An alternate iteration and matching technique for parameter estimation of gable-roof buildings from high-resolution SAR images

Abstract

In urban areas, the reconstruction of gable-roof buildings from high-resolution synthetic aperture radar (SAR) images is more challenging compared to simple parallelepiped buildings. In this paper, an alternate iteration and matching technique is proposed for height and roof inclination angle estimation of gable-roof buildings. First, the different appearances of this building type in SAR images are analyzed. Then, for height and roof inclination angle estimation, an alternate iteration and matching technique is proposed with a nested loop structure. Image matching by maximizing the mutual information between the true and simulated SAR images is performed in the inner loop, while alternate iteration is carried out in the outer loop. This structure is qualified for multi-parameter estimation and achieves a balance between estimation accuracy and computational complexity. Finally, numerical experiment and comparison with previous methods are performed to show the effectiveness and efficiency of our method.