论文部分内容阅读
针对高分辨率极化合成孔径雷达(SAR)影像解译中面向像素方法难以充分利用影像信息的问题,提出一种基于超像素与Span-Pauli分解的非监督分类方法.利用分水岭方法易于过分割的特点,将分水岭分割得到的特征相似、空间相邻的像素集合视为超像素,并作为分类算法的基本处理单元.利用极化参数Span及Pauli基对极化SAR数据进行迭代分类,以实现适用于高分辨率SAR影像的非监督分类.采用X波段高分辨率SAR数据进行实验,结果表明:基于超像素的分类方法能有效抑制分类结果中的椒盐现象,将总体精度提高到了73.81%;在准确区分水体、道路的基础上,提高了复杂的农作物类别的分类精度.
Aiming at the problem that the pixel-oriented method can not make full use of the image information in the high-resolution polarimetric SAR imaging interpretation, an unsupervised classification method based on superpixel and Span-Pauli decomposition is proposed. The watershed method is easy to over-segmentation , The characteristics of watershed segmentation are similar, and the spatially adjacent pixel sets are considered as superpixels and serve as the basic processing unit of the classification algorithm.Polarization SAR data are iteratively classified using polarization parameters Span and Pauli to achieve Which is suitable for unsupervised classification of high resolution SAR images.Experimental results using X - band SAR data show that the superpixel - based classification method can effectively suppress salt and pepper in the classification results and increase the overall accuracy to 73.81%. On the basis of accurately distinguishing water bodies and roads, the classification accuracy of complex crop types is improved.