This thesis presents a methodology to improve the results of standard naval route extrac- tion algorithms based on Automatic Identification System (AIS) data. The techniques discussed focus on the mitigation of undesired statistical effects that reduce reliability and accuracy of the results. Key improvements include the refinement and tailoring of density-based clustering algorithm parameters based on the geographical characteristics of the routes, the addition of dedicated functionalities aimed at minimizing anomalous patterns that may arise from statistical methods, and the addition of routes that could not previously be assessed as partially outside of the defined geographical area of interest. Testing and end product visualization show promising results, with naval routes that appear similar to actual possible navigational patters, with minimal presence of anomalies or unlikely behaviors. Despite this, the presented methodology may require adjustments when applied to a different region, and further optimization and refinement in order to obtain even more value out of the AIS data employed. Future works could incorporate different kinds of data sources, such as radar data or satellite imagery. The implementation of the methodology could be decoupled from the original dataset and made more versatile. Overall, this study represents a valid starting point for future research due to the logical transferability of the framework and the reasonable computational effort required given increased route quality and informational value. Moreover, the broad applicability of accurate and reliable standard naval routes, both in open waters and inland waterways, offers a valuable contribution to numerous practical and research domains.
Questa tesi presenta una metodologia per migliorare i risultati degli algoritmi di estrazione delle rotte navali standard basati sui dati AIS (Autonomous Identification System). Le tecniche discusse si concentrano sulla mitigazione degli effetti statistici indesiderati che riducono l’affidabilità e l’interpretabilità dei risultati. I miglioramenti chiave includono il perfezionamento e l’adattamento dei parametri di algoritmi di clustering density-based in base alle caratteristiche geografiche delle rotte, l’aggiunta di funzionalità dedicate mirate a minimizzare effetti anomali che possono sorgere a causa dell’utilizzo di metodi statistici, e l’aggiunta di rotte che non potevano essere precedentemente valutate poiché parzialmente al di fuori dell’area geografica di interesse definita. I test e la visualizzazione del prodotto finale mostrano risultati promettenti, con rotte navali che appaiono simili ai possibili pattern di navigazione reali, con una presenza min- ima di anomalie o comportamenti improbabili. Nonostante ciò, la metodologia presentata potrebbe richiedere aggiustamenti quando ap- plicata a una regione diversa, e ulteriori ottimizzazioni e perfezionamenti per ottenere ancora più valore dai dati AIS impiegati. Lavori successivi potrebbero incorporare diversi tipi di fonti di dati, come i dati radar o le immagini satellitari. L’implementazione della metodologia potrebbe essere separata dal dataset originale e resa più versatile. Nel complesso, questo studio rappresenta un valido punto di partenza per future ricerche grazie alla trasferibilità logica del framework e al ragionevole sforzo computazionale richiesto, dato l’aumento della qualità delle rotte e del loro valore informativo. Inoltre, la vasta util- ità di rotte navali standard accurate e affidabili, sia in acque aperte che in vie d’acqua interne, offre un contributo prezioso a numerosi ambiti pratici e di ricerca.
A methodology to reduce statistical anomalies in standard naval route extraction
TOGNINI, ELISA
2023/2024
Abstract
This thesis presents a methodology to improve the results of standard naval route extrac- tion algorithms based on Automatic Identification System (AIS) data. The techniques discussed focus on the mitigation of undesired statistical effects that reduce reliability and accuracy of the results. Key improvements include the refinement and tailoring of density-based clustering algorithm parameters based on the geographical characteristics of the routes, the addition of dedicated functionalities aimed at minimizing anomalous patterns that may arise from statistical methods, and the addition of routes that could not previously be assessed as partially outside of the defined geographical area of interest. Testing and end product visualization show promising results, with naval routes that appear similar to actual possible navigational patters, with minimal presence of anomalies or unlikely behaviors. Despite this, the presented methodology may require adjustments when applied to a different region, and further optimization and refinement in order to obtain even more value out of the AIS data employed. Future works could incorporate different kinds of data sources, such as radar data or satellite imagery. The implementation of the methodology could be decoupled from the original dataset and made more versatile. Overall, this study represents a valid starting point for future research due to the logical transferability of the framework and the reasonable computational effort required given increased route quality and informational value. Moreover, the broad applicability of accurate and reliable standard naval routes, both in open waters and inland waterways, offers a valuable contribution to numerous practical and research domains.File | Dimensione | Formato | |
---|---|---|---|
TESI_MAGISTRALE_ELISA_TOGNINI.pdf
non accessibile
Dimensione
7.41 MB
Formato
Adobe PDF
|
7.41 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/230031