The availability of thematic maps has significantly increased over the last few years. Validation of these maps is a key factor in assessing their suitability for different applications. The evaluation of the accuracy of classified data is carried out through a comparison with a reference dataset and the generation of a confusion matrix from which many quality indices can be derived. In addition to the most commonly used indices, e.g. Overall Accuracy and Kappa, the literature proposes the use of less known indices such as the Ground Truth, Allocation Disagreement, and Quantity Disagreement. In this work, besides the theoretical analysis of all available indexes, the quality of three high-resolution global datasets (GL30, GUF, GHSL) has been evaluated through a suite of open source tools: GRASS GIS, QGIS, and Python programming language scripts (developing some ad hoc scripts for the last one). A further analysis was intended to evaluate the spatial correlation of errors using a Python library, i.e. PySAL. The results show a satisfactory accuracy of the considered high-resolution global datasets, with emphasis that less detailed classification system of the map it provides better accuracy of the global land cover product. The thesis has given basis for development of plug-in that, when constructed, will be implemented in open source GIS software and give access to more validation indices that at this moment are not available in most used commercial software.
La disponibilità di mappe tematiche negli ultimi anni è notevolmente aumentata. La validazione di tali mappe rappresenta un fattore chiave per valutare la loro utilità in diverse applicazioni. La valutazione dell'accuratezza di un dato classificato viene effettuata attraverso un confronto con un dataset di riferimento e la generazione di una matrice di confusione dalla quale è possibile derivare numerosi indici di qualità. Oltre agli indici più comunemente utilizzati, quali Overall Accuracy e Kappa, la letteratura propone l'utilizzo di indici meno noti come gli indici di Ground Truth, Allocation Disagreement e Quantity Disagreement. In questo lavoro, accanto all’analisi teorica di tutti gli indici riporati in letteratura, la qualità di tre dataset globali ad alta risoluzione (GL30, GUF, GHSL) è stata valutata mediante una suite di strumenti open source, in particolare GRASS GIS, QGIS e script generati tramite linguaggio di programmazione Python. Un'ulteriore analisi è stata dedicata alla correlazione spaziale degli errori mediante una libreria Python, PySAL. I risultati indicano una soddisfacente accuratezza dei dataset globali ad alta risoluzione considerati ed evidenziano come l'accuratezza migliori al diminuire del numero di classi.
Validation of the high-resolution global land cover maps in Lombardy
BRATIC, GORICA
2016/2017
Abstract
The availability of thematic maps has significantly increased over the last few years. Validation of these maps is a key factor in assessing their suitability for different applications. The evaluation of the accuracy of classified data is carried out through a comparison with a reference dataset and the generation of a confusion matrix from which many quality indices can be derived. In addition to the most commonly used indices, e.g. Overall Accuracy and Kappa, the literature proposes the use of less known indices such as the Ground Truth, Allocation Disagreement, and Quantity Disagreement. In this work, besides the theoretical analysis of all available indexes, the quality of three high-resolution global datasets (GL30, GUF, GHSL) has been evaluated through a suite of open source tools: GRASS GIS, QGIS, and Python programming language scripts (developing some ad hoc scripts for the last one). A further analysis was intended to evaluate the spatial correlation of errors using a Python library, i.e. PySAL. The results show a satisfactory accuracy of the considered high-resolution global datasets, with emphasis that less detailed classification system of the map it provides better accuracy of the global land cover product. The thesis has given basis for development of plug-in that, when constructed, will be implemented in open source GIS software and give access to more validation indices that at this moment are not available in most used commercial software.File | Dimensione | Formato | |
---|---|---|---|
2018_04_Bratic.pdf
accessibile in internet per tutti
Descrizione: Complete thesis work
Dimensione
3.63 MB
Formato
Adobe PDF
|
3.63 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/10589/138899