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

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。

An approach to extract the L-shaped feation （VHR） synthetic aperture radar （SAR） im eatures ages is of the building proposed. Firs rom very high resolu- y, 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 func- tion 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 ima- ges effectively, and the results are in good agreement with the ground truth.