Accurate 3D Reconstruction of Dynamic Objects by Spatial-Temporal Multiplexing and Motion-Induced Error Elimination

Area of Science:

• Computer Vision
• Robotics
• Metrology

Background:

• Accurate 3D reconstruction of dynamic objects is crucial for applications like object recognition and robotic manipulation.
• Simultaneously achieving high accuracy and robustness to motion in 3D reconstruction remains a significant challenge.

Purpose of the Study:

• To present a novel method for high-accuracy and motion-robust 3D reconstruction of dynamic objects.
• To address and eliminate reconstruction errors induced by object motion between image frames.

Main Methods:

• A structured-light multiplexing method using only 3 spatial-temporally encoded patterns for efficient image acquisition.
• Extraction of temporal and spatial codewords to ensure high accuracy and resolve phase ambiguity in stereo matching.
• Derivation of projection pixel motion among frames using extracted spatial codewords to correct motion-induced errors (MiE).

Main Results:

• The proposed method achieves high reconstruction accuracy and precision for dynamic objects across various motion speeds.
• Experimental validation demonstrates superior performance compared to existing methods for different types of motion, including translation and deformation.
• Successful recovery of object surfaces from sequences of images, unaffected by object motion during acquisition.

Conclusions:

• The novel method effectively overcomes the limitations of traditional 3D reconstruction techniques for dynamic scenes.
• It provides a robust solution for accurate 3D reconstruction of objects undergoing complex movements.
• The technique holds significant potential for advancing fields reliant on precise dynamic 3D data.