Super-resolution technique for dense 3D reconstruction in fringe projection profilometry

Area of Science:

• Computer Vision
• Metrology
• 3D Sensing

Background:

• Fringe projection profilometry (FPP) is a key 3D reconstruction technique.
• Higher fringe pattern resolution is crucial for detailed and accurate 3D point clouds in 3D sensing.
• Current methods for FPP super-resolution require enhanced hardware.

Purpose of the Study:

• To propose a novel deep learning approach for FPP super-resolution.
• To reconstruct high-definition 3D shapes without requiring additional hardware.
• To improve the accuracy and detail of 3D reconstructions.

Main Methods:

• Development of a dual-dense block super-resolution network (DdBSRN).
• Integration of a novel dual-dense block structure within a multi-path network architecture.
• Training the network with smaller data samples to achieve a fully functional model.

Main Results:

• The DdBSRN method demonstrates stable and robust performance.
• The proposed method significantly outperforms standard interpolation techniques.
• Achieved superior accuracy and 3D detail in reconstructed shapes.

Conclusions:

• The DdBSRN offers an effective software-based solution for FPP super-resolution.
• This approach enables high-definition 3D shape reconstruction with enhanced accuracy and detail.
• The method is efficient, requiring less training data and no hardware upgrades.