Sentinel-2 Multi-Spectral Instrument (MSI) (S-2) images have been used for mapping burned areas within the borders of the Vesuvius National park, severity affected by fires during summer 2017. A fuzzy algorithm, previously developed for Mediterranean ecosystems and Landsat data, have been adapted and applied to S-2 images. Major improvements with respect to the previous algorithm characteristics are the use of S-2 band reflectance in post-fire images and as temporal difference (delta pre- and post-fire); the definition of fuzzy membership function based on statistical percentiles derived from training areas. Input bands were selected based on their ability to discriminate burned vs. unburned areas. For each input, a sigmoid function has been defined based on percentiles of the unburned and burned histogram distributions, respectively, derived from training data. After the definition of fuzzy Membership functions (MF) they’re applied to the S-2 images in order to find the Membership Degree (MD), or the probability to belong to the burned class. Input membership degrees for the selected bands have been integrated to derived pixel-based synthetic scores of burned likelihood with Ordered Weighted Averaging (OWA) operators. Different operators were tested to represent different attitudes/needs of the stakeholders between pessimistic and optimistic way. Output score maps provided as continuous values in the [0,1] domain have been segmented to extract burned/unburned areas; the performance of the combined threshold and OWA operator has been evaluated by comparison with Copernicus fire damage layers from the Emergency Management Service (EMS) (https://emergency.copernicus.eu/). Error matrix, omission, commission error and Dice coefficient metrics have been analysed. Results show satisfactory accuracy is achieved for the identification of the most severely affected areas while lower performance is observed for those areas identified as slightly damage and probably affected by fires of lower intensity. Then the same algorithm is implemented in another burned area situated in Portugal and the exportability is tested.
Immagini satellitari Sentinel-2 Multi-Spectral-Instrument (MSI) vengono utilizzate per mappare le aree bruciate nei confini del Parco Nazionale del Vesuvio durante l’estate del 2017. L’algoritmo implementato in logica fuzzy si discosta dai precedenti che utilizzano dati Landsat, in quanto sfrutta le bande spettrali S-2 per la calibrazione dell’algoritmo in termini di riflettanza post-evento (POST) e differenza temporale di riflettanza (pre- e post-evento, DELTA). Inoltre vengono definite delle funzioni fuzzy di Membership (MF) o di appartenenza alla classe di bruciato, che associano a ciascun pixel un grado di appartenenza alla classe bruciato nel dominio [0,1]. Le bande spettrali, o features in input alle funzioni fuzzy vengono selezionate in base alla loro capacità di discriminare le aree bruciate dalle aree non bruciate (analisi di separabilità). Per ogni features viene definita una MF basata sui valori dei percentili delle distribuzioni dei dati delle aree bruciate e non bruciate. Dopo aver definito le MF queste vengono applicate all’immagine satellitare S-2 per ottenere una mappa di evidenze parziali, variabile in [0,1] di appartenenza alla classe di bruciato (Membership Degree, MD). Ciascuna delle mappe di evidenza parziale viene data in input ad operatori di aggregazione Ordered Weighted Averaging (OWA). Differenti operatori vengono applicati per rappresentare una diversa attitudine dell’operatore nel valutare le aree bruciate: da pessimistica a ottimistica. Il risultato dell’aggregazione mediante OWA è uno score sintetico di MD, o evidenza globale, in [0,1]. Le mappe di evidenza globale saranno successivamente confrontate con il riferimento Copernicus EMSR213 e verranno analizzate le metriche di accuratezza (errore di omissione, commissione e coefficiente di Dice). I risultati mostrano un’accuratezza soddisfacente per la detezione delle aree severamente bruciate da incendio soprattutto per l’operatore OWA quasi OR. Infine verrà testata l’esportabilità dell’algoritmo che sarà applicato in un’area percorsa dal fuoco in Portogallo.
Mappatura di aree percorse dal fuoco in ambiente mediterraneo mediante un approccio fuzzy applicato ad immagini Sentinel-2
PIASER, ERIKA
2018/2019
Abstract
Sentinel-2 Multi-Spectral Instrument (MSI) (S-2) images have been used for mapping burned areas within the borders of the Vesuvius National park, severity affected by fires during summer 2017. A fuzzy algorithm, previously developed for Mediterranean ecosystems and Landsat data, have been adapted and applied to S-2 images. Major improvements with respect to the previous algorithm characteristics are the use of S-2 band reflectance in post-fire images and as temporal difference (delta pre- and post-fire); the definition of fuzzy membership function based on statistical percentiles derived from training areas. Input bands were selected based on their ability to discriminate burned vs. unburned areas. For each input, a sigmoid function has been defined based on percentiles of the unburned and burned histogram distributions, respectively, derived from training data. After the definition of fuzzy Membership functions (MF) they’re applied to the S-2 images in order to find the Membership Degree (MD), or the probability to belong to the burned class. Input membership degrees for the selected bands have been integrated to derived pixel-based synthetic scores of burned likelihood with Ordered Weighted Averaging (OWA) operators. Different operators were tested to represent different attitudes/needs of the stakeholders between pessimistic and optimistic way. Output score maps provided as continuous values in the [0,1] domain have been segmented to extract burned/unburned areas; the performance of the combined threshold and OWA operator has been evaluated by comparison with Copernicus fire damage layers from the Emergency Management Service (EMS) (https://emergency.copernicus.eu/). Error matrix, omission, commission error and Dice coefficient metrics have been analysed. Results show satisfactory accuracy is achieved for the identification of the most severely affected areas while lower performance is observed for those areas identified as slightly damage and probably affected by fires of lower intensity. Then the same algorithm is implemented in another burned area situated in Portugal and the exportability is tested.File | Dimensione | Formato | |
---|---|---|---|
2020_04_Piaser.pdf
solo utenti autorizzati dal 02/04/2021
Descrizione: Testo della tesi
Dimensione
9.17 MB
Formato
Adobe PDF
|
9.17 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/10589/165126