Use of underwater video and machine learning for monitoring of lobster stocks (RD131)

Project Summary:

Sustainable fishery management requires evidence to support decision-making. Commercial fishery catch rates provide important information on stock abundance; however, this is problematic when catches are determined by factors additional to abundance. In creel (trap) fisheries, the capture process is determined by behavioural responses to the gear, complicated by interactions between and within species, potentially resulting in catch rates that are de-coupled from underlying patterns of abundance. Fishery-independent monitoring methods for European lobsters are also problematic because of difficulties of observing and quantitatively sampling this cryptic species. At the same time, there is a strong need for evidence to support emerging inshore management regimes, particularly because lobster fishing is of high socioeconomic importance for fragile coastal communities.

This project follows on from a feasibility study into measuring lobster abundance by coupling a low-cost creel-mounted underwater camera system with machine-learning technology to detect and quantify lobster presence near fishing activities (FS069) . Given feasibility demonstrated, this full-scale R&D study is developing this towards a tool for use by:

(i) fishery scientists advising managers on lobster stock status;

(ii) individual fishers to guide targeting of effort; and

(iii) researchers investigating fishery ecology.

Further applications of the technology are also explored. The project is delivered through partnership of Heriot-Watt University in Orkney, the Regional Inshore Fishery Group Orkney Sustainable Fisheries Ltd, Bangor University working in the Isle of Man, and The Data Lab, Scotland’s Innovation Centre for data and artificial intelligence.

For more on this project’s previous work, see here.