Researchers at Stevens Institute of Technology (Hoboken, NJ) have created a 3D imaging system that uses quantum properties of light called parametric modes to create images about 40,000 times crisper than current technologies, paving the way for greatly improved lidar sensing and detection in self-driving cars, satellite mapping systems, deep-space communications, and medical imaging of the human retina.1
The work, led by Yu-Ping Huang, director of the Center for Quantum Science and Engineering at Stevens, addresses a decades old problem with lidar. While photodetectors used in these systems are sensitive enough to create detailed images from just a few photons, it has been tough to differentiate reflected laser light from brighter background light such as sunlight.
The technology is the first real-world demonstration of single-photon noise reduction using a method called quantum parametric mode sorting (QPMS), which was first proposed by Huang and his team in a 2017 Nature paper. Unlike most noise-filtering tools, which rely on software-based postprocessing to clean up noisy images, QPMS checks light's quantum signatures through nonlinear optics to create an exponentially cleaner image at the level of the sensor itself.
Huang and colleagues describe a method for imprinting specific quantum properties onto an outgoing pulse of laser light, and then filtering incoming light so that only photons with matching quantum properties are registered by the sensor. The result: an imaging system that is very sensitive to photons returning from its target, but that ignores virtually all unwanted noisy photons. The team's approach yields sharp 3D images even when every signal-carrying photon is drowned out by 34 times as many noisy photons.
"By cleaning up initial photon detection, we're pushing the limits of accurate 3D imaging in a noisy environment," says Patrick Rehain, a Stevens doctoral candidate and the study's lead author. The hardware-based approach could facilitate the use of lidar in noisy settings where computationally intensive postprocessing isn't possible. The technology could also be combined with software-based noise reduction to yield even better results. "We aren't trying to compete with computational approaches—we're giving them new platforms to work in," Rehain says.
1. Patrick Rehain et al., Nature Communications (2020); https://doi.org/10.1038/s41467-020-14591-8.