3D reconstruction of objects with occlusion and surface reflection using a dual monocular structured light system.

Area of Science:

• Computer Vision
• 3D Reconstruction
• Industrial Automation

Background:

• 3D vision is crucial for industrial applications.
• Occlusion and reflection pose significant challenges in complex scene reconstruction.

Purpose of the Study:

• To present a novel 3D reconstruction framework addressing occlusion and reflection issues.
• To improve the accuracy and efficiency of 3D reconstruction in complex industrial environments.

Main Methods:

• Utilized a dual monocular structured light system for multi-angle point cloud acquisition.
• Developed a decision map for efficient point cloud fusion, avoiding redundant data.
• Implemented a compensation method to reduce fusion area errors.
• Employed gray-code and phase-shifting patterns with a unique compensation function to avoid phase-jumping.

Main Results:

• The proposed framework successfully reconstructs complex scenes, including those with shiny surfaces and occlusions.
• Experimental evaluations demonstrate high accuracy and robustness compared to traditional fusion algorithms.
• The decision map and compensation method significantly enhance point cloud fusion efficiency and reduce reconstruction errors.

Conclusions:

• The novel dual monocular structured light system framework effectively solves occlusion and reflection problems in 3D reconstruction.
• The method offers a robust and accurate solution for complex industrial vision applications.
• This approach advances the capabilities of 3D vision systems in challenging environments.