Three-dimensional reconstruction for space targets with multistatic inverse synthetic aperture radar systems

Yue Zhao, Lei Zhang, Bo Jiu, Hongwei Liu, Zhenfang LiEURASIP J. Adv. Signal Process 2019 · DOI

Introduction

Accurately identifying space targets and their structural components, such as solar panels or cabins, is a cornerstone of space situational awareness. While 2D Inverse Synthetic Aperture Radar (ISAR) provides high-resolution imaging, these snapshots are highly sensitive to target motion and lack critical depth information. Traditional 3D reconstruction methods, such as 3D-FT or Interferometric ISAR (InISAR), often suffer from prohibitive computational costs or impractical hardware requirements for distant targets.

Furthermore, existing factorization methods typically require "cross-range scaling"—a precise knowledge of the target's rotation during observation—to recover actual physical dimensions. This paper proposes a novel multistatic ISAR algorithm that leverages orbital mechanics and epipolar geometry to reconstruct true 3D geometry without the need for complex scaling procedures.

Methodology

The proposed algorithm bypasses traditional bottlenecks by integrating orbital priors into the reconstruction workflow. The process is divided into three primary stages:

  1. Establishing the Projection Matrix: Instead of estimating geometry from radar data alone, the method utilizes the target's known Keplerian orbital elements. By transforming coordinates from an Earth-centered frame to the satellite's body frame, the system estimates precise observation angles for each sensor, creating a projection matrix that maps 3D coordinates to 2D range-Doppler observables.
  2. Scattering Center Association: To solve the "matching problem" across different radar perspectives, the authors adapt epipolar geometry from computer vision. A point in one radar image corresponds to a theoretical "epipolar line" in another. A cost function minimizes the distance between detected points and these lines to ensure accurate matching.
  3. Optimization: The association is framed as a 0–1 linear programming assignment solved via the Jonker-Volgenant algorithm for efficiency. Final 3D coordinates are then derived using a least-squares approach.
The block diagram of the algorithm illustrates the integration of orbital elements and radar image processing to achieve 3D reconstruction.

Key Results and Discussion

The algorithm was validated using simulated point-scatterer models and complex Hubble Space Telescope (HST) models. A major breakthrough of this research is the robustness of the Epipolar Geometry Projective (EGP) association method.

As shown in the performance analysis, the EGP method maintains a nearly 100% correct association rate even when the azimuth angle difference between sensors reaches 30 degrees. This significantly outperforms standard techniques like Nearest Neighbor (NN), Robust Point Matching (RPM), and Coherent Point Drift (CPD), which degrade rapidly as the perspective shift increases.

Comparison of association rates shows the EGP method's superior robustness to large differences in azimuth angles between radar views.

In practical terms, the algorithm successfully recovered real-world dimensions. In HST simulations, it estimated a solar panel width of 3.65m (actual: 3.8m) and a length of 16.2m (actual: 16.6m). Furthermore, the reconstruction error remained stable at Signal-to-Noise Ratios (SNR) above 10 dB, staying within the radar's range resolution limits.

Conclusion

By treating a satellite's orbit as a physical "ruler," this multistatic ISAR approach eliminates the need for cross-range scaling and provides a robust framework for 3D space target imaging. While the current model assumes steady, three-axis stabilized motion, future work will address the challenges of tumbling targets and self-occlusion where parts of the structure block the radar's line of sight.

Summary updated

BibTeX
@article{zhao2019three,
  author = {Yue Zhao and Lei Zhang and Bo Jiu and Hongwei Liu and Zhenfang Li},
  title = {{Three-dimensional reconstruction for space targets with multistatic inverse synthetic aperture radar systems}},
  journal = {EURASIP J. Adv. Signal Process.},
  year = {2019},
  doi = {10.1186/s13634-019-0630-8},
}
Figure