Detection of Commercial Fishing-related Slipping Events using Multimodal Data

Slipping, or deliberate release, of dead or dying fish is a potentially illegal act, depending on the fish species and jurisdiction where the fish was caught. These events are extremely difficult to regulate, due to fishing vessels being far out to sea and spread out across a large area. Therefore, detecting such events manually using inspection boats is infeasible. However, slipping events can lead to release of fish oil, if the fish dies as part of the release process, which can be detected in satellite imagery given good detection capabilities. We propose an approach to detect fishing vessels that might have deliberately released fish they have caught. The method we propose is to analyze multiple sources of data simultaneously and combine it, using positional data, sales notes, oil slick detection and satellite-based vessel detection. Our experimental results show that we are able to provide a ranking of suspicious vessels that can be used for further investigations.