Multichannel autostereoscopic measurement system for micro-structured surfaces based on multi-scale depth fusion.

Area of Science:

• Optics
• Computer Vision
• Metrology

Background:

• Autostereoscopic technology enables precise metrology of micro-structured surfaces.
• Single light field modality in existing systems limits performance.
• Richer data is needed for improved depth estimation and 3D reconstruction.

Purpose of the Study:

• To develop a multichannel autostereoscopic measurement (MAM) system.
• To enhance the accuracy and robustness of 3D reconstruction for micro-structured surfaces.
• To address the limitations of single-modality autostereoscopic systems.

Main Methods:

• A 3D optical channel captures elemental images (EIs) from multiple viewpoints.
• A 2D channel acquires high-resolution (HR) images.
• Data fusion techniques and a deep learning network (UniDepth) are employed for depth estimation and 3D geometry.
• Pyramid representation combines multi-scale depth information for precise reconstruction.

Main Results:

• The proposed MAM system integrates data from 3D and 2D channels.
• Deep learning (UniDepth) estimates 3D geometry from HR images.
• Data fusion compensates for deficiencies and enhances accuracy.
• Multi-scale depth information improves 3D reconstruction precision.

Conclusions:

• The multichannel autostereoscopic measurement system provides richer data for depth estimation.
• The system demonstrates improved quality and robustness in 3D reconstruction.
• This approach overcomes the constraints of single light field modalities in autostereoscopic metrology.