Single-Shot Structured Light Sensor for 3D Dense and Dynamic Reconstruction.

Area of Science:

• Optics and Photonics
• Computer Vision
• 3D Imaging

Background:

• Structured light (SL) 3D reconstruction faces a trade-off between acquisition speed and spatial resolution.
• Existing methods like temporally coded SL offer high density but not dynamic capture, while spatially coded SL provides single-shot capture but sparse reconstruction.

Purpose of the Study:

• To develop a system capable of simultaneous accurate, dense, and dynamic 3D reconstruction.
• To overcome the limitations of current structured light techniques in terms of speed, density, and dynamic scene capture.

Main Methods:

• A speckle-based structured light sensor utilizing two synchronized cameras and a diffractive optical element (DOE) projector.
• A high-accuracy calibration method and an optimized epipolar rectification algorithm for precise stereo image alignment.
• An improved semi-global matching (SGM) algorithm for enhanced stereo matching accuracy, leading to high-quality depth map generation.

Main Results:

• The system successfully generates dense point clouds from the acquired depth maps.
• Experimental results demonstrate the system's effectiveness, outperforming other single-shot 3D systems.
• Submillimeter accuracy was achieved at a close range of 0.4 meters.

Conclusions:

• The proposed speckle-based SL sensor effectively achieves accurate, dense, and dynamic 3D reconstruction.
• This approach offers a significant advancement over existing structured light methods for real-time 3D imaging applications.
• The system's ability to provide high-density, high-accuracy 3D data in a single shot opens new possibilities for dynamic scene analysis.