Fast 3D reconstruction via event-based structured light with spatio-temporal coding.

Area of Science:

• Computer Vision
• Robotics
• 3D Sensing

Background:

• Event-based cameras offer low latency and high dynamics for high-speed applications.
• Existing event-based structured light methods often neglect spatio-temporal consistency for depth estimation.

Purpose of the Study:

• To develop an event-based structured light system with a novel spatial-temporal coding strategy.
• To introduce a Spatio-Temporal Enhanced Matching (STEM) approach for improved 3D reconstruction.

Main Methods:

• A system combining a laser point projector and an event camera was developed.
• A spatial-temporal coding strategy was devised for dual-domain depth encoding in a single shot.
• The STEM approach, comprising STE and STM algorithms, was presented to exploit spatio-temporal synergy.

Main Results:

• The proposed method achieves a high reconstruction rate of 16 frames per second.
• A low root mean square error of 0.56 mm was obtained at a distance of 0.72 m.
• STEM effectively integrates dual-domain information for robust matching and accurate depth computation.

Conclusions:

• The developed event-based structured light system and STEM approach significantly advance high-speed 3D reconstruction.
• The method demonstrates superior performance in accuracy and speed compared to existing techniques.
• Exploiting spatio-temporal consistency in event-based data is crucial for robust depth estimation.