Avtor/Urednik | Riss, Jooelle; Gentier, Sylvie; Genter, Albert | |
Naslov | Granitic core cross sections: numerical modelling | |
Tip | članek | |
Vir | Image Anal Stereol | |
Vol. in št. | Letnik 20, št. Suppl 1 | |
Leto izdaje | 2001 | |
Obseg | str. 550-5 | |
Jezik | eng | |
Abstrakt | Starting from RGB images of granitic core cross sections, this paper is an attempt to derive the most representative numerical models as possible for the core sections: each pixel is given a label (false colour) related to its mineral nature (K-Feldspars, Plagioclase, Quartz, Fe-Mg minerals, etc.); these models are needed by scientists involved in the hydro-mechanical behavior of rock masses taking into account their textural and physical characteristics. Since a numerical model is derived, characterising either the size and shape of various components or the stereological parameters (AA, BA, NA, etc.) or their spatial structure (covariance, etc) is straightforward. Accordingly, we focus our attention on the segmentation process: the partition of the initial image is not easy because we need to get n =4 disjoint sets (one for each mineral component) but the union of these sets must be equal to the initial image itself. It is sometimes easier to work with one channel for one mineral component and with an other one for an other mineral component, so we have often to make a decision in the design of the image processing and segmentation. What mineral component has to be first segmented and how to take the result into account during the segmentation of the others mineral phases? Since quartz is the last phase to crystallise in granitic magma and translucent, it will be the last mineral phase to be identified according to the rule: a pixel that is neither K-Feldspars nor plagioclase or Fe-Mg minerals is given a quartz label. However some problems arise and are discussed in the paper. | |
Deskriptorji | GEOLOGY MATERIALS TESTING PARTICLE SIZE IMAGE PROCESSING, COMPUTER-ASSISTED |