A System for In-Line 3D Inspection without Hidden Surfaces.

Area of Science:

• Computer Vision
• 3D Reconstruction
• Metrology

Background:

• Traditional 3D scanners often suffer from occlusions due to mechanical handling or object support.
• Stereo-vision methods require distinct surface features (singularities) for accurate reconstruction.
• Industrial inspection demands high-throughput, occlusion-free 3D scanning, especially for varied parts.

Purpose of the Study:

• To present a novel 3D scanning technique that overcomes occlusion limitations.
• To develop a system suitable for automated industrial inspection of diverse objects.
• To enable complete 3D object reconstruction without mechanical manipulation.

Main Methods:

• Utilizes a variable array of industrial cameras capturing simultaneous views.
• Objects are scanned during freefall, eliminating support-based occlusions.
• Employs silhouette segmentation and carving algorithms for volumetric representation.
• Applies isosurface generation techniques to create 3D models.

Main Results:

• Achieved occlusion-free 3D reconstruction of complete objects.
• Demonstrated the system's capability without mechanical handling (robot arms, spinning plates).
• Validated the method's effectiveness on objects lacking visual singularities.
• Presented experimental results comparing ideal and reconstructed object geometries.

Conclusions:

• The proposed freefall 3D scanning method offers a robust, occlusion-free solution for object reconstruction.
• This technique is highly suitable for integration into production lines requiring high-volume, automated inspection.
• Further research is needed to address reconstruction limitations in specific geometric configurations and bound reconstruction error.