Three-Dimensional (3D) Nanometrology Based on Scanning Electron Microscope (SEM) Stereophotogrammetry.

Area of Science:

• Materials Science
• Computer Vision
• Metrology

Background:

• Three-dimensional (3D) surface reconstruction from Scanning Electron Microscope (SEM) images is a long-established technique.
• Several commercial stereophotogrammetry software packages are currently available for this purpose.

Purpose of the Study:

• To evaluate the performance of commercially available stereophotogrammetry software packages.
• To assess the accuracy of 3D reconstruction using simulated SEM images of virtual samples with known dimensions.

Main Methods:

• Utilized Monte Carlo simulated SEM images of virtual samples with precisely known geometry.
• Tested two stereophotogrammetry software packages by comparing their reconstructed 3D models against the ground truth of the virtual samples.
• Investigated performance on samples with both rough and uniform (low-contrast) surfaces.

Main Results:

• Both tested software packages performed well on simulated SEM images of samples with rough surfaces.
• Significant height reconstruction errors (approximately 46%) were observed in samples with nearly uniform, low-contrast zones.
• The accuracy of 3D reconstruction is highly dependent on surface topography and contrast.

Conclusions:

• Current stereophotogrammetry software packages require further development for reliable 3D reconstruction from SEM images, particularly those with uniform or low-contrast regions.
• The use of simulated SEM images provides a valuable method for algorithm testing and validation in 3D reconstruction.