Non-Line-of-Sight Imaging

Abstract

Emerging single-photon-sensitive sensors combined with advanced inverse methods to process picosecond-accurate time-stamped photon counts have given rise to unprecedented imaging capabilities. Rather than imaging photons that travel along direct paths from a source to an object and back to the detector, non-line-of-sight (NLOS) imaging approaches analyze photons scattered from multiple surfaces that travel along indirect light paths to estimate 3D images of scenes outside the direct line of sight of a camera, hidden by a wall or other obstacles. Here we review recent advances in the field of NLOS imaging, discussing how to see around corners and future prospects for the field.

Three-dimensional imaging through scattering media based on confocal diffuse tomography

“…Finally, another class of methods reconstructs objects by explicitly modeling and inverting scattering of light. For example, non-line-of-sight imaging techniques invert scattering off of a surface or through a thin layer [20][21][22][23][24][25][26] , but do not account for diffusive scattering. Diffuse optical tomography (DOT) reconstructs objects within thick scattering media by modeling the diffusion of light from illumination sources to detectors placed around the scattering volume 27,28 .…”

Non–line-of-sight imaging over 1.43 km

“…In the experiment, we show the ability to retrieve hidden objects at different hidden distances to the visible wall. The agreement between the joint probability density function and the object's true position shows that our system is able to locate a stationary person up to ∼1.8 m. We perform 10 measurements for each position of the hidden object and observe that the SD is ∼ [10, 10, 2.5] cm and the error (or accuracy) between the mean value and the actual position is ∼ [7,8,3] cm, at the hidden distance of 1 m. These results are comparable to previous experiments (12,13). Furthermore, we demonstrate real-time tracking ( Fig.…”

“…non-line-of-sight imaging | optical imaging | computational imaging | computer vision I n recent years, the problem of imaging scenes that are hidden from the camera's direct line of sight (LOS), referred to as non-line-of-sight (NLOS) imaging (1)(2)(3), has attracted growing interest (4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14). NLOS imaging has the potential to be transformative in important and diverse applications such as medicine, robotics, manufacturing, and scientific imaging.…”

“…In contrast to conventional LOS imaging ( 15 – 17 ), the diffuse nature of light reflected from typical surfaces in NLOS imaging leads to mixing of spatial information in the collected light, seemingly precluding useful scene reconstruction. To address these issues, optical techniques for NLOS imaging that have been demonstrated include transient imaging ( 4 – 8 , 10 , 11 , 18 ), speckle correlations ( 9 , 19 ), acoustic echoes ( 20 , 21 ), intensity imaging ( 14 ), confocal imaging ( 22 ), occlusion-based imaging ( 23 – 27 ), wave-propagation transformation ( 28 , 29 ), Fermat paths ( 30 ), and so forth ( 1 – 3 ). With few exceptions ( 14 , 23 – 27 ), most of the techniques that reconstruct NLOS scenes rely on high-resolution time-resolved detectors and use the information encoded in the time-of-flight (TOF) of photons that scatter multiple times.…”

Imaging and certifying high-dimensional entanglement with a single-photon avalanche diode camera

“…The maturity of the technology has enabled the production of compact arrays 4,5 , as well as the reduction of the cost per device through bulk manufacturing processes 6 . Thus far, imaging devices based on SPADs have demonstrated their capabilities in fluorescence lifetime imaging [6][7][8] , LiDAR (light detection and ranging) [9][10][11][12] , non-line-of-sight imaging 13,14 , imaging through strongly scattering media 15 and time-resolved correlation measurements 16,17 . However, SPAD cameras have yet to make their mark due to their overall efficiency and resolution; the fill factor of the earliest available SPAD cameras was on the order of a few percent 6,18 , which, despite the quantum efficiency of the single SPAD pixels being on par with commercial single element SPADs, equates to a large overall loss.…”