Single-Frequency Microwave Imaging with Dynamic Metasurface Apertures
Timothy Sleasman, Michael Boyarsky, Mohammadreza F. Imani, Thomas Fromenteze, Jonah N. Gollub, David R. Smith
(Submitted on 27 Mar 2017)
Conventional microwave imaging schemes, enabled by the ubiquity of coherent sources and detectors, have traditionally relied on frequency bandwidth to retrieve range information, while using mechanical or electronic beamsteering to obtain cross-range information. This approach has resulted in complex and expensive hardware when extended to large-scale systems requiring ultrawide bandwidth. Relying on bandwidth can create difficulties in calibration, alignment, and imaging of dispersive objects. We present an alternative approach using an electrically-large, dynamically reconfigurable, metasurface antenna that generates spatially-distinct radiation patterns as a function of tuning state. The metasurface antenna comprises a waveguide feeding an array of metamaterial radiators, each of whose properties can be modified by applying a voltage to a diode integrated into the element. By deploying two of these apertures, one as the transmitter and one as the receiver, we realize sufficient spatial diversity to alleviate the dependence on frequency bandwidth and obtain both range and cross-range information using measurements at a single frequency. We demonstrate the method experimentally, using one-dimensional dynamic metasurface apertures and reconstructing various two-dimensional scenes (range and cross-range). Furthermore, we modify a conventional microwave imaging technique—the range migration algorithm—to be compatible with such configurations, resulting in an imaging system that is fast and simple in both software and hardware. The imaging scheme presented in this paper has broad application to radio frequency imaging, including security screening, through-wall imaging, biomedical diagnostics, and synthetic aperture radar.
Microwave imaging systems provide unique sensing capabil- ities that are desirable for a variety of applications. Their fea- tures include the construction of three-dimensional (3D) im- ages, the ability to penetrate optically-opaque materials, and the use of non-ionizing electromagnetic radiation—collective traits that are desirable in applications ranging from security screening to biomedical diagnostics [1–11]. Most conventional microwave systems take the form of mechanically-scanned an- tennas or use electronic beamforming in order to retrieve scene content from backscatter measurements [1, 12–14]. Excellent imaging performance can be obtained from these systems, but these approaches often suffer from implementation drawbacks. Specifically, mechanically-scanned antennas tend to be slow and bulky, while electronic beamforming systems, such as phased
arrays or electronically scanned antennas (ESAs) are complex, expensive, and often exhibit significant power draw.
Since the radiation pattern from an aperture does not vary with range in the far-field, targets distant from a static aperture must be probed with a band of frequencies to resolve depth infor- mation in addition to cross-range information. Such a scenario is common, for example, in synthetic aperture radar imaging, which makes use of a transceiver scanned over a large area . Imaging paradigms that require wide bandwidths complicate the radio frequency (RF) signal generation/detection hardware and necessitate allocation of precious portions of the electro- magnetic spectrum. Large frequency bandwidth also increases the possibility for interference with other electronic devices, fur- ther complicating the hardware and processing techniques. In addition, wideband systems tend to suffer from reduced data acquisition rates (due to extended dwell times needed to settle the phase locking circuitry).
From an imaging perspective, objects exhibiting frequency
dispersion (such as walls in through-wall imaging) are problem- atic for wideband systems [6, 17]. From an implementation per- spective, large bandwidths can also contribute to the complexity of system integration because they increase sensitivity to mis- alignment and necessitate calibration of the aperture/feed layers . Given these numerous challenges, reliance on frequency bandwidth has become the bottleneck of many microwave imag- ing systems. If the requirement of bandwidth can be removed, the imaging platform may be considerably simplified.
This notion is in line with many emerging applications which require a capacity for real-time imaging within strict economic constraints. Efforts have recently been made to simplify the hardware required for high resolution imaging and instead rely more heavily on post-processing algorithms. In this approach, termed computational imaging, spatially-distinct waveforms can be used to interrogate a scene’s spatial content, with compu- tational techniques used to reconstruct the image. This idea, which transfers much of the burden of imaging from hardware to software, has been demonstrated at all ranges of the electro- magnetic spectrum as well as in acoustics. Applying this idea to microwave imaging, the conventional dense antenna arrays (or mechanically-swept systems) can be replaced with several large antennas that offer radiation pattern tailoring ca- pabilities, e.g. metasurface apertures. In some instances, it has been shown that metasurface apertures can radiate frequency- indexed wavefronts that multiplex a scene’s spatial information through backscatter measurements. In these demon- strations, post-processing decodes the measurements to resolve an image in all three dimensions. Since these systems rely on large antennas and collect information across a sizable surface in parallel, the architecture is favorable compared to systems that are densely populated with independent antenna modules. Nonetheless, such systems remain particularly dependent on a wide bandwidth due to their use of frequency diversity to generate a sequence of spatially diverse radiation patterns.
