New Laser-Based System Captures Images Around Corners in Real-Time

Researchers have used deep learning to create a laser-based system to capture images around corners in real-time.

Various studies have created a unique approach to a non-line-of-sight imaging system. However, a researcher at Stanford University and project leader, Christopher A. Metzler said the new method offers more.

Along with a uniquely high image resolution, the new laser-based system offers more speed. As a result of these attributes, the system can provide applications that would otherwise be impossible.

Co-author of the study, Felix Heide pointed out:

“Non-line-of-sight imaging has important applications in medical imaging, navigation, robotics, and defense. Our work takes a step toward enabling its use in a variety of such applications.”

In a published article in the journal Optica, Metzler, and colleagues from other colleges described how the new system works.

Using Deep Learning to Build a Laser-Based System

To create the new imaging system, the researchers used a commercially available camera, as well as a standard, but powerful laser source.

The laser beam bounces off a visible wall onto the hidden object, then back to the wall. As a result, an interference pattern – speckle pattern – which encodes the shape of the hidden object emerges.

So, the researchers had to reconstruct the hidden object using the speckle pattern.

But that’s not as easy as it sounds. It involves solving a computational challenge.

Real-time imaging requires short exposure times. Unfortunately, the process also produces too much noise for existing algorithms to work.

That’s where the deep-learning algorithm comes in.

Co-author of the study and researcher from Southern Methodist University, Prasanna Rangarajan explained:

“Compared to other approaches for non-line-of-sight imaging, our deep learning algorithm is far more robust to noise and thus can operate with much shorter exposure times.”

Instead of costly experimental data, the researchers used deep learning to train the algorithm to solve the reconstruction problem.

In a test, the imaging system reconstructed images of 1-centimeter-tall letters and numbers hidden behind a corner. With an exposure length of 0.25 seconds, the system produced reconstructions with 300-micron resolution.

Applications of the Imaging System

Potential applications include “reading the license plate of a hidden car as it is driving or reading a badge worn by someone walking on the other side of a corner.”

With further development, the system would enable autonomous vehicles to see around parked cars or busy intersections. Also, it could allow spacecraft and satellites to capture better images – for example, inside a cave on an asteroid.

Now the researchers are working to extend the system’s field of view to enable the reconstruction of more substantial objects.