*2.6. Determination of Segmentation Threshold Based on OTSU*

The automatic determination of the object-oriented segmentation threshold affects the final classification result and the automatic process of sea ice range extraction. In this study, the OTSU method was used to automatically determine the threshold. The principle of the OTSU method is to continuously iteratively determine an optimal threshold to maximize the variance between the target and the background. Before conducting the OTSU threshold segmentation, the terrestrial mask pixels need to be removed. This is because the OTSU determines the segmentation threshold based on histogram statistics. Land pixels will affect the structure of the histogram and cause the predicted threshold to deviate. After removing the land pixels, the double peaks in the histogram are clearer. This improves the accuracy of the threshold.

#### *2.7. Accuracy Verification*

In order to better evaluate the robustness and applicability of the method developed in this study, the proposed method was compared with the extraction results of the Support Vector machine (SVM) and K-Means methods, and the three methods were applied to GF1, Landsat-8, and Sentinel-2 images. In order to quantitatively evaluate the accuracy of the sea ice extraction, ArcGIS was used to randomly generate 800 test points in the sea area, and the type was marked based on a planet satellite image with a resolution of 3 m. To ensure that the test points were evenly distributed in the study area and that all types of sea ice and seawater were present, their total accuracy and kappa coefficient (κ) were calculated.
