Technical Report: Towards Spatial Feature Regularization in Deep-Learning-Based Array-SAR Reconstruction

TL;DR

This work tackles the challenge of artifacts and edge fragmentation in DL-based TomoSAR urban reconstruction by integrating spatial feature regularization. It develops two model-driven frameworks (iterative reconstruction with enhancement and light reconstruction with enhancement) and four networks (tomo-IRENet-TV, tomo-IRENet-U, tomo-LRENet-biU, tomo-LRENet-LSTM) that exploit sharp edges and regular geometric shapes through 3D-TV and advanced slice fusion methods. Evaluations on simulated urban models and public SARMV3D-1.0 data demonstrate improved edge preservation, structural integrity, and robustness, with the LRENet-LSTM variant achieving the best accuracy-efficiency trade-off. The results highlight the practical potential of spatial priors for large-scale urban 3D mapping with TomoSAR, while outlining avenues for generalization, noise-robustness, and non-local modeling in future work.

Abstract

Array synthetic aperture radar (Array-SAR), also known as tomographic SAR (TomoSAR), has demonstrated significant potential for high-quality 3D mapping, particularly in urban areas.While deep learning (DL) methods have recently shown strengths in reconstruction, most studies rely on pixel-by-pixel reconstruction, neglecting spatial features like building structures, leading to artifacts such as holes and fragmented edges. Spatial feature regularization, effective in traditional methods, remains underexplored in DL-based approaches. Our study integrates spatial feature regularization into DL-based Array-SAR reconstruction, addressing key questions: What spatial features are relevant in urban-area mapping? How can these features be effectively described, modeled, regularized, and incorporated into DL networks? The study comprises five phases: spatial feature description and modeling, regularization, feature-enhanced network design, evaluation, and discussions. Sharp edges and geometric shapes in urban scenes are analyzed as key features. An intra-slice and inter-slice strategy is proposed, using 2D slices as reconstruction units and fusing them into 3D scenes through parallel and serial fusion. Two computational frameworks-iterative reconstruction with enhancement and light reconstruction with enhancement-are designed, incorporating spatial feature modules into DL networks, leading to four specialized reconstruction networks. Using our urban building simulation dataset and two public datasets, six tests evaluate close-point resolution, structural integrity, and robustness in urban scenarios. Results show that spatial feature regularization significantly improves reconstruction accuracy, retrieves more complete building structures, and enhances robustness by reducing noise and outliers.

Figures (32)

• Figure 1: Simplified illustration of the study.
• Figure 2: Simplified illustration of notions in the study.
• Figure 3: TomoSAR observation geometrical relationship.
• Figure 4: Examples of spatial feature loss. (a) Optical image and 2D SAR image of two regular buildings in the scene. (b) Corresponding reconstruction results, showing holes in the surface of the buildings. (c) Optical image and 2D SAR image of a crane in the scene. (d) Corresponding reconstruction results, revealing a fragmented crane arm.
• Figure 5: Examples of spatial features in radar images. (a) Optical image of the scene. (b) 3D Lidar point cloud of the scene. (c) Radar image of the scene from one angle, with line structures marked in red. (d) Radar image of the scene from another angle, with line structures marked in red, and a regular geometric shape highlighted with a green box.