Research on signal reconstruction of ground-penetrating radar based on matrix pencil method

Beneath the surface of our world lies a hidden landscape of pipes, cables, caves, and structural voids that we cannot see with our eyes. To map this invisible terrain, engineers and geologists rely on ground-penetrating radar, a tool that sends high-frequency radio waves into the earth and listens for the echoes that bounce back from buried objects. It is a non-invasive way to "see" underground without digging, making it essential for finding landmines, locating utility lines, or checking the integrity of tunnels. However, the signals returning from deep underground are often weak and fragile. As these waves travel through soil and rock, they pick up a chaotic layer of static and random noise, much like trying to hear a whisper in a crowded room. This static can drown out the faint reflections from small or deep targets, turning a clear picture of the subsurface into a blurry, unreadable mess. When the signal is too corrupted, the radar cannot distinguish between a harmless patch of dirt and a dangerous cavity, rendering the data useless for safety inspections or construction planning.

To solve this problem, a team of researchers led by Yanan Li and Runlin Luo has developed a new way to clean up these radar signals. Instead of simply filtering out the noise, which often accidentally removes the valuable signal along with it, they treated the radar data as a complex mathematical puzzle. They realized that the useful part of a radar echo behaves like a specific type of mathematical pattern known as a sum of damped exponentials. In plain terms, this means the signal fades away in a predictable way, while the random noise does not follow this pattern. By modeling the signal this way, the researchers could separate the true echo from the static. They used a technique called the matrix pencil method, which involves breaking the data down into its core components to identify the most important "poles" or frequencies that make up the real signal. Once these key components were identified, the team could reconstruct the original, clean signal, effectively stripping away the noise while keeping the subtle details of the underground objects intact.

The researchers tested this approach using three different types of data to ensure it worked in various situations. First, they created a computer simulation of a square cavity buried under layers of clay and limestone. In this virtual environment, they added heavy amounts of white noise to the radar signal, making the reflection from the cavity almost invisible. When they applied their new method, the noise vanished, and the shape of the buried cavity reappeared with striking clarity. They found that the number of mathematical components they chose to keep was critical; too few, and the signal remained noisy, but too many, and the reconstructed wave began to shake and distort. By finding the perfect balance, they restored the signal to a state nearly identical to the original, noise-free version.

Next, the team moved from the computer to a physical laboratory experiment. They built a large sandpit and buried a metal pipe inside it, then scanned it with a real radar system. The raw data from the scan showed the pipe's reflection, but it was jagged and obscured by static. After processing the data with their matrix pencil method, the image became smooth and sharp. The hyperbolic curve that indicates a buried pipe, which is a standard shape in radar imaging, became continuous and easy to read. The researchers measured the improvement and found that the clarity of the image increased significantly, proving that the method could handle real-world physical conditions, not just computer models.

Finally, the team took their method to a full-scale engineering site to test it on a real-world problem: checking the quality of grout behind the segments of a shield tunnel. In tunnel construction, a gap often exists between the concrete rings and the surrounding rock, which is filled with grout to stabilize the structure. If this grout is missing or uneven, it can lead to ground collapse. The researchers scanned a large test platform designed to mimic a real tunnel, which contained several artificial defects, including empty spaces and areas with insufficient grout. In the raw radar images, these defects were hidden behind a wall of noise, with only the strongest signals visible. After applying the reconstruction technique, the hidden defects emerged clearly. Even small voids and subtle anomalies near the tunnel joints became distinct, allowing inspectors to identify potential hazards that would have been missed otherwise.

The results of these tests show that this new method is a powerful tool for improving how we see underground. By focusing on the mathematical structure of the signal itself, the researchers were able to remove high-frequency noise without blurring the edges of the targets. This means that engineers can now detect smaller defects and deeper structures with greater confidence. Whether it is ensuring the safety of a new tunnel, locating a buried pipe, or inspecting the ground for hidden cavities, this technique offers a clearer, more reliable view of the world beneath our feet. The study confirms that by understanding the fundamental nature of the signal, we can recover information that was previously lost to noise, turning a blurry, uncertain image into a precise map of the subsurface.