ACQUISITION AND DISPLAY OF ULTRA HIGH-RESOLUTION BACKSCATTERED ELECTRON IMAGES OF METEORITE SECTIONS R. C. Ogliore, Washington University in St. Louis ([email protected]) Why Ultra-Res BSE? • Ultra-Res BSE: Backscattered electron imaging at ~50 nm per pixel over an entire 1-inch round thin/thick meteorite section • “Virtual Microscope” allows for remote characterization of sample mineralogy/petrology • Facilitates searching for small rare phases (e.g., cosmic symplectite, carbonaceous microclasts) • Minimizes electron beam exposure for beam-sensitive samples • Especially useful for rare/precious/fragile samples: CI chondrites, returned samples from Hayabusa2 and OSIRIS-REx missions Challenges • Acquisition: Sample topography is larger than the depth-of-field at high magnifications — focus changes across sample • Balance: Brightness/contrast variations due to sample-detector geometry and changing emission current • Robustness: Deal with outliers (bad stage move, no features for alignment) • Stitching: Extreme large final image size (up to 250 gigapixels) makes mosaic assembly difficult due to memory requirements • Display: Large final image makes display challenging Solutions to Challenges • Acquisition: First acquire “focus map” over entire sample and calculate objective-lens working distance (by surface fitting or 2D interpolation) for a matrix of high-resolution acquisitions of ~10,000 individual BSE images. • Balance: Use overlap in stitching and recorded emission current to calculate brightness changes across individual BSE images. Solutions to Challenges (continued) • Robustness: Calculate positions/rotations of each image from matched features in Matlab using an affine geometric transform, then remove outliers and interpolate. Each image → one 3×3 affine matrix. • Stitching: Use pyvips (pixel streaming) and affine matrices calculated from Matlab to stitch final mosaic. • Display: Create deepzoom image tiles in pyvips and display in OpenSeaDragon in web browser, with scale bar and other features for easy cataloging of small features. Conclusions • Tescan Mira3 FEG-SEM with ImageSnapper plugin was used for this work, but can be adapted to other SEMs • All code to acquire, stitch, and display ultra-res BSE images is available on github: https://github.com/ ogliore/DeepZoomSEM • Example images: https://presolar.physics.wustl.edu/ meteorite-deep-zoom/ • Future improvements: 16-bit image display with real- time image adjustments, false-color options, processing speed improvements.
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
ACQUISITION AND DISPLAY OF ULTRA HIGH-RESOLUTION BACKSCATTERED ELECTRON IMAGES OF METEORITE SECTIONS
R. C. Ogliore, Washington University in St. Louis ([email protected])
Why Ultra-Res BSE? • Ultra-Res BSE: Backscattered electron imaging at ~50 nm per pixel over an entire 1-inch round thin/thick meteorite section • “Virtual Microscope” allows for remote characterization of sample mineralogy/petrology • Facilitates searching for small rare phases (e.g., cosmic symplectite, carbonaceous microclasts) • Minimizes electron beam exposure for beam-sensitive samples • Especially useful for rare/precious/fragile samples: CI chondrites, returned samples from Hayabusa2 and OSIRIS-REx missions
Challenges • Acquisition: Sample topography is larger than the depth-of-field at high magnifications — focus changes across sample • Balance: Brightness/contrast variations due to sample-detector geometry and changing emission current • Robustness: Deal with outliers (bad stage move, no features for alignment) • Stitching: Extreme large final image size (up to 250 gigapixels) makes mosaic assembly difficult due to memory requirements • Display: Large final image makes display challenging
Solutions to Challenges • Acquisition: First acquire “focus map” over entire sample and calculate objective-lens working distance (by surface fitting or 2D interpolation) for a matrix of high-resolution acquisitions of ~10,000 individual BSE images.
• Balance: Use overlap in stitching and recorded emission current to calculate brightness changes across individual BSE images.
Solutions to Challenges (continued) • Robustness: Calculate positions/rotations of each image from matched features in Matlab using an affine geometric transform, then remove outliers and interpolate. Each image → one 3×3 affine matrix. • Stitching: Use pyvips (pixel streaming) and affine matrices calculated from Matlab to stitch final mosaic.• Display: Create deepzoom image tiles in pyvips and display in OpenSeaDragon in web browser, with scale bar and other features for easy cataloging of small features.
Conclusions • Tescan Mira3 FEG-SEM with ImageSnapper plugin was used for this work, but can be adapted to other SEMs • All code to acquire, stitch, and display ultra-res BSE images is available on github: https://github.com/ogliore/DeepZoomSEM • Example images: https://presolar.physics.wustl.edu/meteorite-deep-zoom/ • Future improvements: 16-bit image display with real-time image adjustments, false-color options, processing speed improvements.
Related Documents
