基于SVM的高分辨率SAR图像建筑物自动提取

通过遥感影像直接提取出建筑物是进行城市土地利用调查和土地执法检察的有效手段。文章利用高分辨率的SAR数据,提出了基于SVM和高分辨率SAR图像的建筑物自动提取方法。该方法利用SAR图像对建筑物信息反映强烈的特点,选择一系列建筑物和非建筑物像元作为训练样本,利用SVM方法构建每个像元的属于建筑物可信度,设定95%的可信度界定为建筑物像元,据此识别出建筑物的位置。以温江地区的COSMO-SkyMed数据进行试验表明,该方法建筑物识别的抽样精度达到93%以上,显示出高分辨率SAR图像在城市土地利用研究中的巨大潜力。

Automatic Building Detection from High Resolution SAR Based on SVM Method

Using remote sensing image to detect buildings directly is a wonderful method for city land use management and checking lawless land use.This paper has proposed a new method of automatically building detection based on high resolution SAR image.The method makes use of high resolution SAR image and detects buildings by Support Vector Machine.First,we select some building pixels and other class pixels as samples.Then,we use SVM method to calculate the reliability of each pixel belongs to building pixel or not.Last,we take 95% as a divided point to identify building's information.A result with above 93% accuracy has been achieved while applying this technique to Wenjiang city using COSMO-SkyMed SAR image,which has demonstrated prospecting applications of high resolution SAR image for urban land use management.