结合凝聚层次聚类的极化SAR海冰分割

Segmentation method for agglomerative hierarchical-based sea ice types using polarimetric SAR data

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。

This paper investigates the characteristics of high resolution full-polarization Synthetic Aperture Rader （SAR） data combined with the selected texture characteristic, and uses the theory of the agglomerative hierarchical clustering algorithm to present a sea ice segmentation method using full-polarimetric SAR data. By comparing the result of this method with the results of the K-Means, ISODATA, and fuzzy c-means methods, we verified that our method can reduce the number of small patches and realize good segmentation accuracy. The results verify the validity of this method.