Towards Automatic SAR-Optical Stereogrammetry over Urban Areas using Very High Resolution Imagery

TL;DR

This paper explores the feasibility of combining SAR and optical VHR imagery for 3D urban reconstruction, analyzing accuracy, proposing a matching strategy, and demonstrating meter-level positioning with real data.

Contribution

It introduces a novel stereogrammetric approach for SAR-optical data fusion in urban areas, including a robust matching strategy and an analysis of accuracy expectations.

Findings

3D positioning accuracy in the meter range.

Feasibility of SAR-optical stereogrammetry with VHR imagery.

Matching heterogeneous multi-sensor data remains challenging.

Abstract

In this paper we discuss the potential and challenges regarding SAR-optical stereogrammetry for urban areas, using very-high-resolution (VHR) remote sensing imagery. Since we do this mainly from a geometrical point of view, we first analyze the height reconstruction accuracy to be expected for different stereogrammetric configurations. Then, we propose a strategy for simultaneous tie point matching and 3D reconstruction, which exploits an epipolar-like search window constraint. To drive the matching and ensure some robustness, we combine different established handcrafted similarity measures. For the experiments, we use real test data acquired by the Worldview-2, TerraSAR-X and MEMPHIS sensors. Our results show that SAR-optical stereogrammetry using VHR imagery is generally feasible with 3D positioning accuracies in the meter-domain, although the matching of these strongly hetereogeneous multi-sensor data remains very challenging. Keywords: Synthetic Aperture Radar (SAR), optical images, remote sensing, data fusion, stereogrammetry