Abstract: Traditional coastline extraction methods only aim at a certain type of coast effectively.For areas containing a variety of types of coastal areas, it is necessary to cut the image classification, extract the coastline, and stitching images to extract the complete coastline.The process is not only complicated, but also difficult to divide different types of coastal areas strictly.Therefore, a new coastline method suitable for mixed coastal area is proposed-automatic coastline extraction method for simulating sea water rising process.Firstly, a classification method based on spectral difference decision tree (SDDT) is applied to classify remote sensing images.According to the difference of the types of land, the two categories are marked as "tidal area" and "non tidal area".Secondly, by calculating the area of each patch area, the area blocks less than a preset threshold are defined as the noise region, and all noise areas are removed.Finally the maximum rectangle filling algorithm(MRFA) is used to simulate the flood tide to gradually fill the "tidal area" and extract the coastline.The coastline extracted by this method is continuous, and is not affected by a small number of misclassified patches in the classification map.