衍射层析模型下穿墙雷达三维学习成像方法

针对现有穿墙雷达三维稀疏成像中，存在网格时延构建字典矩阵所需内存过大以及凸优化稀疏成像算法阈值参数不确定影响重建图像质量的问题，提出了一种基于衍射层析稀疏模型的学习近似消息传递三维成像方法。该方法在衍射层析成像算法上通过构造快速傅里叶变换算子来建立三维成像稀疏模型，然后修正近似消息传递算法求解稀疏解，并将其迭代过程映射成多层神经网络，最后通过数据驱动自适应学习多层神经网络中的可调参数，从而实现三维学习成像。仿真和实验数据处理结果表明，该方法不仅减小了系统所需内存，还避免了参数的人工调整对成像质量的影响。

Three-Dimensional Learning-Imaging Method of Through-the-Wall Radar with Diffraction Tomography Model

The existing three-dimensional sparse imaging of through-the-wall radar has too much memory required to construct the dictionary matrix with grid delay. Moreover, the uncertain threshold parameters of convex optimization sparse imaging algorithms affect the quality of reconstructed images. A three-dimensional imaging method of learning approximate message passing with diffraction tomography sparse model is proposed in this paper. A three-dimensional imaging sparse model is firstly built by constructing a fast Fourier transform operator for the diffraction tomography algorithm. Subsequently, we modify the approximate message passing algorithm to reconstruct its sparse solution, and map iterative process of sparse solution into a multilayer neural network. Finally, we use data-driven adaptive learning of the adjustable parameters in the multilayer neural network to achieve three-dimensional learning-imaging. Simulation and experimental data processing results show that this method not only reduces the memory required by the system, but also avoids the impact of manual adjustment of parameters on imaging quality.