SiWa: See into Walls via Deep UWB Radar

Opportunity

In modern architectures, walls are becoming increasingly complex, yet they are susceptible to deterioration and damage over time. Rather than resorting to the drastic measure of breaking down and reconstructing a damaged wall, a more effective maintenance approach involves non-destructive detection of in-wall conditions. Presumably, it would be most helpful to open a “virtual window” into walls to expose the structure inside. Aided by such “visual” cues, one may diagnose the health of the wall structure without unnecessary destructions. Furthermore, early detection of leakage pathways and corrosion in reinforcing bars allows for repairs using less invasive methods.

One promising approach to non-destructive structural inspection is radio frequency (RF-) sensing. In fact, the wireless sensing community has witnessed a rapid growth of sensing capability enabled by RF signals in the past decade. By far, several sensing technologies have been implemented to “see” the location, shape and motion of objects, as well as their material compositions. Some of these systems utilise the diffraction of RF signals to achieve an "X-ray"-like ability to see through obstacles such as drywall. However, they often lack specific capabilities of seeing through walls, especially those built by concrete and bricks, which require processing schemes distinct from those commonly applied to RF signals propagating mainly through air.

Technology

Designing a fully functional system capable of seeing into walls presents significant challenges. First, the absence of low-cost yet high-performance hardware complicates the effort. Second, while new hardware, such as the IR-UWB radar, has emerged in recent years, effectively leveraging signals propagating within walls for diagnostic purposes remains highly complex. The majority of RF-related research has concentrated on signals propagating through air, resulting in a limited number of experimental algorithms available for this task. Moreover, experimental algorithms often struggle when confronted with real-world scenarios. Finally, experimental algorithms often fail when facing realistic situations.

To tackle these challenges, we propose the “see through the wall technology” SiWa. Constructed based on a commodity-grade IR-UWB radar, SiWa is a hand-held device for swiping across a wall surface. It then processes the reflected signals by emulating synthetic aperture radar (SAR) and spectrometer for in-wall structural imaging and material identification, as illustrated in Figure 1.

In particular, SiWa aims to i) detect in-wall structures (e.g., rebar) and defects (e.g., cracks) via imaging (Figure 2), and ii) recognise their status (e.g., whether a reinforcing bar is corroded or a crack contains water seepage) via material identification. To avoid the complicated calibration of model-based approaches, SiWa exploits an end-to-end deep learning pipeline to process signals in a self-adaptive manner.

Figure 1: SiWa collects radio frequency reflections as a user swipes its probe across a wall surface, thereafter it intelligently processes the signals for in-wall structural imaging and material identification (e.g., rebar layout, and whether a rebar is corroded).

Figure 2: A field experiment: three scanning strips (with ground truth cracks marked) and their corresponding radar imaging results.

Applications & Advantages

• We design SiWa as a low-cost, easy-to-carry prototype with a competitive see into wall performance.
• SiWa is capable of obtaining high imaging resolution and classification accuracy for in-wall structures and objects.
• SiWa can also identify the status of certain structure and objects, e.g., whether a reinforcing bar is corroded or a crack contains water seepage.