Fusion 400

Turning data into knowledge earns FUSION Best Runner-up Paper Award 2013.

CMRE's Maritime Security program works to contribute to NATO's common operating picture through common design standards, data fusion, and the use of historical data to provide enhanced knowledge of vessel activities. These efforts have been recognized at the 16th International Conference on Information Fusion (Istanbul, 9-12 July 2013), where a paper written by a CMRE team was awarded best runner-up.  The paper, titled "Traffic Knowledge Discovery from AIS Data," was written by Giuliana Pallotta, Michele Vespe and Karna Bryan.

The paper proposes an unsupervised, incremental learning approach to extract historical traffic patterns from Automatic Identification System (AIS) data.  AIS provides a rich source of cooperative information on vessel movement via growing networks of shared coastal receivers as well as commercially available AIS records from space.  The rapidly increasing amount of data simply cannot be processed by operators.  A compact representation synthesizing this vast amount of data gives operational utility to data which would otherwise be ignored.

The paper describes CMRE's methodology, called "Traffic Route Extraction for Anomaly Detection" (TREAD,) which infers traffic routes, ports, and stationary areas from raw AIS data.  TREAD can be used to classify and predict vessel behaviors and detect low-likelihood behaviors, or anomalies. The award- winning paper also describes data and results from CMRE's coastal AIS receiver, which is available to collaborators to further their own research requiring contextual information layers.  One example of this is context-based tracking, in which traffic routes can inform filtering algorithms in the absence of sensor measurements.

The FUSION Conference also awarded "best paper" to the Gerhard Kurz and Uwe Hanebeck Karlsruhe Institute of Technology (KIT) (Germany) for "Recursive Fusion of Noisy Depth and Position Measurements for Surface Reconstruction."  Second-best runner-up was "A New Approach for Doppler-only Target Tracking" submitted by University of Florence (Italy) and Selex (Italy).