Efficient three-dimensional near-field imaging with physics-informed deep learning for MIMO radar

Date

2024-6

Author

Oral, Okyanus

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

49
views

12
downloads

Cite This

Near-field radar imaging systems are used in a wide range of applications, such as medical diagnosis, through-wall imaging, concealed weapon detection, and nondestructive evaluation. In this thesis, we consider the inverse problem of reconstructing the three-dimensional (3D) complex-valued reflectivity distribution of the near-field scene from the sparse multiple-input multiple-output (MIMO) array measurements. Motivated by recent advances, we develop physics-informed deep learning techniques for the image reconstruction and array optimization tasks encountered in near-field MIMO radar imaging. Firstly, we develop a novel plug-and-play (PnP) reconstruction method that exploits deep priors and regularization on the magnitude. Our approach provides a unified general framework to effectively handle arbitrary regularization on the magnitude of a complex-valued unknown and is equally applicable to other radar image formation problems including SAR. Secondly, we focus on existing learned direct inversion methods that enable real-time imaging and perform modifications to improve these methods. We demonstrate the effectiveness of all developed approaches under various compressive and noisy observation scenarios using both simulated and experimental data. We also analyze the resolution achieved at compressive settings with sparse MIMO arrays. The developed methods enable not only state-of-the-art performance for 3D real-world targets but also fast computation. Lastly, we develop a novel method for joint optimization of MIMO arrays and reconstruction methods. We illustrate the performance of the jointly optimized imaging system by utilizing various reconstruction methods and different observation settings, and compare the performance with the commonly used MIMO arrays.

Subject Keywords

3D near-field MIMO radar imaging, Complex-valued reconstruction, Plug-and-play methods, Deep learning, Joint optimization

URI

https://hdl.handle.net/11511/110033

Collections

Graduate School of Natural and Applied Sciences, Thesis