The subject of this thesis is the detection and the mapping of burned areas, using multispectral and SAR satellite images. This need arises from the necessity to have a fast response when catastrophic events occur, such as fires, on a large scale, but which would be difficult to monitor on earth. In these cases, the presence of satellite missions dedicated to Earth observation are crucial in assessing the damage, making it faster and more accurate. This study aims to map the burned areas by using parameters derived from SAR images, with the support of optical images. This choice was dictated by the characteristics of the two subjects: on one hand the optical images which have a certain seniority in the field of Earth observation with many auxiliary studies, on the other hand the development of radar images allowed to provide images with continuous coverage, regardless of day and night and the presence of clouds and smoke. The combination of these aspects helped to provide useful information for the purpose of the analysis. The mapping was made by using supervised classification algorithms, with the SAR image data as variable data, while the optical image data as response data, to associate the burned/unburned labels with the values of the SAR parameters, thus training the classifier to recognize the pixels of the image autonomously. The analysis was carried out on some fires that spread in the state of California in the recent years, and which had an impact on a huge area.
In questa tesina viene trattato il tema del rilevamento e della mappatura delle aree bruciate, attraverso l’utilizzo di immagini satellitari multispettrali e SAR. Questa esigenza nasce dalla necessità di avere una risposta tempestiva laddove si verifichino eventi catastrofici, come gli incendi, su larga scala, ma che sarebbero difficili da monitorare via terra. In questi casi, la presenza delle missioni satellitari destinate all’osservazione della Terra risulta determinante nella valutazione dei danni, rendendola più rapida e accurata. Questo studio ha lo scopo di mappare le aree bruciate attraverso l’utilizzo di parametri derivati dalle immagini SAR, con l’appoggio delle immagini ottiche. Tale scelta è stata dettata dalle particolari caratteristiche dei due soggetti: da un lato le immagini ottiche che godono di una certa anzianità nel campo dell’osservazione della Terra con numerosi studi a supporto, dall’altro lo sviluppo delle immagini radar che ha permesso di fornire immagini con una copertura continua, indipendentemente da giorno e notte e dalla presenza di nuvole e fumo. La combinazione di questi aspetti ha contribuito a fornire informazioni utili per lo scopo dell’analisi. La mappatura avviene tramite algoritmi di classificazione supervisionati, usando i dati delle immagini SAR come variable data, mentre i dati delle immagini ottiche come response data, in modo da associare ai valori dei parametri SAR l’etichettatura bruciato/non bruciato, allenando così il classificatore a riconoscere i pixel dell’immagine autonomamente. L’analisi è stata fatta su alcuni incendi che hanno colpito lo stato della California negli ultimi anni, e che hanno avuto un impatto su un vasto territorio.
Burned area mapping using Synthetic Aperture Radar
Laabadi, Jihad
2020/2021
Abstract
The subject of this thesis is the detection and the mapping of burned areas, using multispectral and SAR satellite images. This need arises from the necessity to have a fast response when catastrophic events occur, such as fires, on a large scale, but which would be difficult to monitor on earth. In these cases, the presence of satellite missions dedicated to Earth observation are crucial in assessing the damage, making it faster and more accurate. This study aims to map the burned areas by using parameters derived from SAR images, with the support of optical images. This choice was dictated by the characteristics of the two subjects: on one hand the optical images which have a certain seniority in the field of Earth observation with many auxiliary studies, on the other hand the development of radar images allowed to provide images with continuous coverage, regardless of day and night and the presence of clouds and smoke. The combination of these aspects helped to provide useful information for the purpose of the analysis. The mapping was made by using supervised classification algorithms, with the SAR image data as variable data, while the optical image data as response data, to associate the burned/unburned labels with the values of the SAR parameters, thus training the classifier to recognize the pixels of the image autonomously. The analysis was carried out on some fires that spread in the state of California in the recent years, and which had an impact on a huge area.File | Dimensione | Formato | |
---|---|---|---|
Thesis.pdf
non accessibile
Dimensione
3.25 MB
Formato
Adobe PDF
|
3.25 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/10589/188925