Quantifying the Influence of Surface Texture and Shape on Structure from Motion 3D Reconstructions.

Area of Science:

• Photogrammetry and 3D Reconstruction
• Optical Metrology
• Computer Vision

Background:

• Optical geometrical measurements depend on object surface properties.
• Structure from Motion (SfM) requires surface texture for feature detection and accurate 3D point-cloud triangulation.
• Quantitative studies on surface texture's influence on SfM accuracy are lacking.

Purpose of the Study:

• To investigate the impact of varying object texture levels on SfM reconstruction accuracy.
• To introduce quantitative metrics for evaluating SfM reconstruction quality.
• To compare the influence of texture level with variations in capturing range.

Main Methods:

• Utilized reference artifacts with well-defined geometries.
• Employed Structure from Motion (SfM) using ContextCapture software and a 50 Mpx DSLR camera.
• Introduced metrics for lateral resolution, vertical geometric variation, and spatial-frequency information.
• Compared SfM results with calibrated optical microscopes.

Main Results:

• Developed a pipeline for quantitatively assessing texture's influence on SfM.
• Demonstrated that both capturing range and texture level affect reconstructed geometries.
• Found that increased range reduces spatial resolution, while insufficient texture increases noise and potential errors.

Conclusions:

• The proposed methodology enables quantitative investigation of texture effects in SfM.
• Texture level and capturing range are critical factors influencing SfM accuracy.
• The study provides a replicable framework for evaluating SfM performance under varying surface conditions.