Semi-Automatic SAR Image Land Cover Labeling Pipeline

Abstract

This paper deals with the semi-automatic SAR(Synthetic Aperture Radar) image land-cover labeling system. The term semi-automatic refers to a sequential process in which the input SAR image is automatically labeled through the GPS information and the open API provided by the nation, and the user can manually modify the labeling information with the annotation tool. Since there may be some differences from the actual land information, the label is finally corrected by manually modifying it with annotation tool. The semi-automatic labeling system proposed in this paper is made to label water/forest/farmland/building by using SAR image and latitude/longitude information as input, but it can also be used to develop a land-related image analysis system with different purposes. The entire pipeline system is tested with the Sentinel-1A(S1A) image, and found that it works well in an acceptable range.