SAR样本数据集

数据主编：王智睿（目标认知与应用技术国家级重点实验室，中国科学院空天信息创新研究院） AIR-PolSAR-Seg-2.0是一套面向大规模复杂场景的极化SAR地物分类数据集，旨在推动极化SAR散射特性分析、地物分类等核心技术的发展与突破。该数据集由三景不同区域的高分三号L1A级复数SAR影像构成，空间分辨率为8 m，包含HH、HV、VH和VV共4种极化方式，涵盖水体、植被、裸地、建筑、道路、山脉等6类典型地物。数据集具有场景复杂规模大、强弱散射多样、边界分布不规则、类别尺度多样以及样本分布不均衡的特点。此外，本数据集将三景完整的SAR影像裁剪为24,672张512×512像素的切片，方便开展深度学习类方法的实验验证。 详细使用说明请参考：大规模复杂场景极化SAR地物分类数据集(AIR-PolSAR-Seg-2.0)下载使用说明.pdf

Data Editor: Zhirui Wang (National Key Laboratory of Target Recognition and Application Technology, Aerospace Information Research Institute, Chinese Academy of Sciences). AIR-PolSAR-Seg-2.0 is a polarimetric SAR (PolSAR) land cover classification dataset targeting large-scale complex scenarios, aiming to promote the development and breakthroughs of core technologies including polarimetric SAR scattering characteristic analysis and land cover classification. This dataset consists of three GF-3 L1A-level complex SAR images from different regions, with a spatial resolution of 8 m, featuring four polarization modes: HH, HV, VH, and VV, and covers six typical land cover types including water bodies, vegetation, bare land, buildings, roads, and mountains. This dataset exhibits the characteristics of large-scale and complex scenes, diverse strong and weak scattering signals, irregular boundary distributions, varied category scales, and imbalanced sample distribution. Additionally, the three complete SAR images in this dataset have been cropped into 24,672 slices of 512×512 pixels, facilitating experimental validation of deep learning-based methods. For detailed usage instructions, please refer to the document Download and Usage Instructions for Large-Scale Complex Scene Polarimetric SAR Land Cover Classification Dataset (AIR-PolSAR-Seg-2.0).pdf