Real-time scalable depth sensing with hybrid structured light illumination.

Area of Science:

• Computer Vision
• Optical Sensing
• Robotics

Background:

• Structured light techniques like time multiplexing (TM) and spatial neighborhood (SN) are crucial for depth sensing.
• TM offers high accuracy but is slow, while SN provides low delay but lower accuracy.
• Integrating TM and SN advantages is challenging for scalable depth sensing.

Purpose of the Study:

• To develop a novel, scalable depth sensing paradigm that merges the benefits of TM and SN.
• To create hybrid structured light patterns for adaptable depth reconstruction.
• To design a scene-adaptive framework for optimizing depth maps based on motion.

Main Methods:

• Designed hybrid structured light patterns combining phase-shifted fringe (TM) and pseudo-random speckle (SN).
• Developed a scene-adaptive depth sensing framework utilizing motion detection.
• Implemented a real-time (20 fps) depth sensing system to validate the approach.

Main Results:

• Depth reconstruction achieved using TM principles for static scenes (multiple frames) and SN principles for dynamic scenes (single frame).
• The scene-adaptive framework successfully generated optimal global or region-wise depth maps.
• The developed system demonstrated a real-time performance of 20 frames per second.

Conclusions:

• The proposed hybrid structured light method effectively integrates TM and SN for scalable depth sensing.
• The scene-adaptive framework enables efficient depth map generation balancing accuracy and speed.
• This approach offers a novel solution for real-time depth sensing in diverse applications.