Stereoscopic Building Reconstruction Using High-Resolution Satellite Image Data

Abstract

This paper proposes a new method for building detection and reconstruction from satellite images. In our approach, we propose to use divergence-based centroid neural network to carry out the grouping of 3D line segments. By grouping 3D line segments into the principal 3D lines which can constitute 3D rooftop model, the system significantly reduces the unnecessary line segments from low level feature extraction. Our grouping process consists of two steps. We carry out the first grouping process to group fragmented or duplicated 3D lines into the principal 3D lines which can be used to construct the rooftop model, and construct the groups of lines that are parallel each other in the second step. From the grouping result, 3D rooftop models are reconstructed by the final clustering process. We test the proposed method with high-resolution IKONOS stereo images. The experimental result proved the efficiency of the proposed method in the reconstruction of the rectilinear type of 3D rooftop model from high-resolution satellite imagery.