OBIA for combining LiDAR and multispectral data to characterize forested areas and land cover in a tropical region
2012
Dupuy, S. | Laine, G. | Tormos, T.
Prioritizing and designing forest restoration strategies requires an adequate survey to inform on the status (degraded or not) of forest types and the human disturbances over a territory. Very High Spatial Resolution (VHSR) remotely sensed data offers valuable information for performing such survey. We present in this study an OBIA methodology for mapping forest types at risk and land cover in a tropical context (Mayotte Island) combining LiDAR data (1 m pixel), VHSR multispectral images (Spot 5 XS 10 m pixel and orthophotos 0.5 m pixel) and ancillary data (existing thematic information). A Digital Canopy Model (DCM) was derived from LiDAR data and additional information was built from the DCM in order to better take into account the horizontal variability of canopy height: max and high Pass filters (3m x 3m kernel size) and Haralick variance texture image (51m x 51m kernel size). OBIA emerges as a suitable framework for exploiting multisource information during segmentation as well as during the classification process. A precise map (84% total accuracy) was obtained informing on (i) surfaces of forest types (defined according to their structure, i.e. canopy height of forest patches for specific type); (ii) degradation (identified in the heterogeneity of canopy height and presence of eroded areas) and (iii) human disturbances. Improvements can be made when discriminating forest types according to their composition (deciduous, evergreen or mixed), in particular by exploiting a more radiometrically homogenous VHSR multispectral image.
Afficher plus [+] Moins [-]Mots clés AGROVOC
Informations bibliographiques
Cette notice bibliographique a été fournie par National Institute for Agricultural Research
Découvrez la collection de ce fournisseur de données dans AGRIS