多方位高分辨率SAR的三维目标自动重建(一) 目标像的检测

Automatic reconstruction of 3D objects from multi-aspect high-resolution SAR Part I:object image extraction

用多方位SAR图像融合获取立体目标信息是高分辨率SAR的一项关键应用技术。以t长方体为建筑物目标模型,提出从多方位米级分辨率SAR图像自动重建三维立体建筑物目标的方法。全文分两部分,第一部分介绍从米级分辨率SAR图像检测提取建筑物目标像的方法。先用恒虚警率(CFAR)检测器检测各方位SAR图像中建筑物目标像的边缘,然后用脊滤波细化边缘,再根据条状建筑物像的直线特征,用分块平行线Hough变换检测平行线段,最终检测得到平行四边形的建筑物目标像。文中用虚拟场景模拟的SAR图像和机载全极化X波段Pi-SAR图像做试验,提取了建筑物目标像。第二部分以本文为基础进行三维建筑物目标群的重建。

Reconstruction of three-dimensional stereo objects from multi-aspect high-resolution SAR images is of great importance for SAR technology applications. In this paper, a building object is simply modeled as a cuboid, and automatic reconstruction of three-dimensional （3D） objects from multi-aspect metric-resolution SAR images is developed. In Part I, a method for building image extraction from metric-resolution SAR image is studied. The edge detector of constant false alarm rate （CFAR） is first employed to detect the edge of polarimetric SAR image. Due to the linear feature of building images, parallel Hough transform technique is introduced to further extract the parallelogram-like images of the building wall scatterings. Image extraction is successfully performed using both simulated image and X-band Pi-SAR images. Reconstruction of building objects from their multi-aspect images will be discussed in Part II.