南极典型冰架冰裂隙数据集（2015、2016、2020）

我们提出利用U-net网络进行冰裂隙识别探测的算法，可以实现南极冰裂隙的自动化探测。基于Sentinel-1 EW 1月、2月的数据，为了抑制SAR图像的相干斑噪声，选择Probabilistic Patch-Based Weights（PPB）算法进行滤波，然后选择具有代表性的样本输入U-net网络进行模型训练，根据训练的模型进行冰裂隙的预测。以南极5个典型冰架（Amery、Fimbul、Nickerson、Shackleton、Thwaiters）为例分类结果的平均准确率可达94.5%，其中裂隙区域的局部准确率可达78.6%，召回率为89.4%。We propose an algorithm for Antarctic ice crack detection based on U-net, enabling automated ice crack identification. Based on Sentinel-1 EW data from January and February, the Probabilistic Patch-Based Weights (PPB) algorithm was selected for filtering to suppress speckle noise in SAR images. Representative samples were then fed into the U-net network for model training, and ice crack prediction was conducted using the trained model. Taking five typical Antarctic ice shelves (Amery, Fimbul, Nickerson, Shackleton, Thwaiters) as case studies, the average classification accuracy reaches 94.5%, with the local accuracy of crack regions up to 78.6% and the recall rate at 89.4%.

搜集汇总

背景与挑战

背景概述

该数据集是一个针对南极典型冰架（包括Amery、Fimbul、Nickerson、Shackleton和Thwaiters）的冰裂隙数据集，覆盖2015、2016和2020年。数据通过U-net网络和Sentinel-1 SAR图像实现自动化探测，经过PPB滤波处理，具有高准确率（平均94.5%），并以shpfile格式提供，空间分辨率为1m-10m，适用于冰川和遥感研究。