TerraSAR-X dual-pol time-series for mapping of wetland vegetationReportar como inadecuado

TerraSAR-X dual-pol time-series for mapping of wetland vegetation - Descarga este documento en PDF. Documentación en PDF para descargar gratis. Disponible también para leer online.

* Corresponding author 1 LETG - Rennes - Littoral, Environnement, Télédétection, Géomatique LETG - Littoral, Environnement, Télédétection, Géomatique 2 ECOBIO - Ecosystèmes, biodiversité, évolution Rennes 3 IETR - Institut d-Electronique et de Télécommunications de Rennes

Abstract : Mapping vegetation formations at a fine scale is crucial for assessing wetland functions and for better landscape management. Identification and characterization of vegetation formations is generally conducted at a fine scale using ecological ground surveys, which are limited to small areas. While optical remotely sensed imagery is limited to cloud-free periods, SAR time-series are used more extensively for wetland mapping and characterization using the relationship between distribution of vegetation formations and flood duration. The aim of this study was to determine the optimal number and key dates of SAR images to be classified to map wetland vegetation formations at a 1:10,000 scale. A series of eight dual-polarization TerraSAR-X images HH-VV was acquired in 2013 during dry and wet seasons in temperate climate conditions. One polarimetric parameter was extracted first, the Shannon entropy, which varies with wetland flooding status and vegetation roughness. Classification runs of all the possible combinations of SAR images using different k number of images subsets were performed to determine the best combinations of the Shannon entropy images to identify wetland vegetation formations. The classification runs were performed using Support Vector Machine techniques and were then analyzed using the McNemar test to investigate significant differences in the accuracy of all classification runs based on the different image subsets. The results highlight the relevant periods i.e. late winter, spring and beginning of summer for mapping vegetation formations, in accordance with ecological studies. They also indicate that a relationship can be established between vegetation formations and hydrodynamic processes with a short time-series of satellite images i.e. 5 dates. This study introduces a new approach for herbaceous wetland monitoring using SAR polarimetric imagery. This approach estimates the number and key dates required for wetland management e.g. restoration and biodiversity studies using remote sensing data.

Keywords : SVM SAR imagery vegetation formations change detection multitemporal image analysis

Autor: Julie Betbeder - Sébastien Rapinel - Samuel Corgne - Eric Pottier - Laurence Hubert-Moy -

Fuente: https://hal.archives-ouvertes.fr/


Documentos relacionados