J. Eur. Opt. Society-Rapid Publ. 22, 40( 2026) 391
Figure 1. a) Top-view schematic of the NLOS setup. b) Front-view of the relay wall. The black square represents the projected target position.
detection points, can be designed in a more photon-efficient manner.
Both confocal and non-confocal arrangements share fundamental hardware components comparable to those of direct time-of-flight( ToF) LiDAR systems used for threedimensional imaging in the direct field of view( FoV). However, in NLOS systems the laser pulse paths are more complex than those encountered when measuring distances in direct line of sight. First, at least three reflections before detection drastically reduce the receivable photon flux compared with single reflection in direct ToF LiDAR. Second, the achievable lateral resolution of reconstructed NLOS scenes cannot be controlled solely via detector pixel resolution or the number of illumination points, as in LiDAR. Instead, not the number of pixels but the temporal resolution combined with the spatial extend of the illumination and detection scheme on the relay wall define the achievable depth and lateral resolution. These requirements place heavy demands on the measurement setup. Existing lab-grade systems are often expensive and far from compact, relying on pico- and femtosecond laser sources, single-photon-sensitive detectors, and complex time-correlated singe-photon counting electronics. Moreover, free-space propagation loss scales with the inverse fourth power of the object distance to the relay wall [ 7 ], compared with the inverse-square loss in LiDAR. This is often compensated by higher laser power, which comprises eye-safety and limits applicability to laboratory environments.
In this work, an eye-safe NLOS localization system is presented that addresses the constraint of restricted laser emission through a high-efficiency photon-detection design and achieves inexpensive, compact implementation using nanosecond-pulsed laser diodes. Specifically, a two-dimensional single-photon-avalanche-diode( SPAD) array with integrated timing capabilities and camera optics is developed and used to observe, in parallel, the detection region on the relay wall. A single illumination point combined with the dense grid of detection points meets the requirements for a NLOS measurement. Scan-free NLOS imaging [ 13, 14 ] has been reported previously with this single illumination system design using picosecond pulsed laser sources. Extending this prior work, we examine the performance gains for localization using additional illumination position. Simulated time-of-arrival photon statistics, validated by measurements with the implemented system, are used for the investigation. Despite the longer pulse width compared to comparable NLOS systems, three-dimensional localization of a target in a NLOS scenario is achieved at a resolution determined solely by detector parameters.
2 Material and methods
2.1 Measurement setup
The measurement setup comprises two pulsed laser diodes and a time-resolved multi-pixel detector [ 15 ] with camera optics that image the relay wall. Each diode operates at the accessible emission limit specified for eye-safe operation, meaning that neither the repetition rate nor the pulse energy can be increased without reducing the other in order to increase the system’ s performance without losing civilian applicability. Nevertheless, the spatial separation and individual emission angles allow the use of more than one diode in the system, as only one laser can be imaged onto a hypothetical retina at a time. Therefore, spatially separated sources are used to comply with laser safety standards, even though the total emission into the scene is multiplied. The dual illumination scheme corresponds to the simplest case of multiple illumination and provides two symmetry planes for the following analysis. Although more sources could be implemented, illuminating arbitrary positions using a mirror system with a single laser source for instance, circumvents the exploitation of multiplied total emission. Depending on the reflecting angles of the mirror, a nearby retina could receive the laser power from multiple emission angles. Moreover, the laser diodes used are cost-efficient and solid-state-based compared to galvo mirror designs.
The arrangement is a common NLOS imaging scenario with a flat relay wall, an occluder, an observing system, and an NLOS scene containing a single target placed in 1.5 m distance to the relay wall( Fig. 1a).
The SPAD-array detector [ 15 ] has 24 32 pixels with individual time-to-digital converters( TDCs) that generate ToF histograms under a first-photon detection scheme. With each laser pulse, each pixel yields at most one timestamp. This time of arrival corresponds to the first induced avalanche in the SPAD within one measurement cycle of 1 / f rep = 80ls duration, where f rep denotes the laser pulse repetition frequency of 12.5 kHz. Because signal photons, ambient light, and dark counts can all trigger these avalanches, histogramming of the time stamps is required to