unknown
5 Akvaplan-niva (current employee)
1 Akvaplan-niva (prior employee)
Authors (11)
- Lionel Camus
- Kanchana Bandara
- Trude Kristin Borch
- Ragnhild Pettersen
- Morten Thorstensen
- Sigmund Sevatdal
- Tormod Henry Skålsvik
- Marc Picheral
- Leif Edvard Bildøy
- Marta Moyano
- Jerome Coindat
Abstract
The salmon industry is currently challenged by sea lice infestations (Lepeophtheirus salmonis). This calls for a need to develop an in situ early warning system to detect, count and report the presence of free swimming sea lice in the water to stakeholders. Herein, we propose a novel approach based on an in-situ imaging sensor, an underwater vision profiler version 6 (UVP6) and onboard data processing using artificial intelligence for real-time detection and classification. We have developed AI algorithms for free swimming sea lice, adults lice (male and female) and ca. 30 different plankton taxa. The UVP6 was deployed from an autonomous buoy from a winch to collect data across depth and the algorithms were deployed in the UVP6 onboard microprocessor. Data were streamed to a dashboard for end users. The results show that the performance of the AI running in the sensor is not sufficient quality compared to running the AI in the cloud which offers more powerful processing capacity. Overall, a system has been successfully developed, tested, debugged and optimized. This proof of concept points out the need to further improve on the AI component, the data processing pipeline and the sensor hardware in order to offer a robust solution for commercial use.