一种改进的全极化SAR影像面向对象分类方法

发布时间：2018-05-09 20:34

本文选题：全极化SAR影像 + 面向对象分类　；参考：《中国矿业大学学报》2015年05期

【摘要】：面向对象分类过程,首先对图像进行分割得到对象,然后将对象进行分类,分割效果直接影响最终分类精度.针对这一问题,提出一种改进的全极化合成孔径雷达(SAR)影像面向对象分类方法,在分类时首先通过计算各对象内部像元类别比例对对象进行判断,若所有类别比例均没有达到某个阈值,则认为此对象存在分割偏差,对其进行基于像元的分类,反之则进行面向对象分类,最后整合像元级和对象级分类结果.分类算法采用改进分类器动态选择法(ICDS)对差异性较大的3个基分类器Wishart、核-KNN和Wishart-KNN进行决策级融合.以AIRSAR,EMISAR的全极化SAR影像为数据进行分类实验.结果表明:改进算法充分利用了对象级和像素级分类的优点,从而得到高精度的分类结果,该算法具有良好的应用前景.
[Abstract]:In the process of object oriented classification, the object is segmented first, then the object is classified. The segmentation effect directly affects the final classification accuracy. In order to solve this problem, an improved object oriented classification method for fully polarized synthetic aperture radar (SAR) images is proposed. If the proportion of all categories does not reach a certain threshold, then the segmentation deviation of the object is considered, and the object based classification is carried out. Otherwise, the object oriented classification is carried out, and the results of pixel and object level classification are finally integrated. The improved classifier dynamic selection method (ICDSs) is used to fuse the three base classifiers, namely, Wishart, Kernel KNN and Wishart-KNN. The full polarization SAR image of AIRSARY EMISAR is used as the data for classification experiment. The results show that the improved algorithm makes full use of the advantages of object level and pixel level classification, and obtains high precision classification results. The algorithm has a good application prospect.
【作者单位】：国土环境与灾害监测国家测绘地理信息局重点实验室中国矿业大学;宜春学院数学与计算机学院;江苏省地理信息技术重点实验室南京大学;
【基金】：国家自然科学基金项目(41171323) 中国地质调查局地质调查工作项目(1212011120229,12120115040601) 江苏高校优势学科建设工程项目(SZBF2011-6-B35)
【分类号】：TN957.52

【相似文献】