Session: MESA-01-01 Artificial Intelligence and Emerging Technologies for Mechatronics and Embedded Systems

Paper Number: 69799

Start Time: August 17, 11:10 AM

69799 - A Feature Encoding Approach and a Cloud Computing Architecture to Map Fishing Activities 

In recent years Vessel Monitoring System (VMS) and Automatic Identification System (AIS) have been developed for vessels longer than 12/15 m in length overall, and their use has proved very effective to support equitable resource allocation and sustainable fisheries management. In this scenario, small scale vessels (< 12 m in length) remain untracked and largely unregulated, even though they account for 83% of all fishing activity in the Mediterranean Sea. In this paper we present an architecture that makes use of a low-cost LoRA / cellular network to acquire and process positioning data from small scale vessels, and a feature encoding approach that can be easily extended to process and map small scale fisheries. Indeed, the use of a such low-cost and open source technology coupled to automated statistical analyses could open up potential for more integrated, accessible and transparent platforms to inform coastal resource management, fisheries management and cross-border marine spatial planning. The feature encoding method uses a Markov chain to model transitions between successive behavioural states (e.g., fishing, steaming) of each vessel and probabilistically classify its activity. Cross-validations by k-fold and Leave One Boat Out showed higher accuracy in comparison with a standard method based on non-encoded data.

Presenting Author: Alessandro Galdelli Universit&#224; Politecnica delle Marche

Authors:

A. Galdelli Università Politecnica delle Marche
A. Mancini Università Politecnica delle Marche
E. Frontoni Università Politecnica delle Marche
A. N. Tassetti Institute of Marine Biological Resources and Biotechnologies National Research Council