摘要:
以Radarsat-2为例,提出了一种利用多极化SAR影像并结合后向散射模型提取建筑物高度的方法.以北京城区为实验区,首先,分析了SAR影像中建筑物二次散射所对应的亮线连通区域,并统计出该区域对雷达后向散射截面的贡献量;然后,基于平行六面体假设,将建筑物主长度及其与雷达方位向的夹角定量化,并给出计算不同极化散射矢量的方法;最后,利用几何光学-物理光学(geometrical optics-physical optics,GO-PO)模型一阶近似解的后向散射模型估计建筑物高度,并通过比较多个局部训练区提取结果,探讨不同极化信息的提取效果并给出最优极化组合.实验结果表明,通过极化特征组合比仅利用单极化信息提取建筑物高度的精度更高,81.43%建筑物误差小于5m,均方根误差4.45m,与ASTER GDEM相关系数为0.9095,提取结果可靠.
Abstract:
With Radarsat-2 as an example, a method of building height extraction from multi-polarization SAR imagery was proposed based on backscattering model.First, the connected component of double-scattering of the buildings in the image was analyzed and its contribution to radar cross section was got simultaneously, which was a case study in urban areas of Beijing.Second, different polarization-scattering vectors were calculated based on parallelepiped-assumption, which was supported by quantifying buildings' correlation length and the angle between radar's azimuth and buildings' main direction.Finally, optimal polarized combination was utilized, which was extracted by using backscattering model from the solution of geometrical optics-physical optics(Go-Po)first-order approximation and comparing the results from different regionally training areas at the same time.The experimental results show that optimal polarized combination produces much less errors than single-polarization imagery in extracting the height of entire experimental area, with 81.43% of buildings having errors less than 5 meters, root mean square error being 4.45, and correlation coefficient with ASTER GDEM being 0.909 5, which proves that the result in height extraction is reliable.
下载: