利用多尺度Hough变换提取高分辨率SAR图像建筑物L型结构
An Approach to Extract L-shaped Feature of Building from VHR SAR Image Using Multi-scale Hough Transform
-
摘要: 提出了一种利用多尺度Hough变换从高分辨率SAR图像提取建筑物L型结构的方法。针对高分辨率SAR图像建筑物L型结构的特点,建立了建筑物L型结构简化几何模型,并采用从粗到精的思路利用多尺度Hough变换提取建筑物L型结构方向线。并提出了一种L型结构组合度函数对提取直线进行编组,确定建筑物L型结构的方向和拐点。最后,采用基于扫描线的方法计算L型结构线宽,得到完整规则的建筑物L型结构。多幅真实机载高分辨率SAR图像实验结果表明,本方法可以有效地提取高分辨率SAR图像建筑物L型结构,提取结果与实际位置吻合较好。Abstract: An approach to extract the L-shaped features of the building from very high resolution(VHR) synthetic aperture radar(SAR) images is proposed.Firstly,a simple geometric model for the L-shaped feature is defined,and an approach based on multi-scale Hough Transform is carried out to extract orientations.And then,a new combination degree function is employed to group linear structures into L-shaped features,and corners of them are determined.Finally,the line width of L-shaped feature is estimated using the imporved scan-line technology.The experimental results on several real airborne VHR SAR images prove that the proposed approach extract the L-shaped features of buildings from VHR SAR images effectively,and the results are in good agreement with the ground truth.
下载:
