Fig. 1. Time-multiplexing NLOS imaging framework. (a) Experimental setup of the time-multiplexing NLOS imaging based on splitting the emitted pulse laser. BS 1 divides the emitted pulse laser into two beams. One beam (red lines) directly enters BS 2, while the other beam (yellow lines) undergoes two reflections before entering BS 2. The convex lens and telescope are used to overcome the divergence of the beam. A beam dump is used to collect the remaining beams. A galvanometer directs the two beams with different delays to two separate points and scans a rectangular area on the relay surface. At the receiving end, a lens is used to collect the returned echoes from the sensing point. (b) Photograph of the experimental setup.

Fig. 2. Two separate histograms. (a) Raw photon-arrival-time histogram within one period containing 2 first bounces and 2 third bounces from two separate sampling points, respectively. The period is 25 ns, corresponding to a 40 MHz repetition frequency. (b) The processed and unmixed histograms. The processing includes three steps: 1) subtract the pre-collected background, 2) slightly smooth the histogram using a mean filter, and 3) clip into two parts, given the added delay.

Fig. 3. Experimental imaging results, which show the imaging results of three targets: number “5”, letter “T”, and letter “L,” which are made of carton.

Fig. 4. Simulation of the time-multiplexing NLOS imaging. (a) The schematic of time-multiplexing NLOS imaging. The laser sends four pulses with different delays to the relay surface, directing to four sampling points L1, L2, L3, and L4. The pulses scatter at the relay surface, then diffuse through the hidden scene and scatter off object p. The pulses scatter again at the object, and detector d collects the returned echoes at sensing point s. (b) The simulated histogram, which contains 4 first bounces and 4 third bounces from four sampling points of the relay surface. (c) The reconstructed NLOS imaging results using simulated histograms.

Fig. 5. Comparison of the imaging simulation with and without beam splitting in two different laser powers. (a) The reconstruction results with the 16 × 16 sampling points (16 × 16 scans) at low laser power (without beam splitting). (b) The image with 16 × 32 sampling points (16 × 16 scans) at low laser power (with beam splitting), where the laser power is assumed to be one-fourth of the laser power due to the twice splitting. (c), (d) The results at high laser power. (e)–(h) The comparisons with 32 × 32 and 32 × 64 sampling points on the relay surface. (i)–(l) The comparisons with 64 × 64 and 64 × 128 sampling points on the relay surface.