Flood Assessment using Multi-temporal Modis Satellite Images

Flood assessment using unsupervised techniques for multi-temporal MODIS satellite images is presented. Classical methods like mean shift algorithm is compared with artificial neural network method like self organizing maps for automatic water pixel identification and extraction. The extracted results help in identification of flooded and non-flooded places. Different methods are applied and comparative study of unsupervised methods involving mean shift and self organizing maps are carried-out. In order to evaluate the algorithmic performance, root mean square error and receiver operating characteristics is used as performance evaluation indices. The results reported will provide useful information for multi-temporal time series satellite image analysis which can be used for current and future research in disasters management.