Within the INDISAL project, the partners developed camera solutions and visual technology that enables the automatic observation of characteristics of the salmon and also allows the biometric re-identification of individual salmons. New and innovative underwater light and imaging solutions are built that can provide underwater recordings with an unprecedented high quality. This, in turn, allowed to exploit and develop state-of-the-art computer vision and machine learning techniques to automatically identify individual salmons and measure key fish-status and/or welfare variables. A "finger-print" like biometric identification of individual salmons along with the automatic observation of characteristics of the salmon lays the foundation to measure and understand their welfare and growth within commercial fish farms. The project owner, Sealab AS, develops innovative steerable high quality camera solutions that allow to established 24/7 data connections to commercial fish farms, thereby enabling remote access to high quality camera recordings at any time from a salmon-cage at a commercial fish farm. The longterm data-gathering was also essential for the development of robust algorithms as this allowed to test developed approaches under real lighting and water conditions. A full-scale experiment for data gathering from a sub-cage and for the testing of biometric identification approaches under realistic lighting conditions in a commercial fish farm was performed. The developed and tested salmon identification methods demonstrated for the first time the ability that individual salmons could be re-identified in underwater-video-streams that were not recorded at the same day and under different lighting conditions. We also demonstrated the first approach for the measurement of the apparent mouth motion of salmon, leading to a "mouth-opening-frequency" measurement as a behavioral status variable of the salmon that also has a high likelihood to be a valuable welfare indicator for salmon as well.
The insights and results gained within the INDISAL project will have a significant impact in aquaculture. The re-identification of salmon opens up for a non-invasive observation of individual or groups of salmon over a prolonged time-period enabling the study of temporal changes in the appearance and behavior of the fish. Objective characterization measurements of the salmon, along with a biological interpretation, will improve the understanding of their relation to stress and will subsequently increase the ability to improve the welfare of salmon.
The developed and tested salmon identification methods demonstrated for the first time the ability that individual salmons could be re-identified in underwater-video-streams that were not recorded at the same day and under different lighting conditions.
We also demonstrated the first approach for the measurement of the apparent mouth motion of salmon, leading to a "mouth-opening-frequency" measurement as a behavioral status variable of the salmon that also has a high likelihood to be a valuable welfare indicator for salmon as well.
The re-identification of salmon opens up possibilities for a non-invasive observation of individual or groups of salmon over a prolonged time-period, enabling the study of temporal changes in the appearance of the fish. This includes the growth of individual salmon and/or the development of visual damages over time.
The idea of the INDISAL project is to develop an individual biometric "finger-print" identification for salmon, enabling the gathering of status information for observed individual salmons over time in industrial sea based salmon farming. For this we develop and improve innovative underwater light solutions that help -- together with camera technology -- to provide excellent video-recordings and algorithms that can identify and analyze the characteristics of salmon from video streams.
The innovation enables the salmon farming industry to have a real time overview of the current state of each individual salmon (i.e. growth-rate, quality, health status), while the collected time-line data of individual salmons can be used to analyze in detail which operational events had an impact on both individual fish and the whole fish-population. This could be utilized to improve fish welfare and to optimize the biomass production of the complex and dynamic fish-cage population. In addition, this opens completely new possibilities to observe and investigate the development and behaviour of salmons in fish-cages.