Gottardi, Beatrice
(2020)
Automatic methods for crop classification by
merging satellite radar (sentinel 1) and optical (sentinel 2)
.
data and artificial intelligence
analysis.
[Laurea magistrale], Università di Bologna, Corso di Studio in
Analisi e gestione dell'ambiente [LM-DM270] - Ravenna
Documenti full-text disponibili:
![[img]](http://amslaurea.unibo.it/style/images/fileicons/application_pdf.png) |
Documento PDF (Thesis)
Disponibile con Licenza: Salvo eventuali più ampie autorizzazioni dell'autore, la tesi può essere liberamente consultata e può essere effettuato il salvataggio e la stampa di una copia per fini strettamente personali di studio, di ricerca e di insegnamento, con espresso divieto di qualunque utilizzo direttamente o indirettamente commerciale. Ogni altro diritto sul materiale è riservato
Download (13MB)
|
Abstract
Land use and land cover maps can support our understanding of coupled human- environment systems and provide important information for environmental modelling and water resource management. Satellite data are a valuable source for land use and land cover mapping. However, cloud-free or weather independent data are necessary to map cloud-prone regions. Merging radar with optical images would increase the accuracy of the study.
Agricultural land cover is characterized by strong variations within relatively short time intervals. These dynamics are challenging for land cover classifications on the one hand, but deliver crucial information that can be used to improve the machine learning classifier’s performance on the other hand.
A parcel-based map of the main crop classes of the Netherlands was produced implementing a script on GEE and using Copernicus data. The machine-learning model used is a Random Forest Classifier. This was done by combining time series of radar and multispectral images from Sentinel 1 and Sentinel 2 satellites, respectively. The results show the potential of providing useful information delivered by entirely open source data and uses a cloud computing-based approach. The algorithm combines the two satellites data of one year in a multibands image to feed in the classifier. Standard deviation and several vegetation indexes were added in order to have more variables for each 15-day-median image composite. The process paid particular attention to time variability of mean values of each field. This will provide useful information both for understanding differences among crops and variability over the phenology of the plant. The accuracy assessment demonstrates that several crop types (i.e. corn, tulip) can be better classified with both radar and optical images while others (i.e. sugar beet, barley) have an increased accuracy with only radar. The overall accuracy of RFC with optical and radar is 76% while it is 74% if only radar is used.
Abstract
Land use and land cover maps can support our understanding of coupled human- environment systems and provide important information for environmental modelling and water resource management. Satellite data are a valuable source for land use and land cover mapping. However, cloud-free or weather independent data are necessary to map cloud-prone regions. Merging radar with optical images would increase the accuracy of the study.
Agricultural land cover is characterized by strong variations within relatively short time intervals. These dynamics are challenging for land cover classifications on the one hand, but deliver crucial information that can be used to improve the machine learning classifier’s performance on the other hand.
A parcel-based map of the main crop classes of the Netherlands was produced implementing a script on GEE and using Copernicus data. The machine-learning model used is a Random Forest Classifier. This was done by combining time series of radar and multispectral images from Sentinel 1 and Sentinel 2 satellites, respectively. The results show the potential of providing useful information delivered by entirely open source data and uses a cloud computing-based approach. The algorithm combines the two satellites data of one year in a multibands image to feed in the classifier. Standard deviation and several vegetation indexes were added in order to have more variables for each 15-day-median image composite. The process paid particular attention to time variability of mean values of each field. This will provide useful information both for understanding differences among crops and variability over the phenology of the plant. The accuracy assessment demonstrates that several crop types (i.e. corn, tulip) can be better classified with both radar and optical images while others (i.e. sugar beet, barley) have an increased accuracy with only radar. The overall accuracy of RFC with optical and radar is 76% while it is 74% if only radar is used.
Tipologia del documento
Tesi di laurea
(Laurea magistrale)
Autore della tesi
Gottardi, Beatrice
Relatore della tesi
Correlatore della tesi
Scuola
Corso di studio
Indirizzo
CURRICULUM CLIMATE-KIC
Ordinamento Cds
DM270
Parole chiave
Earth Observation,Artificial Intelligence,Crop Mapping,GEE
Data di discussione della Tesi
25 Marzo 2020
URI
Altri metadati
Tipologia del documento
Tesi di laurea
(NON SPECIFICATO)
Autore della tesi
Gottardi, Beatrice
Relatore della tesi
Correlatore della tesi
Scuola
Corso di studio
Indirizzo
CURRICULUM CLIMATE-KIC
Ordinamento Cds
DM270
Parole chiave
Earth Observation,Artificial Intelligence,Crop Mapping,GEE
Data di discussione della Tesi
25 Marzo 2020
URI
Statistica sui download
Gestione del documento: