Automatic extraction of coastline based on Google Earth engine

Abstract

Traditional remote sensing image-based coastline extraction is limited by data volume and processing speed, and the extracted coastline is susceptible to noise. Therefore, this paper proposes a method based on the Google Earth Engine geospatial platform, which combines threshold segmentation, the Otsu method, and morphological algorithms. First, remote sensing images are preprocessed on the platform, and the Normalized Difference Water Index (NDWI) is calculated. Then, the Otsu method is used to calculate the NDWI threshold for water-land segmentation, resulting in a binary water-land image. Next, morphological methods are employed to remove noise points, fill holes, and extract the coastline. The proposed method was tested using Hainan Island based on synthetic images from August 2021, and the average accuracy of land-water segmentation reached 98.49%. The results demonstrate that this method can accurately extract coastlines and significantly improve extraction efficiency. This research method provides a reference for the automatic extraction of coastlines over large areas and long time series.