Imaging on the Edge: Mapping Object Corners and Edges with Stereo X-Ray Tomography.

Area of Science:

• Medical Imaging
• Computer Vision
• Materials Science

Background:

• X-ray computed tomography (XCT) provides 3D volumetric imaging but requires numerous projections, limiting temporal resolution.
• Dynamic processes and high-temporal-resolution imaging are hindered by the time-consuming nature of traditional XCT data acquisition.
• Previous stereo X-ray methods enabled rapid 3D reconstruction of fiducial markers using only two projections.

Purpose of the Study:

• To demonstrate stereo X-ray techniques for 3D reconstruction of sharp object corners, removing the need for internal fiducial markers.
• To enable deformation measurement of manufactured components under load using enhanced temporal resolution.
• To explore the use of synthetic data for model training when real annotated data is scarce.

Main Methods:

• Developed stereo X-ray imaging to reconstruct 3D object corners from only two projection images.
• Applied techniques to manufactured components for deformation analysis under load.
• Investigated and validated model training using synthetic data mimicking real-world stereo X-ray characteristics.

Main Results:

• Successfully achieved reliable 3D reconstruction of sharp corners using just two X-ray projections.
• Demonstrated the method's effectiveness with real-world stereo X-ray images.
• Confirmed applicability without requiring annotated real training datasets.

Conclusions:

• Stereo X-ray 3D reconstruction is feasible for object corners using synthetic training data.
• The approach expands the applicability of stereo X-ray methods in resource-limited scenarios.
• Enables faster, marker-free 3D analysis for dynamic processes and component deformation.