通过大量的实验，研究了3种分形特征提取算法在高分辨率SAR图像分类中的适应情况和彼此的差异性。对分形特征提取的图像量化等级、窗口大小、点对数的确定以及坏值剔除数量进行了深入的研究。在此基础上，总结了适合高分辨率SAR图像地物分类的分形特征提取的一般规律。最后，提出了利用分形特征的统计量作为特征向量，运用模糊C均值算法对图像进行分类，并取得了很好的效果。

The adaption abihty and differences of three fractal feature abstract methods in the high resolution SAR image classification are researched by a great deal of experimentations, and a deep research of quantity numbers, size of windows, point pairs an.d bad values are made. Based what have done, the general laws of fractal feature abstract in the high resolution SAR image classification are summarized. At last, the paper proposes using the statistical features of the fractal feature, uses fuzzy C-clustering method,classes the images, a good effect can be gained.

TN957.52

雷达信号处理国防科技重点实验室资助项目（51431020205DZ2401）