Application of machine learning algorithms for forest monitoring by satellite SAR data

Abstract

Hurricanes cannot be controlled by humans, and are increasing in number day by day. Hurricanes are responsible for large-scale loss of life and assets. They appear within a very short time, and are unstoppable by people once they have started. Therefore, for effective risk management, damage should be assessed after the disaster. Satellite radar images (SAR) have advantages because the radar sensor can work in all-weather conditions, is not affected by clouds, so the use of SAR imagery is useful in identifying damage and loss of assets. In our project, we selected the Jesenik area, because a hurricane occurred on March 17, 2018, and there were substantial losses in forest areas in particular where there are many homes. Sentinel 1 (S1) images have been used, some from the pre-disaster period and others from the post-disaster period. Backscatter values are analyzed in both images. It is expected that the difference between post-disaster images will be greater than the pre-disaster images. (In case of extensive damage). After applying the segmentation algorithm, we find out the segmentation of different area. The results show a polygon for damages detected by SAR images.