Structured Light Scanning

Unlike passive methods like photogrammetry, structured light scanning is an active sensing technique. It projects a controlled, known light pattern (e.g., stripes, grids, or pseudo-random dots) onto the target object. This active projection provides its own illumination, making the system less dependent on ambient lighting and enabling the capture of textureless or uniformly colored surfaces that are challenging for passive methods.

The core mechanism is geometric triangulation. The system consists of a known baseline between a projector and a camera. The projector casts the pattern, and the camera observes how the pattern deforms on the object's surface. By knowing the original projected pattern and the baseline geometry, the system can calculate the 3D coordinates of each point on the surface through triangulation, similar to stereo vision but with one 'eye' being a pattern projector.

• Key Equation: Depth (Z) is inversely proportional to the observed disparity (shift) in the pattern.

To achieve high accuracy and resolve correspondence ambiguities, advanced temporal coding methods are used. The most common is phase-shifting profilometry (PSP), where a series of sinusoidal fringe patterns are projected with incremental phase shifts. By analyzing the phase of the deformed fringes at each pixel, a precise, unwrapped phase map is computed, which directly correlates to depth. This method provides dense, sub-pixel accuracy and is robust to surface reflectivity variations.

Structured light scanners are renowned for high metrological accuracy and spatial resolution. They can achieve micron-level precision for close-range setups, making them ideal for industrial inspection, reverse engineering, and quality control. The resolution is determined by the camera sensor density and the fineness of the projected pattern, allowing for the capture of fine surface details, scratches, and intricate geometries.

The technique has specific material limitations. Highly specular (shiny) surfaces cause mirror-like reflections that scatter the pattern, leading to missing data. Transparent or translucent materials (e.g., glass, clear plastic) allow light to pass through or scatter internally, corrupting the surface measurement. These challenges are often mitigated by applying a temporary matte spray coating to the object before scanning.

• vs. Photogrammetry/SfM: Active vs. passive. More accurate for textureless objects but requires controlled setup.
• vs. LiDAR: Uses visible light patterns, not lasers. Typically higher resolution for close-range but shorter effective range.
• vs. Time-of-Flight (ToF): Measures depth via geometric triangulation, not the time delay of a light pulse. Generally more accurate but slower due to multiple pattern projections.
• vs. Stereo Vision: Replaces one camera with a pattern projector, eliminating the challenging passive correspondence problem.

Photogrammetry is the science of obtaining reliable measurements and 3D reconstructions from photographs. Unlike active scanning, it is a passive technique that relies on ambient light.

• Key Principle: Uses triangulation from multiple overlapping 2D images to compute 3D coordinates.
• Comparison to Structured Light: Photogrammetry is ideal for large-scale outdoor scenes and historical artifacts where projecting a pattern is impractical. Structured light excels at capturing fine, textureless surface details in controlled environments.
• Common Pipeline: Involves Structure from Motion (SfM) followed by Multi-View Stereo (MVS).

RGB-D Reconstruction creates a 3D model by fusing synchronized color (RGB) and depth (D) images. Structured light scanners (e.g., Intel RealSense) are a primary source of RGB-D data.

• Sensor Fusion: Combines the detailed color texture from the RGB camera with the precise geometric data from the depth sensor.
• Volumetric Integration: Depth frames are integrated into a unified volumetric representation, commonly a Truncated Signed Distance Function (TSDF), to progressively build a watertight model.
• Applications: Real-time 3D mapping for robotics, SLAM, and interactive AR/VR experiences.

Light Field Reconstruction aims to capture or synthesize the full plenoptic function—the intensity of light rays traveling in every direction through every point in space.

• Contrast with Geometry-Based Scanning: While structured light captures explicit surface geometry, light field methods capture the appearance from many viewpoints, enabling effects like post-capture refocusing and parallax.
• Data Format: Represents a 4D function (2D spatial + 2D angular dimensions).
• Hybrid Approaches: Advanced systems may combine structured light for geometry with multi-camera arrays to sample the light field for photorealistic view synthesis.

Inverse Rendering is the process of inferring underlying scene properties (geometry, material, lighting) from 2D images. It is the inverse of the traditional computer graphics rendering pipeline.

• Relation to Scanning: Structured light scanning directly measures geometry. Inverse rendering often infers it, along with other properties, from appearance.
• Differentiable Rendering: Modern approaches use differentiable rendering to optimize neural scene representations via gradient descent, explaining subtle shading and reflectance cues.
• Output: Produces disentangled, editable scene attributes (e.g., a PBR material map) rather than just a colored mesh.

A Truncated Signed Distance Function (TSDF) is a volumetric representation crucial for fusing multiple depth scans, including those from structured light sensors.

• Mechanism: Each voxel in a 3D grid stores the signed distance to the nearest surface (positive outside, negative inside), truncated to a fixed band. A weight is also stored for running averages.
• Fusion: New depth frames are integrated by updating the TSDF volume, smoothly averaging measurements and reducing noise.
• Surface Extraction: The final zero-crossing of the TSDF (where distance=0) defines the surface, extracted using algorithms like Marching Cubes to create a mesh.

Stereo Matching is a passive depth estimation technique that finds corresponding pixels in a pair of rectified images from a stereo camera to compute a disparity map.

• Fundamental Difference: Stereo matching uses passive triangulation with natural texture, while structured light uses active triangulation with a projected pattern, providing robustness in textureless regions.
• Disparity to Depth: Disparity (pixel shift) is converted to depth using known camera baseline and focal length.
• Hybrid Systems: Some advanced structured light systems use a stereo camera pair to observe the deformed pattern, improving robustness and resolution.