Abstract
In-trawl camera systems can make commercial and scientific trawling more efficient, sustainable and improve catch welfare. Trawl image data can inform about species and size-specific catches and their behaviours with a high temporal and spatial resolution. IMR in Norway has an ambition to implement camera systems for scientific sampling in all trawl stations. To support this, the CRIMAC Centre for research-based innovation is developing image-based sampling systems both for commercial fisheries and scientific surveys. We have developed a data pipeline where images from the in-trawl camera system are automatically downloaded, processed, analysed and combined with acoustic data. We use deep learning algorithms to detect and identify a range of different species and are currently improving the count estimates and developing automated length measurements. The next step is to develop automated open-close codend mechanisms and real-time data processing and transfer.
In this presentation we will show examples of how image data can improve acoustic target classification by combining image and acoustic data. We will also present results from automatic monitoring of mackerel length distribution and density during trawl hauls. Real time species and size information can be highly valuable to optimize commercial catches. Trawl-image based sampling in scientific surveys also requires a good understanding of species and size dependent behaviour in the trawl.