Vegetation fuel type classification using optimised synergy of sentinel data and texture feature

This paper aims to map vegetation fuel types using a combination of remote sensing data in a complex and diverse plant cover of central Portugal. This study employs Sentinel-1 (S1) and Sentinel-2 (S2) bands, digital elevation model (DEM), and vegetation indices (VIs). Gray-level co-occurrence matrix (GLCM) texture features were generated for the first three principal components (PCs), after applying principal component analysis (PCA) on the S2A spectral bands. First, the fuel type classes based on the FirEUrisk Hierarchical Multipurpose Fuel Classification System (FirEUrisk-HMFCS) were established, then the Random Forest (RF) classifier was employed. Moreover, the feature selection method was used to improve classifier performance. The proposed methodology increased the overall accuracy (OA) of the classification up to 91.89% due to the consideration of the feature selection in the synergy of multisource data, and the role of texture feature data.