Applicability of self-sampling for biological data collection within the Belgian fisheries sector to improve stock assessments for commercial fish species (VISIM 2)

The EFMZV project VISIMII aims to introduce Machine Vision in the Belgian beam trawl fishery. The aim is a higher, more extensive data collection on catches and discards in this particular fishery. The researchers are also counting on more data for the so-called data-limited fish species, such as turbot and brill, among others. The Machine Vision technology is expected to strengthen and facilitate the required collaborations between scientist and fishermen.

The researchers plan to develop a mobile compact system that can be easily installed on board at the start and dismantled after each trip. At the fish auction and on commercial fishing vessels, we are testing to what extent we can use deep learning to achieve an accurate length measurement, volume measurement and species recognition of fish passing by on a conveyor belt. We investigate the accuracy of the system and the possibilities and problems during implementation. In addition VISIMII starts from the fishermen and their working environment where we not only implement but also focus on integration of the product to keep the impact on the current operation as low as possible.

Today, only a small percentage of the catches of the Belgian fleet are documented by ILVO. A shortage of data for less common species such as turbot and brill makes it very difficult to estimate their distribution and densities in the fishing areas. The implementation of a more efficient, faster, automated system for data collection aboard fishing vessels can remedy this. The fishermen also have an interest in the improved datasets: in the long run they will have a better view on where the best fishing grounds are located at a given moment, based on length distributions per catch. In the future, this kind of data can be used for stock assessments and catch prediction models as well.