This feasibility study explored the challenges facing the development of a fully integrated data collection system, and a camera interface to provide automated image capture and species identification from onshore fish grader machines.
The intention is that data collected by the camera can be matched to vessel and weight data routinely collected by the grader machine to provide an automated source of data on fish landings. The study explored the options and obstacles facing the development of such a system, alongside the development of a cost-effective early camera prototype. The project collaborated with Plymouth Trawler Agents.
A method of capturing individual fish images was devised using well-established and freely available python Computer Vision (CV) packages compatible with Raspberry Pi hardware [Raspberry Pi is a trademark of the Raspberry Pi Foundation]. This arrangement used motion-detection to identify fish passing by on the grader machine belt. This data collection method is proposed to be the conceptual foundation for training a ML algorithm which we expect would be capable of automatically identifying fish species, as well as an estimate of individual fish lengths.
Centre for Environment, Fisheries and Aquaculture Science (Cefas)