Automated ship image analysis using Artificial Intelligence

The MIK (Maritiem Informatie Kruispunt) as well as the Navy have identified the need to automatically process the current camera imagery to allow behaviour and post-event analysis. This project addresses this need by developing and validating deep learning algorithms suitable for specific tasks, and additionally explores the possibility to extend this future capability to other spectral bands.
The main objective of this project is to prototype an automated image analysis platform for optical imagery. A secondary objective is to explore the extension of this capability to the SWIR and the Thermal IR. The full capability will be developed for post-event analysis and the extraction of only a selection of features will be made real-time. Special attention will also go to the creation of relevant databases, by physical data gathering, but also exploring the use of synthetic datasets. This makes the project complementary to the running DEFRA project, called SIREN, allowing useful synergies mainly for data acquisition and building of databases. To our knowledge, no prior studies have examined true behaviour analysis going further than looking at the vessel propagation path. The change detection problem with respect to the configurations of the vessels, and the evaluation of other spectral domains is only sporadically addressed.
We expect this project to provide a demonstrator that fuses the pre-defined anomalous behaviour in order to assess the probability of a threat being present. The key outcome is the demonstration of an automated imagery analysis platform that could shape the future analysis capacity of the MIK.

The project Ship AI is financed under the "Innovation4Defence" call.