针对全极化高分辨率合成孔径雷达(SAR)多噪声的特点,选择合适的海冰SAR影像纹理特征量,并基于凝聚层次聚类的思想,提出一种全极化SAR海冰分割方法。将该方法的结果与K-Means、迭代自组织聚类(ISODATA)和模糊C均值算法等经典分割方法相比较,发现该方法的碎斑明显减少,且分割结果较为准确,证明了该方法的有效性。 Aiming at the characteristics of all-polarimetric high-resolution synthetic aperture radar (SAR) multi-noises, we choose the appropriate feature quantity of sea ice SAR image texture and propose a fully polarimetric SAR sea ice segmentation method based on the idea of ​​cohesive hierarchical clustering. Comparing the result of this method with classical segmentation methods such as K-Means, ISODATA and fuzzy C-means algorithm, it is found that the method reduces the number of patches and the segmentation result is more accurate. Effectiveness.