Stereo Structured Light Technology

Key Takeaways

  • Active Projection Enhances Matching: Stereo structured light projects known coded patterns onto scenes, significantly improving feature matching stability in low-texture and featureless regions, solving the texture dependency problem of traditional stereo vision.
  • High-Precision Close-Range Sensing: Combining stereo triangulation with structured light encoding, this technology achieves sub-millimeter depth accuracy within 0.3-3 meters, making it ideal for fine manipulation tasks in embodied AI robots.
  • Environmental Light Resistance: Using near-infrared wavelengths and temporal modulation techniques, structured light systems maintain high depth quality in complex lighting environments, offering stronger robustness compared to pure passive stereo vision.

What is it?

Stereo structured light technology is a hybrid depth sensing method that combines active structured light projection with passive stereo vision. The system consists of one or more structured light projectors, stereo cameras (left and right cameras), and supporting image processing algorithms.
The structured light projector projects specific coded patterns (such as stripes, grids, speckles, or Gray codes) onto the scene, forming deformed textures on object surfaces. The stereo cameras synchronously capture images containing the projected patterns and reconstruct depth information by calculating the disparity of corresponding points in the left and right images.
Core Definition: Stereo structured light fundamentally overcomes feature matching difficulties in low-texture regions (such as white walls or smooth surfaces) by actively adding scene textures, finding wide application in 3D scanning, industrial inspection, and embodied AI robotics.
Unlike traditional passive stereo vision, the key advantage of this technology is that it does not rely on natural scene textures but instead creates controllable texture features through active projection, enabling stable and reliable depth reconstruction.

How does it work?

The stereo structured light system workflow consists of four key stages: pattern projection, dual-image capture, feature matching, and depth reconstruction.

Pattern Projection and Encoding

The system projects pre-designed coded patterns through a projector. Common encoding methods include:
  • Sinusoidal Fringe Grating: Calculates phase through phase-shifting methods, suitable for high-precision measurement
  • Binary Gray Code: Achieves unique identification through spatial encoding, suitable for fast measurement
  • Random Speckle: Uses statistical properties of speckle patterns for matching, suitable for real-time applications
  • Blue-Green Line Laser: Performs cross-sectional scanning through multiple parallel laser lines, suitable for profile measurement
Projected pattern wavelengths typically use near-infrared (NIR, such as 850nm or 940nm) to reduce interference with visible light and improve environmental light resistance.

Stereo Synchronous Capture

Left and right cameras capture images containing projected patterns under strictly synchronized timing. Synchronization accuracy is critical for depth calculation, typically requiring microsecond-level precision. Camera exposure time and gain need joint optimization based on ambient light intensity and projection brightness to ensure patterns are clearly visible without overexposure.

Feature Matching and Decoding

The system finds corresponding points in left and right images. Since structured light provides rich texture information, matching algorithms can be based on:
  • Phase Information: For fringe gratings, matching by calculating local phase differences
  • Spatial Encoding: For Gray codes, matching by decoding spatial unique identifiers
  • Feature Points: For speckle patterns, matching through feature descriptors (such as SIFT, SURF)
Compared to traditional stereo vision relying on natural textures, the active textures provided by structured light significantly improve matching success rate and accuracy.

Depth Calculation and Post-Processing

Based on triangulation principles, the depth Z calculation formula is:
Mathematical Foundation: Z = (f · B) / d, where f is camera focal length, B is baseline distance (distance between left and right camera optical centers), and d is disparity (pixel displacement of corresponding points in left and right images).
The system also requires the following post-processing:
  • Distortion Correction: Eliminates lens radial and tangential distortion
  • Denoising Filtering: Applies bilateral filtering or guided filtering to reduce noise
  • Hole Filling: Completes occlusion regions and matching failure areas
  • Point Cloud Generation: Converts depth maps to 3D point cloud data

Why does it matter?

Stereo structured light technology solves the core pain point of traditional passive stereo vision failure in low-texture environments while maintaining the high resolution and cost advantages of stereo systems.
Key Advantage: In embodied AI, robots need to perform fine manipulation tasks in unstructured environments, such as object grasping, threading, and parts assembly. Stereo structured light can provide sub-millimeter depth accuracy at close range, meeting fine manipulation requirements.
Compared to single-camera structured light, stereo structured light better handles occlusion and reflection issues through information redundancy from two viewpoints, improving depth data completeness. Compared to pure ToF technology, stereo structured light offers higher spatial resolution at close range, suitable for texture-rich scene understanding.
Additionally, the hardware cost of this technology is relatively controllable, enabling mass production through standard industrial cameras and customized projection modules, suitable for large-scale commercial deployment.

Applications

Embodied AI Robotics

Stereo structured light is widely used in robot vision perception systems, especially for fine manipulation tasks such as robotic arm grasping, object recognition, 3D object reconstruction, and spatial localization. In robot navigation, this technology can be used for high-precision SLAM (Simultaneous Localization and Mapping) and dynamic obstacle detection.

Industrial Inspection and Metrology

Production lines use stereo structured light for non-contact 3D profile measurement, part dimension inspection, welding quality assessment, and surface defect analysis. Its high precision makes it suitable for automotive manufacturing, aerospace, and precision machining fields.

3D Scanning and Reverse Engineering

Portable 3D scanners use stereo structured light technology to digitize physical objects for digital modeling, applied in cultural heritage preservation, medical devices, body scanning, and digital twins.

Smart Healthcare

Medical imaging devices use this technology for facial scanning, oral scanning, and limb measurement, assisting surgical planning and prosthetic customization. Its non-contact nature avoids discomfort associated with traditional measurement methods.

Human-Computer Interaction

AR/VR devices and gesture recognition systems use stereo structured light for high-precision hand tracking and spatial mapping, enabling more natural interaction experiences.

SGI Solution

Suzhou Guanshi Intelligence Technology Co., Ltd. (SGI) provides end-to-end vision solutions based on stereo structured light technology, covering optical design, hardware integration, algorithm development, and system calibration.
Our Full-Stack Capabilities: From custom optical module design and sensor driver development to mass production calibration, SGI controls the entire value chain to ensure optimal performance.
Core technical expertise includes:
  • Custom Optical Module Design: Designing appropriate projector focal length, field of view (FOV), and encoding patterns based on application requirements, ensuring optimal depth quality within target working distance.
  • Stereo Camera Selection and Integration: Offering various industrial-grade camera module options, including high-resolution CMOS sensors and global shutter solutions to meet dynamic scene requirements.
  • Algorithm Development and Optimization: SGI has developed efficient feature matching and depth filtering algorithms that reduce computational load while maintaining accuracy, supporting real-time processing (30-60 FPS).
  • System Calibration and Correction: Providing complete calibration workflows including camera intrinsic calibration, extrinsic calibration, projector calibration, and stereo rectification, ensuring depth output linearity and consistency.
  • Embedded Deployment Support: Algorithm porting and performance optimization for edge computing platforms (such as NVIDIA Jetson, Rockchip, Raspberry Pi), balancing power consumption, heat dissipation, and real-time performance.
SGI can provide full-process support from prototype design to mass production delivery based on specific customer requirements, ensuring stability and reliability of stereo structured light systems in practical applications.
R50 Structured Light RGB-D Camera For industrial inspection and intelligent interaction scenarios, featuring 0.3–3.5 m high-precision depth sensing and RGB aligned output. GV_D100 Industrial Structured Light Camera For industrial automation and machine vision scenarios, featuring 0.2–10 m wide-range depth sensing and industrial-grade anti-interference capability. Robot Vision Application Scenarios View robot vision scenarios from customer requirements and deployment perspectives.

Related Topics