Enhancing Maritime Behaviour Analysis through Novel Feature Engineering and Digital Shadow Modelling: A Case Study in Kiel Fjord
This repository contains an implementation of an extension and utilisation of the Surface Vessel Nautical Behaviour Analysis (SV-NBA) framework for in-depth spatio-temporal analysis of maritime surface vessels’ behaviour in Kiel Fjord.
To install all the requirements, one needs to first install:
- conda
- poetry
A detailed list of the required libraries can be found in:
- poetry.toml
The proper installation must then be done with poetry and conda.
The collected data comprises of AIS signals received in a monitoring period from 24.03.22 to 30.06.2023. Of those days there were 296 for which all outgoing AIS messages in the Kiel Fjord region were recorded.
The data used in the experiment has a time resolution of 10 seconds, while the data in the data/assets are sample data obtained through downsampling, with a resolution of 10 minute.
This framework illustrates a transformer-based autoencoder that takes input data X, transforms it into an intermediate representation Y through an encoder, and then reconstructs it into output Z using a decoder. During this process, a loss function Loss(X, Z) is used to optimize the model for accurate reconstruction of the input data. Finally, the intermediate representation Y is utilized for behavior analysis, categorizing data points into different behavior types.
[gaf, sga, lha]@informatik.uni-kiel.de
We use the MIT license, see
- LICENSE