天基ISAR空间目标重构算法

Space target reconstruction algorithm for spaceborne ISAR

关键词:  空间态势感知  天基ISAR  目标重构  旋转模式判决  矩阵正则化

Key words:  space situation awareness  spaceborne ISAR  target reconstruction  rotation pattern classification  matrix regularization

摘要:

面向在轨目标的天基逆合成孔径雷达（ISAR）成像是空间态势感知（SSA）的一项关键支撑技术。传统的二维距离-多普勒（RD）图像虽可部分揭示目标的散射特性,但其本质缺乏预测轨道机动和实现非合作目标识别所必需的三维几何信息。而现有基于图像序列的多视角重构方法在天基应用场景下存在固有局限:天基平台与目标卫星间的相对轨道运动会造成有效观测时间受限与成像投影平面不稳定的问题。为应对上述挑战,提出一种可变观测模式的目标重构算法。首先,该算法直接利用各特显点距离徙动（RM）轨迹中蕴含的目标结构信息,规避了易引入误差的图像配准过程。其次,推导了二维与三维旋转模式下距离徙动轨迹的表征模型,并提取了用于旋转模式分类的判别特征。最后,提出一种基于高阶多普勒系数估计的高精度距离徙动估计技术；对于不同的旋转模式,结合截断核范数正则化的因式分解方法可实现受观测误差影响时的二维或三维目标重构。仿真结果表明,所提空间目标重构算法可有效实现散射点提取以及距离徙动矩阵重建,进一步实现距离徙动矩阵的正则化收敛,从而获得不同旋转状态下空间目标重构结果,验证了所提算法的有效性与灵活性。

Abstract:

The advancement of spaceborne inverse synthetic aperture radar (ISAR) imaging for on-orbit target represents a critical technology for space situational awareness (SSA). While conventional two-dimensional (2D) range-Doppler (RD) imaging provides valuable scattering intensity distributions, it inherently is the projection of the target’s three-dimensional (3D) structure, thereby losing critical geometric information essential for predicting orbital maneuvers and enabling non-cooperative target recognition. Current multi-views reconstruction methods based on image sequences face inherent limitations in spaceborne scenarios: the relative orbital motion between the spaceborne platform and target satellite induces limited observation time and unstable imaging projection plane. To address these challenges, a target reconstruction algorithm with variable observation mode is proposed. First, the structural information contained in range migration (RM) trajectories is directly exploited to avoid the error-prone image alignment process. Second, we derive a unified geometric model characterizing the range migration (RM) evolution under both 2D and 3D rotation patterns, and extract discriminative features for rotation pattern classification. Finally, we develop a high-precision RM estimation algorithm based on higher-order Doppler coefficient estimation. For different rotation modes, a truncated nuclear norm (TNN) regularization combined with factorization framework enables the reconstruction of 2D or 3D targets under observation error conditions. Simulation results demonstrate that the proposed target reconstruction algorithm effectively achieves scattering point extraction and RM matrix reconstruction. It further ensures the regularized convergence of the RM matrix, thereby obtaining spatial target reconstruction results under various rotation states. This validates the effectiveness and flexibility of the proposed algorithm.