Alternative efforts have focused on reducing the dependence of metasurface apertures on a wide bandwidth by leveraging electronic tuning [34–36]. These dynamic metasurfaces are com- posed of individually addressable metamaterial elements whose resonance frequencies can be varied as function of external volt-
(black arrows) are simultaneously radiating and the signals are age. By selecting the resonance frequency of each element— which influences the amplitude and possibly phase of radiation emitted from that point—spatially-distinct radiation patterns can be generated which are capable of multiplexing a scene’s cross-range spatial content without requiring a large frequency bandwidth. Using dynamic metasurfaces, high-fidelity image reconstruction has been experimentally demonstrated using a small bandwidth, even down to a single frequency point [34, 35, 37]. However, in these demonstrations range resolu- tion is limited due to the small bandwidth of operation. In the extreme case of a single frequency point, imaging was viable but range resolution completely disappeared, yielding images that only identified objects in the cross-range dimension.
The seemingly indispensable bandwidth requirement, con- sidered necessary to resolve objects in the range direction, is not a fundamental limitation arising from the physics of microwave imaging systems. This can be understood by interpreting im- age formation through spatial frequency components (i.e. in k-space). In this framework, sampling each spatial frequency component (k-component) can yield information in the corre- sponding spatial direction. Analysis in k-space has been success- fully employed in synthetic aperture radar (SAR) for decades to ensure images free of aliasing. More recently, it has been used to realize phaseless imaging and to implement massive, multi- static imaging systems for security screening [3, 10, 11, 33, 38].
More generally, wavefronts propagating in all directions (sampling all spatial frequency components through the decom- posed plane waves) can retrieve information along all direc- tions, even when a single spectral frequency is used. The key to achieving this capability is to operate in the Fresnel zone of an electrically-large aperture, where electromagnetic fields exhibit variation along both the range and cross-range directions, allow- ing the spatial frequencies to be sampled with no bandwidth.
To properly sample the required k-components at a single fre- quency, a set of relatively uncorrelated radiation patterns must be generated spanning a wide range of incidence angles. This type of wavefront shaping has been explored from a k-space per- spective in the optical regime, with the most famous examples being the early works on holography [39, 40]. In these works, a monochromatic source was sculpted into complex, volumetric shapes in the near-field of a recorded hologram. At microwave frequencies, however, generating waveforms to sample all k- components in an imaging application has not been straightfor- a) ward. In a recent attempt, a simple antenna was mechanically scanned along a long path and range/cross-range imaging was demonstrated using a single spectral frequency. This approach The k-space illumination at an off-centered location in a scene for the cases of a) SIMO and b) MIMO systems. Note that the k-space has significantly enhanced support in both kx and ky for the MIMO case.
2. IMAGING BACKGROUND A. Single-Frequency Imaging
In microwave imaging, range resolution can be determined by the time domain resolution of the measured signal, or Since 2π this algorithm takes advantage of fast Fourier transform tech- δx/y = ∆k niques, it can reconstruct scenes at real-time rates. Given that x/y have not been directed at obtaining range information and do not possess favorable form factors Considering the results of k-space analysis and the ideas behind computational imaging, we investigate dynamic meta- surface apertures as a simple and unique means for realizing single-frequency Fresnel-zone imaging. The proposed concept is illustrated in Fig. 1, where an electrically-large, linear, transmit- receive aperture is formed by a pair of dynamic metasurfaces; one acting as transmitter and the other as receiver. In Fig. 1b the rays from two arbitrary points on the transmit (Tx) and receive (Rx) antennas are drawn and the corresponding spatial frequen- cies in the scene are shown. Since we are assuming Fresnel zone operation, k-components in 2D (both range and one cross-range dimension) can be probed from a 1D aperture radiating a series of distinct patterns at a single frequency.
dynamic metasurface apertures are planar and easily-fabricated, the added benefit of single frequency operation further simplifies the RF hardware requirements. The imaging system proposed in Fig. 1a therefore promises a fast, low-cost device for appli- cations such as security screening, through-wall imaging, and biomedical diagnostics.
We begin by reviewing the underlying mechanisms of mi- crowave and computational imaging. Next, we outline the spe- cific aperture that has been implemented and detail its operation in a single-frequency imaging scheme. Experimental images are shown to highlight the performance and a point spread function analysis is carried out across the field of view [13, 46]. We also demonstrate that single-frequency imaging is robust to practical misalignments—a property especially attractive in the emerg- ing application of imaging from unmanned aerial/terrestrial vehicles . The reduction of bandwidth to a single frequency provides a wealth of benefits from a hardware/processing per- spective, yet still the proposed system performs quality imaging that is desirable for a variety of applications.