High-resolution Structure from Motion modelling and 3D printing of Scanning Electron Microscopy data

Structure from Motion photogrammetry calculates 3D point clouds through identification of matching features in an overlapping series of pictures, from which textured 3D surfaces can be derived. This method has become increasingly popular in field geology because with the help of drone pictures, high-resolution digital outcrop models, digital elevation models or orthoimages can be produced at very high quality but low-costs.

Here, we use secondary electron images with micron-scale resolution to reconstruct the 3D geometry of a c. 400 μm quartz mineral fish using photogrammetry. The quartz fish from a marble ultramylonite from Thassos (Greece) has been extracted by an in situ etching technique (Bestmann et al., 2000, JSG, 22, 1789-1807). 57 secondary electron images captured at various stage rotations and stage tilts in a TESCAN Vega II scanning electron microscope were automatically aligned using the Structure from Motion software Agisoft Metashape (version 1.8.4). In order to increase the precision of the algorithm the background information of the images was removed using Adobe Photoshop and 15 marker points were identified in the images, which also helped to define a scaled coordinate system. We calculated a dense point cloud (c. 2.8 million points) from which a 3D model (c. 600000 faces) was derived on which the secondary electron image information was textured.

The tiled 3D model can be used to precisely measure parameters like volume, surface or shapes of the quartz fish either in Agisoft Metashape or from the exported 3D model using more specialized 3D analysis software (e.g. CloudCompare). Furthermore, features at the nanometer-scale like size and orientations of the grain boundaries or crystal faces of the dissolved calcite crystals, which surrounded the quartz fish, can be quantitatively investigated. After cleaning and down-sampling of the exported polygon mesh, the 3D surface can be transformed into a volume and eventually 3D printed. This method offers a great potential for quantitative investigations of the geometry and spatial relationship of microstructures and printed 3D models are a great haptic tool, which can be used in teaching and public outreach.