Compressive Imaging – Queen’s University Belfast
Development of innovative imaging technologies has been at the forefront of the research conducted within the electromagnetics community for decades. Today, imaging systems form an indispensable part of our daily lives, from optical cameras to LiDAR and millimetre-wave (mmW) radars with applications ranging from automotive sensors, non-destructive testing, to remote sensing and security screening to name a few. Across the electromagnetic spectrum, radiation at microwave and mmW frequencies is particularly advantageous as electromagnetic waves at these frequencies are not ionizing (unlike X-rays), can see through materials that are opaque at optical frequencies and operate in all-weather conditions.
MmW imaging radars conventionally rely on a multi-pixel based raster scanning principle to image a limited field-of-view located in front of the radar. This raster scanning requirement can be realized using several techniques, such as mechanical raster scanning shown in Fig. 1(a) and phased array based all-electronic scanning solutions depicted in Fig. 1(b). Limitations with such imaging technologies are that raster scanning can be slow (a significant challenge for real-time data acquisition) and require an excessive amount of data (a significant challenge for real-time image reconstruction).
To address these challenges, there has been an increasing shift towards unusual solutions, leveraging compression in the physical layer. Computational imaging facilitated by compressive sensing is one such technique that has recently received significant traction [1-3]. Computational imaging uses single-pixel compressive antennas to radiate wave-chaotic modes to sample the scene information using quasi-random bases as depicted in Fig. 1(c).
This quasi-random sampling replaces the raster scanning requirement. In computational imaging, the back-scattered data from the imaged scene is compressed into a single channel and an estimate of the scene information can then be retrieved by correlating the compressed back-scattered measurements and the transfer function of the wave-chaotic compressive antennas. As a result, the mmW images of the scene can be retrieved from a significantly reduced number of measurements in comparison to the raster scanning based techniques. Computational compressive imaging is an all-electronic technique, and hence does not require any mechanically moving parts. Moreover, the synthesis of quasi-random modes to facilitate compressive imaging can be achieved in real-time using a simple frequency-sweep or dynamically modulating the radiation pattern of the wave-chaotic antennas.
The development of compressive sensing theory and its recent applications in mmW imaging systems has shown to offer a significant potential to reshape the next generation imaging technologies. It is a question of “how” rather than “if” we will soon witness the transition of this technology from lab-based research efforts to real-life applications.
Compressive Imaging – Queen’s University Belfast
© GSA 2021
