The application of Object Based Image Analysis to Petrographic Micrographs R. Marschallinger and P.Hofmann GIScience Research Institute, Austrian Academy of Sciences, Schillerstr. 30, A-5020 Salzburg, Austria In this paper, we describe the application of Object Based Image Analysis for the knowledge-based, automated mineral classification from petrographic micrographs. Digital color images, acquired by an optical, petrographic microscope at different stage rotations and with parallel and crossed polarization filters, are the input to a sequence of different image segmentation and fuzzy classification steps. As compared with traditional, pixel-based algorithms, Object Based Image Analysis incorporates not only the spectral characteristics of various mineral phases, but also their topological and genetic properties, resulting in superior image classification results. By means of rule sets, image analysis can be flexibly adapted to different rock types. Keywords Petrographical Microscopy; Petrographical Image Analysis; Object Based Image Analysis 1. Introduction Petrographic thin section microscopy, based on transmitted as well as on reflected light, has a long tradition both in applied and academic geosciences. A broad spectrum of microscopic techniques for the identification of rock-forming minerals and ore minerals has been established over the past century [1]. As by education and experience, a geoscientist can straightforwardly identify a rock’s constituent mineral phases, quantify fabric parameters and infer a rock’s genesis, involving just rock thin sections and a petrographic microscope [2,3]. Despite these obvious advantages, there are inherent drawbacks: petrographic microscopy is a time-consuming, iterative approach that involves expert knowledge and lots of experience in combining multiple, mostly “soft” optical classification criteria (compare Table 1); optical- based quantification of mineral chemistry is unreliable unless impossible and an optical microscope’s magnification range is quite limited. With a palette of microscope add-ons like polarization filters, slots, specialized lens systems, apertures and the application of associated microscopy “tricks”, petrographic microscopy is a typical expert domain that has a tendency to yield irreproducible results. Although textual descriptions of micro-petrographical findings remain important, the need for more quantitative and reproducible data from optical microscopy has long been recognized [4]. Digital image analysis and image classification applied to petrographic micrographs, in part necessitating sophisticated hardware [5], have been important steps towards quantification. Today, petrographic image analysis systems work with pixel-based image analysis algorithms [6]; they perform reliably in selected fields, e.g. for the automatic extraction of reservoir rock properties [7,8]. However, routinely applied to magmatic or metamorphic rocks, pixel-based algorithms commonly fail: on the one hand, there are overlapping RGB characteristics of the constituent minerals, on the other hand, it is impractical to abstract the before-mentioned, multi-criteria expert handling by means of traditional, pixel- based image analysis methods, because these address mainly the spectral characteristics of the minerals in a petrographic thin section. 2. Material For demonstrating our Object Based Image Analysis approach to the automatic classification of petrographic micrographs, we choose a Metatonalite from the Pennine basement of the Tauern Window (Eastern Alps, Austria). The sample has been selected because of its rich set of textural features that are directly linked to the rock’s history: the Metatonalite has undergone a two-stage evolution covering the late Variscan, intrusive emplacement and the younger Alpidic metamorphism [9,10]. The rock fabric mirrors this history (Fig. 1): on the one hand, the primary magmatic texture with idiomorphic Plagioclases, Biotite clots, Quartz aggregates and minor occurrence of Potassium Feldpars has been preserved. On the other hand, in the course of the Alpidic regional metamorphism, the rock has been re- equilibrated to greenschist facies pressure-temperature conditions: originally Ca-rich, magmatic Plagioclases have been transformed to Albite/Oligoclase, with concurrent growth of small Epidote/Clinozoisite and White Mica minerals inside the Plagioclases. The distribution of Epidote minerals portrays an original, magmatic chemical zoning of the Plagioclases. Biotite has been transformed to metamorphic greenish-brown variants, unmixing the titanium component as Titanite. Almandine-rich Garnet is attributed to the Alpidic event, too. Accessories are Zircon, Apatite and Ore minerals. Microscopy: Science, Technology, Applications and Education A. Méndez-Vilas and J. Díaz (Eds.) 1526 ©FORMATEX 2010 ______________________________________________