基于改进变差函数的高分辨率SAR图像建筑区提取  

Extraction of Building Areas from High-resolution SAR Images Based on Advanced Variogram Method

作　　者：王燕红[1,2] 程博[1] 尤淑撑[3] 武盟盟 

机构地区：[1]中国科学院遥感与数字地球研究所,北京100094 [2]中国科学院大学,北京100049 [3]中国土地勘测规划院,北京100035

出　　处：《遥感信息》2014年第2期1-6,共6页Remote Sensing Information

基　　金：国家科技合作专项项目(2012DFA20930)

摘　　要：变差函数是一种简单有效的纹理提取算法,在建筑区提取方面有较好效果,但是易受SAR影像噪声影响、大窗口易造成提取结果边缘模糊、稳健性差。为此本文提出了一种以中值滤波方法为基础的改进算法,不仅保留了变差函数在区分建筑区与非建筑区的优势,而且使窗口取值更加灵活,改善了高分辨率SAR提取建筑区时受强反射点、噪声干扰的缺点。对新方法和现有方法进行了分析和对比,结果表明,改进的方法虚警率及漏检率更低总体误差更小,能够更完整地提取建筑区。Variogram is a simple and effective method of exacting texture and achieves a satisfied result in exacting building areas.But the existing variogram model is not robust against the speckle noise in SAR images and the edge of exacting area is apt to fuzzy.To extract building areas from SAR images,this paper proposes an advanced variogram approach based on median filter.The new method not only can keep the advantage of discriminate the building areas and non-building areas,but also decrease the false alarm caused by the strong reflect spot and speckle noise.The experiments between the existing method and the proposed one is compared and analyzed,the results show that the advanced method has a lower false alarm rate,a lower missing rate,and a smaller overall error.

关 键 词：SAR图像 建筑区 改进变差函数 中值滤波 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]