| 摘要: |
| 在毫米波雷达干涉测量中阵列天线数较少，会导致雷达成像的分辨率低，无法满足对多目标进行微形变监测的需求。本文提出了基于MIMO毫米波雷达三维合成孔径成像（SAR）的q-SPICE-MAP超分辨率成像算法。该算法将单快拍下的广义稀疏迭代协方差估计（q-SPICE）算法扩展到多快拍维度，基于MIMO-SAR成像平台进行方位向-高度向超分辨率成像；针对q-SPICE算法作为协方差估计算法，其成像结果缺少目标的相位信息，通过最大后验概率（MAP）准则对超分辨率成像结果进行相位反演。通过多组仿真实验表明，在目标角度间隔小于阵列孔径分辨率情形下，相较于IAA算法、多测量向量稀疏贝叶斯学习（MMV-SBL）算法和单快拍q-SPICE-MAP算法，本文所提算法能够对目标进行清晰成像，微形变测量的均方根误差（RMSE）和平均绝对误差（MAE）显著降低，验证了其具有良好的超分辨率成像和微形变监测能力。 |
| 关键词: MIMO毫米波雷达 微形变监测 超分辨率成像 相位反演 协方差估计 |
| DOI： |
| 分类号:TN957.51 |
| 基金项目:国家自然科学基金资助项目(61561010);广西创新驱动发展专项资助项目 (桂科AA21077008);广西无线宽带通信与信号处理重点实验室2022年主任基金资助项目(GXKL06220102,GXKL06220108);八桂学者专项经费资助项目(2019A51);桂林电子科技大学研究生教育创新计划资助项目(2022YXW07,2022YCXS080);2022年广西高等教育本科教学改革工程项目(2022JGB196);桂林电子科技大学学位与研究生教改项目(2022YXW07,2023YXW02);广西研究生教育创新计划资助项目(YCSW2022271),广西研究生教育创新计划项目(编号：2024YCXS032). |
|
| q-SPICE-MAP: Micro-Deformation Monitoring Based on MIMO-SAR Super-Resolution Imaging |
|
|
|
|
| Abstract: |
| In millimeter-wave radar interferometry, a limited number of array antennas often lead to low imaging resolution, which fails to meet the requirements for micro-deformation monitoring of multiple targets. The q-SPICE-MAP super-resolution imaging algorithm was proposed based on three-dimensional Multiple-Input Multiple-Output (MIMO) Synthetic Aperture Radar (SAR). The single-snapshot generalized sparse iterative covariance-based estimation (q-SPICE) algorithm was extended to the multi-snapshot dimension, and azimuth-elevation super-resolution imaging was performed based on the MIMO-SAR platform. To address the inherent lack of phase information in q-SPICE, which is a covariance estimation-based approach, the Maximum A Posteriori (MAP) criterion was integrated for phase retrieval so that the super-resolution results could be reconstructed. Through multiple sets of simulation experiments, it was demonstrated that when the angular interval between targets was smaller than the array aperture resolution, clearer imaging was achieved by the proposed algorithm compared to the IAA algorithms, the Multiple Measurement Vector Sparse Bayesian Learning (MMV-SBL) algorithm, and the single-snapshot q-SPICE-MAP algorithm. Furthermore, the Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) of micro-deformation measurements were significantly reduced, validating its superior capability in super-resolution imaging and micro-deformation monitoring. |
| Key words: MIMO millimeter-wave radar micro-deformation monitoring super-resolution imaging phase retrieval covariance estimation |