Single Shot Panoramic Depth Acquisition

33 concepts25 papers1 proposal

To navigate an autonomous drone through a crowded room, map a narrow pipeline, or track moving objects across an urban intersection, a system needs to see in all directions at once. Traditional 3D sensing architectures struggle with this requirement because they trade time for space.

A standard approach to capturing a 360-degree environment relies on sequential scanning: spinning a LiDAR deck (Thomas Lowe et al. [Paper]), rotating a camera on a motorized gimbal (Kennard Chan et al. [Paper]), or stitching together a sequence of narrow-field images. If the robotic platform moves or objects in the scene shift during the acquisition cycle, the time delay introduces severe geometric misalignments, ghosting artifacts, and high computational latency.

To overcome these limitations, a distinct paradigm has emerged: single-shot panoramic depth acquisition. By co-designing specialized optics with structured illumination or time-resolved sensors, these systems capture complete, spatially synchronized 360-degree depth profiles in a single sensor exposure.

Bending Light to See Everywhere at Once

The foundation of single-shot panoramic sensing lies in optical configurations that compress a spherical or cylindrical field of view onto a single flat image sensor. This is primarily achieved through two methods:

  1. Catadioptric Systems: These setups combine a standard camera lens with a precisely shaped curved mirror (such as a cone, parabola, or hyperbola) mounted directly in front of the gaze. For example, Ryota Hiura et al. demonstrate how a conic mirror projects the surrounding 360-degree horizon so that all vertical lines in the environment map perfectly to radial lines originating from the center of the image sensor [Paper]. Similarly, Tomoki Hayashi et al. utilize a hyperbolic mirror to capture a complete tabletop workspace and its participants simultaneously using a single downward-facing camera [Paper].
  2. Panoramic Annular Lenses (PAL): Instead of mirrors, PAL systems use specialized refractive and reflective surfaces within a single lens assembly to capture a wide horizontal band. As shown by Lei Sun et al., this hardware approach yields a flat cylindrical perspective of the entire surrounding horizon in a single shot, eliminating the weight and latency of mechanical gimbals on aerial platforms [Paper].

While these passive optical designs provide complete situational awareness, extracting accurate 3D geometry from a single warped 2D image is mathematically underdetermined, especially in dark or featureless environments.

Active Omnidirectional Structured Light

To resolve depth instantaneously without relying on ambient lighting or surface textures, researchers integrate omnidirectional optics with active structured light projection.

Instead of sweeping a single laser line across a room over time, these systems project a continuous 360-degree plane of light. When this light plane intersects the surrounding environment—such as the interior walls of a pipe or tunnel—it forms a visible ring. An omnidirectional camera directed at the same scene captures the deformations of this ring in real time.

      [ Hyperbolic Mirror ]
             /    \
            /      \
  [Wall] <--  Ring  --> [Wall]
            \ Laser /
             \     /
         [ Camera Lens ]

The intuition behind why this works is rooted in rigid triangulation geometry. If the sensor is inside a perfectly uniform cylinder, the camera views the projected laser ring as a perfect circle. If the sensor encounters a dent, an expansion, or an obstacle, the ring warps. Because the spatial relationship between the laser projector, the curved mirror, and the camera sensor is fixed and calibrated, the exact 3D coordinate of every point along that ring can be calculated from its pixel position in a single frame.

Atsushi Yamashita et al. pioneered this approach for pipe inspection, combining a 360-degree laser projector with a hyperboloidal catadioptric camera to map internal pipe geometry without any rotating parts [Paper, Paper]. For larger, confined infrastructures like transport tunnels, Takuya Igaue et al. demonstrate that pairing a ring laser with wide-angle fisheye lenses allows a compact rig to capture high-density cross-sectional profiles continuously as the platform moves forward [Paper]. This concept also scales to open environments; Jin Shin et al. utilize multiple discrete line lasers directed at a hyperbolic mirror to provide simultaneous 360-degree obstacle ranging for mobile robots [Paper].

Alternative Modalities: Stereo and Time-of-Flight

The core philosophy of single-shot panoramic acquisition—decoupling spatial coverage from temporal scanning—extends to other depth-sensing modalities, such as passive stereo and active Time-of-Flight (ToF).

Single-Frame Panoramic Stereo

Traditional stereo vision requires multiple cameras separated by a baseline distance. To achieve a panoramic stereo view without a massive camera array, an elegant solution captures light rays tangential to a defined viewing circle, mimicking the interpupillary distance of human eyes across all azimuth angles. In Panoramic Stereo Videos with a Single Camera, the authors introduce a multi-faceted "Coffee Filter Mirror" that splits a single camera sensor's perspective into a synchronized, dual-view panorama, capturing instantaneous 360-degree parallax. On a smaller scale, Trung Ngo Thanh et al. employ a multi-mirror catadioptric sensor that provides a central wide view alongside peripheral mirrors to deliver multi-baseline stereo information within a single image frame [Paper].

Panoramic Time-of-Fight

Instead of calculating depth via triangulation geometry, ToF sensors measure the phase shift or travel time of emitted light. By pairing integrated fisheye lenses directly with a ToF sensor array, Hideaki Kanayama et al. demonstrate that a system can capture a complete indoor RGB-D map in a single capture instance, eliminating the ghosting and stitching errors that plague traditional architectural scanners [Paper]. Taking this concept to an extreme, the authors of Trapping Light for Time of Flight use a mirror-filled "light trap" combined with time-resolved sensing to computationally unroll reflected light paths, reconstructing a full 360-degree 3D model of an object from a single static sensor position.

Hybrid and Active Trade-offs

While single-shot panoramic sensing provides instantaneous, artifact-free spatial context, it introduces a fundamental physical trade-off: it spreads the finite resolution of an image sensor across a massive field of view, reducing angular resolution.

To balance this trade-off, researchers often deploy hybrid architectures. Sung-Heum Kim et al. propose a "scout-inspector" design, where a static catadioptric camera acts as a persistent 360-degree spatial buffer to detect environmental changes, which then instantly guides a high-resolution pan/tilt camera to zoom in on specific targets [Paper].

Alternatively, Yan Meng introduces an active vision approach called ViDAR, which mounts a visual-inertial sensor on a high-speed rotating motor controlled by a reinforcement learning agent [Paper]. Rather than capturing the entire scene in a single passive shot, the system dynamically and intelligently directs its gaze to track features in all directions, expanding its effective field of view while maintaining high local resolution.

Key Foundational Work

Evaluation of Effectivity of Omnidirectional Image Sensor COPIS in a Real Environment (1994) — contributed 1 concept elements to this direction.
Acquiring a Radiance Distribution to Superimpose Virtual Objects onto a Real Scene (1999) — contributed 1 concept elements to this direction.
History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 24 papers
  2. Version 1 · 2026-10-04 · 17 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_omnidirectional_structured_light_sensing_c173e660,
  title = {omnidirectional structured light sensing},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/omnidirectional-structured-light-sensing/rdr_c173e660c0a24c86852313b89913b432},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-22},
}

This exact version (version 2, frozen text):

@misc{lacuna_omnidirectional_structured_light_sensing_c173e660_v2,
  title = {omnidirectional structured light sensing},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/omnidirectional-structured-light-sensing/rdr_c173e660c0a24c86852313b89913b432/v/2},
  note = {Version 2, frozen 2026-10-04. Accessed 2026-11-22},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake