in early detection of different fish behaviours during feeding process. Its ability in real time acquisition, data processing and anticipating decisions-making provides great potential to reduce the associated costs and minimizing the impact of the waste on the seabed.

Acknowledgement

The experiments are part of two projects, one of them financed by the European Maritime and Fisheries Fund of the European Commission (DEMO-BLUESMARTFEED) [project agreement number EASME/EMFF/2017/1.2.1.12/S1/05/SI2.789750] and the other financed by HORIZON2020 inside the research infrastructure AQUAEXCEL2020 (SMARTFEEDINSALMON) [project identification code AE120015].

The authors thank PLAGTON and PISCIALBA for their participation as end-users and CAMAR as partner of the project. Also, many thanks to SINTEF Ocean AS for giving its facilities and SalMar as operator.

Bibliography

Li, D., Wang, Z., Wu, S., Miao, Z., Du, L., & Duan, Y. (2020). Automatic recognition methods of fish feeding behavior in aquaculture: A review. Aquaculture, 528(May), 735508. https://doi.org/10.1016/j.aquaculture.2020.735508.

Maniva, Y., 1976. Attraction of bony Fish, Squid and Crab by Sound. Sounds Reception in Fish, Elsevier.

Samueloff, M. F. a. B., 2000. Fish Feeding Control in Aquaculture on the Basis of Sound Emitted by Fish. Tel Aviv, Patent nº WO 00/03586.