2020-2021年南北极sentinel-1冰裂隙产品数据集

提出利用U-net网络进行冰裂隙识别探测的算法，可以实现格陵兰冰盖典型冰川冰裂隙的自动化探测。基于Sentinel-1 IW每年7、8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。以格陵兰2个典型冰川（Jakobshavn、Kangerdlussuaq）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%。属性表中type字段1代表横向裂隙，2代表外展裂隙，3代表冰川上壁，4代表雁行裂隙，5代表冰瀑，7代表冰裂。An algorithm for ice fracture detection using the U-Net network is proposed, enabling automated detection of typical glacier ice fractures over the Greenland Ice Sheet. Based on Sentinel-1 IW data acquired in July and August each year, the Probabilistic Patch-Based Weights (PPB) algorithm was selected for filtering to suppress speckle noise in SAR images. Then, representative samples were selected and input into the U-Net network for model training, and ice fracture prediction was performed using the trained model. Taking two typical glaciers in Greenland (Jakobshavn and Kangerdlussuaq) as examples, the average accuracy of the classification results reaches 94.5%, with the local accuracy of the fracture region reaching 78.6% and the recall rate at 89.4%. In the attribute table, the "type" field: 1 represents transverse cracks, 2 represents extensional fractures, 3 represents the upper wall of glaciers, 4 represents en echelon cracks, 5 represents icefalls, and 7 represents ice crevasses.

搜集汇总

背景与挑战

背景概述

该数据集基于2020-2021年Sentinel-1卫星每年7、8月的影像，采用PPB算法滤波和U-net机器学习方法，实现了格陵兰冰盖典型冰川冰裂隙的自动化探测，平均准确率达94.5%。数据集包含51个shp格式矢量文件，属性表中定义了横向裂隙、外展裂隙等6种冰裂隙类型，适用于冰裂隙识别研究。