Research map / 3D vision / Depth and stereo
Opportunistic Single-Photon Time of Flight
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Shows that scattered light from unknown pulsed lasers carries a sinc-comb spectral signature that a free-running SPAD can detect, synchronize to, separate, and use for passive mm-scale 3D imaging.
Problem
Passive 3D imaging using the fast, faint ambient light from asynchronous, unknown, out-of-view pulsed lasers, without hardware synchronization or control of the sources.
Approach
Pulsed lasers leave a sinc-comb pattern in the temporal frequency spectrum of incident flux that is laser-specific and scene-invariant; the method detects combs by searching repetition frequencies, localizes them via harmonic hopping to mHz precision, reconstructs per-pixel pulse trains for computational synchronization, and fits laser positions and depth geometrically.
Key results
- Passive 3D reconstruction of 3D-printed staircase targets achieves 2.3 mm average RMSE versus 4.9 mm for Azure Kinect, and 2.5 mm at SBR 0.01.
- Laser repetition frequency estimated to within about 1 mHz at high SBR and 10 mHz at low SBR.
- Wall-plane RMSE of 3.4 mm and 3.6 mm in a room-scale scene with three out-of-view lasers at about 10 MHz.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use opportunistic single-photon ToF instead of UWB flux probing | Single-photon pulse reconstruction · Pulse reconstruction, SBR 0.5 | pulse FWHM: 120 ps vs 110 ps | Helped |
| Use opportunistic single-photon ToF instead of UWB flux probing | Single-photon pulse reconstruction · Pulse reconstruction, SBR 0.01 | pulse reconstruction | Helped |
| Use opportunistic single-photon ToF instead of Azure Kinect | 3D staircase reconstruction · 3D staircase targets | depth RMSE: 4.9 mm vs 2.3 mm | Helped |
| Reduce SBR to 0.01 | Opportunistic single-photon ToF · 3D staircase targets | depth RMSE: 2.3 mm; 2.5 mm | No clear effect |
| Move the camera and laser sources independently between acquisitions | Opportunistic single-photon ToF · Dynamic acquisition | registered point-cloud RMSE: 1.1 cm | Mixed |
| Aggregate timestamps from neighboring pixels | Laser detection under low-SBR conditions · SBR 0.01, 10×10 pixel neighborhood | laser detection | Helped |
Surprising findings
- Overhead room lights (SBR about 0.01) prevent detection with single-pixel probing, but 10x10 pixel timestamp aggregation enables laser detection.
Limitations
- Uses a single-pixel SPAD with galvo scanning to emulate a 2D SPAD camera rather than a real 2D camera.
- Assumes stationary camera and lasers during exposure and direct-light dominance (negligible indirect light in main model).
- Geometric optimization is non-convex and sensitive to occlusions and shadows.
Experimental setup
Metrics: RMSE (mm) frequency estimation error (mHz) inlier ratio pulse FWHM (ps)
Baselines: UWB flux probing (Wei et al.) Azure Kinect
Builds on: ultra-wideband imaging (Wei et al. 2023) CFAR frequency detection Barron adaptive robust loss
Closest papers by meaning
- Towards 3D Vision with Low-Cost Single-Photon Cameras (CVPR 2024 · similarity 0.85)
- Snapshot Lidar: Fourier Embedding of Amplitude and Phase for Single-Image Depth Reconstruction (CVPR 2024 · similarity 0.84)
- PlatoNeRF: 3D Reconstruction in Plato's Cave via Single-View Two-Bounce Lidar (CVPR 2024 · similarity 0.84)
- Separating the Wheat from the Chaff: Spatio-Temporal Transformer with View-interweaved Attention for Photon-Efficient Depth Sensing (AAAI 2025 · similarity 0.83)
- Time-Efficient Light-Field Acquisition Using Coded Aperture and Events (CVPR 2024 · similarity 0.82)
- Neural Inverse Rendering from Propagating Light (CVPR 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.