Elevation Estimation-Driven Building 3-D Reconstruction From Single-View Remote Sensing Imagery

Building 3-D reconstruction from remote sensing images has a wide range of applications in smart cities, photogrammetry, and other fields. Methods for automatic 3-D urban building modeling typically employ multiview images as input to algorithms to recover point clouds and 3-D models of buildings. However, such models rely heavily on multiview images of buildings, which are time-intensive and limit the applicability and practicality of the models. To solve these issues, we focus on designing an efficient DSM estimation-driven reconstruction framework (Building3-D), which aims to reconstruct 3-D building models from the input single-view remote sensing image. Existing DSM estimation networks suffer from the imbalance between local and global features, which leads to oversmooth DSM estimates at instance boundaries. To address this issue, we propose a Semantic Flow Field-guided DSM Estimation (SFFDE) network, which utilizes the proposed concept of elevation semantic flow (ESF) to achieve the registration of local and global features. First, in order to make the network semantics globally aware, we propose an elevation semantic globalization (ESG) module to realize the semantic globalization of instances. Furthermore, in order to alleviate the semantic span of global features and original local features, we propose a local-to-global elevation semantic registration (L2G-ESR) module based on ESF. Our Building3-D is rooted in the SFFDE network for building elevation prediction, synchronized with a building extraction network for building masks, and then sequentially performs point cloud reconstruction and surface reconstruction (or CityGML model reconstruction). On this basis, our Building3-D can optionally generate CityGML models or surface mesh models of the buildings. Extensive experiments on ISPRS Vaihingen and DFC2019 datasets on the DSM estimation task show that our SFFDE significantly improves upon state-of-the-art, and
$\delta _{1}$
,
$\delta _{2}$
, and
$\delta _{3}$
metrics of our SFFDE are improved to 0.595, 0.897, and 0.970. Furthermore, our Building3-D achieves impressive results in the 3-D point cloud and 3-D model reconstruction process.