An algorithm based on a support vector machine (SVM) is proposed for hydrometeor classification. The training phase is driven by the output of a fuzzy logic hydrometeor classification algorithm, i.e., the most popular approach for hydrometer classification algorithms used for ground-based weather radar. The performance of SVM is evaluated by resorting to a weather scenario, generated by a weather model; the corresponding radar measurements are obtained by simulation and by comparing results of SVM classification with those obtained by a fuzzy logic classifier. Results based on the weather model and simulations show a higher accuracy of the SVM classification. Objective comparison of the two classifiers applied to real radar data shows that SVM classification maps are spatially more homogenous (textural indices, energy, and homogeneity increases by 21% and 12% respectively) and do not present non-classified data. The improvements found by SVM classifier, even though it is applied pixel-by-pixel, can be attributed to its ability to learn from the entire hyperspace of radar measurements and to the accurate training. The reliability of results and higher computing performance make SVM attractive for some challenging tasks such as its implementation in Decision Support Systems for helping pilots to make optimal decisions about changes in the flight route caused by unexpected adverse weather.
Roberto, N., Baldini, L., Adirosi, E., Facheris, L., Cuccoli, F., Lupidi, A., et al. (2017). A support vector machine hydrometeor classification algorithm for dual-polarization radar. ATMOSPHERE, 8(8) [10.3390/atmos8080134].
A support vector machine hydrometeor classification algorithm for dual-polarization radar
GARZELLI, ANDREA
2017-01-01
Abstract
An algorithm based on a support vector machine (SVM) is proposed for hydrometeor classification. The training phase is driven by the output of a fuzzy logic hydrometeor classification algorithm, i.e., the most popular approach for hydrometer classification algorithms used for ground-based weather radar. The performance of SVM is evaluated by resorting to a weather scenario, generated by a weather model; the corresponding radar measurements are obtained by simulation and by comparing results of SVM classification with those obtained by a fuzzy logic classifier. Results based on the weather model and simulations show a higher accuracy of the SVM classification. Objective comparison of the two classifiers applied to real radar data shows that SVM classification maps are spatially more homogenous (textural indices, energy, and homogeneity increases by 21% and 12% respectively) and do not present non-classified data. The improvements found by SVM classifier, even though it is applied pixel-by-pixel, can be attributed to its ability to learn from the entire hyperspace of radar measurements and to the accurate training. The reliability of results and higher computing performance make SVM attractive for some challenging tasks such as its implementation in Decision Support Systems for helping pilots to make optimal decisions about changes in the flight route caused by unexpected adverse weather.File | Dimensione | Formato | |
---|---|---|---|
2017-Atmosphere-08-00134.pdf
accesso aperto
Descrizione: 2017-Atmosphere
Tipologia:
PDF editoriale
Licenza:
Creative commons
Dimensione
5.84 MB
Formato
Adobe PDF
|
5.84 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/11365/1012681