USING SERIAL GRINDING AND IMAGING TECHNIQUES TO PRODUCE THREE-DIMENSIONAL MODELS OF SAMPLES WITH WEAK DENSITY CONTRAST

Detailed three-dimensional measurements, such as diameter, length, aspect ratio, and/or orientation, of fossilized organisms can provide valuable insights into the morphology, life cycle, habit, and evolution of early metazoans. In order to efficiently make such measurements on embedded fossils, researchers produce digital reconstructions using a variety of tomographic techniques. Unfortunately, samples that lack strong density contrast are poor candidates for traditional x-ray scanning methods, and, as a result, often are difficult to reconstruct. In such cases, the use of serial grinding and imaging, which relies solely on color and texture in order to differentiate objects, may be advisable.

We present results from the Grinding, Imaging, Reconstruction Instrument (GIRI) at Princeton University, which consists of an automated serial grinding and imaging pipeline, in conjunction with a newly developed neural network image classification technique. GIRI is a computer numerical controlled surface grinder which has been retrofitted with an imaging stage that consists of an 80 megapixel digital back coupled to a 120mm macro lens. The system is capable of resolving objects as small as 10.4 microns over a 40.4 x 53.7mm field of view. Although the process is destructive, GIRI produces a comprehensive, high resolution digital archive for further analysis. Our image processing technique, which consists of a user-friendly training stage and a predefined convolutional neural network, can quickly and accurately trace objects of interest for use in three-dimensional visualization software.

We share several case studies, including new digital reconstructions of Precambrian tubular fossils, models of early Cambrian metazoans, and analyses of porosity and permeability from an oolite.