【摘要】 针对复杂场景中目标雷达散射特性数据缺失，获取成本高、理论计算效率低的问题，开展了结合物理光学法与多任务高斯过程回归的目标电磁散射特性快速预示研究。首先，介绍了单输入单输出高斯过程回归模型，并推导了物理光学法中频率与远区散射场之间的关系。然后，结合高斯过程回归中的先验均值函数和后验均值函数，构建频域散射特性代理模型，实现目标散射特性的快速预示。最后，提出了应用于目标散射特性预示的单输入多输出的高斯过程回归模型，通过构建多任务的目标散射特性数据集，能够同时实现多个任务的预示。仿真结果表明，相较于传统的单任务高斯过程回归模型，该方法在保证预测精度的同时，能够大幅提升预示效率，预测时间约为单任务高斯过程的67.5%。更多还原

【Abstract】 To address the challenges of data scarcity, high acquisition costs, and low theoretical computational efficiency in radar cross-section(RCS) characterization of targets under complex scenarios, this study develops a rapid prediction framework for electromagnetic scattering properties through physics opticsinspired and multi-task Gaussian process regression. Firstly, a single-input single-output(SISO) Gaussian process regression model is established, with theoretical derivation of the frequency-to-far-zone scattered field relationship based on physical optics. Secondly, by integrating prior and posterior mean functions from Gaussian process regression, a frequency-domain scattering surrogate model is constructed to enable rapid characterization. Finally, a multi-input multi-output(MIMO) Gaussian process regression architecture is proposed, which achieves simultaneous multi-task prediction through the establishment of a unified scattering characteristics dataset. Simulating results demonstrate that comparing with the traditional single-task Gaussian process regression model, this proposed method can significantly improve the prognostic efficiency while guaranteeing the prediction accuracy, and the prediction time is about 67.5% of the single-task Gaussian process.更多还原