Ocean Watch

Project created by Sofia Hedén and part of Industrial Immersion Exchange program 2024
212d ago update

This project aimes to use data gathered in marine environments to create a real time anomaly detection model. The model should be able to identify different types of anomalies found at sea. Primarily ships and other vessels are to be investigated, but the possibility of extending the model to also include other types of anomalies exist, for example oil leaks based on satellite imagery. 

We aim to create a model that can utilize multiple data sources and predict whether a ship is behaving in accordance with the rules and regulations, or if it is somehow deviating from the norm (anomaly).

pages

Resources

2024-06-17 09:15 Image

For our anomaly detection, we are working with AIS (automatic identification system) data to get the positions of the ships and their relative velocities, which can be seen by the blue dots. Our current objective is to make a predictive path of each ship (shown by the green line) so that we can detect when anomalies occur. If a ship slows down in a strange way, or moves in a direction that is anomalous (shown by red lines) then we can send out a message so that an operator can examine the situation in more detail. We also plan on integrating more information such as radar data, which can be used to validate that a ship is truly at the position that their AIS data implies, to make sure that ships are giving real, accurate AIS data.

Attributes

Security
Prediction