GSA Connects 2021 in Portland, Oregon

Paper No. 213-4
Presentation Time: 8:50 AM


MANZUK, Ryan1, SINGH, Devdigvijay2, MEHRA, Akshay3 and MALOOF, Adam C.2, (1)Princeton UniversityGeosciences, Guyot Hally, Princeton, NJ 08544-0001, (2)Department of Geosciences, Princeton University, Princeton, NJ 08544, (3)Dartmouth CollegeDepartment of Earth Sciences, 19 Fayerweather Hill Road, Hanover, NH 03755

Recent advances in the fields of computer vision and machine learning have enabled accurate classification and segmentation of petrological and paleontological samples (among other geological materials). To integrate these computational methods and build off of insights made through microscope-based techniques, we first need method for taking standardized, high-resolution (e.g., micron-scale) images of polished slabs and thin sections with a field of view large enough to fit samples containing fossils and bedforms. An imager for geological materials also can leverage reflectance minima of mineral spectra in the NIR range, as well as UV activation of apatitic, calcitic, and organic fossil components as additional sources of contrast that can aid quantitative classification but are overlooked by the human eye and standard red-green-blue cameras. We describe a new multispectral macro-photography setup that can acquire 1:1 magnification images at 3.8~μm per pixel spatial resolution over a 21 cm2 field of view, equipped with 8-band (470-940 nm) spectral resolution, and ultraviolet (365 nm) activation for reflected-light photography of polished slabs. The macro setup is integrated with the Grinding, Imaging, and Reconstruction Instrument (GIRI), so that thousands of multispectral images can be acquired automatically as GIRI grinds through a specimen to produce 3D image stacks. Additionally, we have designed a 5-band (470-940 nm) multispectral light table with automated rotating polarizers. This table allows use of the camera as a high-throughput means for multispectral imaging of thin sections with the potential for quantitative studies that leverage mineral extinction under cross-polarized light. Because the transmitted and reflected light setups must contend with chromatic aberration, we demonstrate a technique to increase image sharpness through blur quantification and post-processing. Our setup and methods provide a fast and accurate way to (1) build reproducible and accessible image archives of rock units, (2) classify and segment those images, and (3) quantitatively compare facies and fossil assemblages.