基于聚类的合成孔径雷达图像分割算法研究

Study on Synthetic Aperture Radar Image Segmentation Algorithm Based on Clustering

针对SAR图像的分割问题,对K均值聚类算法进行了研究,分析了自适应动态K均值聚类算法,改进了最大适应度函数聚类的样本分离准则。毫米波SAR图像分割表明,对于城区建筑及路、桥场景的分割,改进算法比基本的K均值聚类算法、动态K均值聚类算法及自适应动态K均值聚类算法性能要好,指标更优。在运算效率上,改进算法与自适应动态K均值聚类算法效率相当。

Aiming at SAR image segmentation issue, K mean value clustering algorithm is studied; adaptive dynamic K mean value clustering algorithm is analyzed, and sample separation criteria of maximum fitness function clustering is im- proved. MMW SAR image segmentation shows that the improved algorithm has better performance and more excellent in- dex than that of basic K mean value clustering algorithm, dynamic K mean value clustering algorithm and adaptive dy- namic K mean value clustering algorithm for segmentation of urban area building, road, bridge scenes. The improved al- gorithm has correspondent efficiency with adaptive dynamic K mean value clustering algorithm in operation efficiency.