Detecting Urban Floods with Small and Large Scale Analysis of ALOS-2/PALSAR-2 Data

When a large-scale flood disaster occurs, it is important to identify the areas in short time order effectively support affected afterwards. Synthetic Aperture Radar (SAR) promising for detection. A number of change detection methods have been proposed detect flooded with pre- and post-event SAR data. However, remains difficult built-up due complicated scattering microwaves. To solve this issue, paper we propose idea analyzing local changes data as well larger-scale changes, which may improve accuracy detecting floods areas. Therefore, aimed at evaluating effectiveness multi-scale analysis using ALOS-2/PALSAR-2 First, several features were determined by calculating standard deviation images, difference correlation coefficient images sizes kernels. Then, segmentation on both small large scales was applied image calculated explanatory variables each segment. Finally, machine learning models tested their performance comparing small-scale approach approach. Ten-fold cross-validation used validate model, showing that highest offered AdaBoost improved F1 Score from 0.89 0.98 analysis. The main contribution manuscript that, our results, can be inferred shows better quantitative