Flooded area extraction & damage assessment from multispectral remote sensing imagery based on image classification [Thesis (M.Sc.)]
2012
Alouane Yosra
Floods are water-related natural disaster which affect and threat all the aspects of human life. Loss of property as well as displacement and economic degradation are obvious consequences following flood events not to mention the loss of precious human lives. Mapping floods is a key tool to improve flood management and to mitigate its catastrophic effects. In this context, the present thesis aims to explore the use of optical remote sensing imagery in performing flooded areas mapping and damage assessment. The specific thesis objectives include: 1) to perform a comparative analysis of a range of classification approaches when combined with medium spatial resolution multispectral imagery from Landsat TM imagery for mapping flooded area, 2) to evaluate the impact of extra spectral information (band ratios and/ or indices) to improving the flooded area extraction from TM imagery and 3) to perform an assessment of the flooding damage extent exploiting freely-distributed operational land use/land cover products. For this purpose, different classification methods for extracting the flooded areas are parameterized and subsequently applied to a Landstat TM imagery acquired over the region of Evros River (Northern Greece) during a flooding event occurred in 2010. In particular the pixel-based classifiers Support Vector Machines (SVM), Artificial Neural Networks (ANN) and object-based classification are implemented and compared.
显示更多 [+] 显示较少 [-]