基于张量压缩感知的雷达目标三维成像技术

Recently, radar imaging technique by means of compressive sensing has become a popular research topic at home and abroad, which has a great potential on reducing the burden of data collection, storage and transmission and improving the imaging quality. Currently, compressive sensing are mainly dedicated to one-dimensional or two-dimensional radar imaging, and when it is applied to three-dimensional radar imaging, most of the available methods usually convert the data cube into a long vector along with huge dictionary and measurement matrices, which will result in heavy computation burden of solving large-scale optimization problems and huge memory consumption for storing dictionary and measurement matrices. In this project, we extend the compressive sensing to the tensor case, break the bottlenecks encountered in high-dimensional signal processing using compressive sensing from the perspective of tensor, and investigate the application of tensor compressive sensing to the three-dimensional radar target imaging . The research contents include: (1) tensor modeling of three-dimensional radar echo; (2) construction and optimization of measurement matrix with low-memory usage, and design of radar data compressive sampling schemes; (3) development of fast and stable sparse tensor reconstruction algorithms, and experimental validation of the proposed method. Research of tensor compressive sensing and its application in three-dimensional radar imaging in this project is a prospective task and full of challenges, which has great theoretical significance and application value for the design and realization of new imaging radar system and high-dimensional signal processing.

压缩感知雷达成像技术是当前国内外研究的热点和难点，在降低数据采集、存储和传输负担以及提高成像质量等方面具有很强的应用潜力。目前该领域研究主要针对一维和二维雷达成像，然而，在对目标进行三维成像时，传统压缩感知成像方法需要将数据进行向量化操作，会带来大规模优化问题求解以及大型字典和测量矩阵存储等问题。本项目结合用于高维信号处理的张量理论和压缩感知理论，从张量的视角突破传统压缩感知方法在高维信号处理中遇到的瓶颈，探索基于张量压缩感知的雷达目标三维成像方法，研究内容包括：（1）雷达目标回波稀疏张量建模；（2）低存储测量矩阵构造优化与雷达目标三维回波压缩采样方法；（3）快速稳健的雷达目标三维成像张量重构算法设计及实验验证。本项目所涉及的张量压缩感知理论在三维雷达成像中的应用研究是一个具有前瞻性和挑战性的课题，对新型雷达成像系统设计与实现以及高维信号处理具有重要的理论意义和实用价值。

压缩感知雷达成像在降低数据采集、存储和传输负担以及提高雷达目标成像质量方面具有很强的应用潜力。但是，在三维成像应用中，常规压缩感知方法通常将信号重排为向量再进行处理，面临信息损失和大规模计算问题。本项目在深入研究雷达目标回波数据的低秩特性和雷达目标稀疏特性的基础上，借助张量这一处理高维问题的有力工具，提出一系列高效的雷达目标稀疏三维成像方法，主要包括：（1）基于散射中心理论，构造三维稀疏字典，建立雷达目标回波三维张量稀疏模型，更好揭示雷达目标回波、目标三维散射系数以及与频率、方位和俯仰方向稀疏字典直接的关系；（2）利用回波的低秩特性和目标图像的稀疏特性，根据不同的数据稀疏采样方案，提出基于张量表示的三维成像方法，高效实现目标三维成像；（3）提出一种基于联合低秩和稀疏先验信息的雷达目标二维成像方法，能够提高不同稀疏采样方案下的目标成像质量；（4）基于信源的空域稀疏分布特性，面向L型阵列和均匀线性阵列，提出基于稀疏张量表示模型的到达角估计方法，能够大幅度降低稀疏重构的计算复杂度。本项目在IEEE Trans. Image Processing 等权威期刊、会议发表学术论文8篇。相关研究成果有助于解决当前雷达目标三维成像数据量大、计算复杂度高等问题，促进稀疏表示理论和雷达成像技术的发展，为新型三维雷达设计与实现提供理论和技术支撑，并进一步将相关技术向水下目标探测、高光谱成像和核磁共振成像等应用领域拓展。