HR: 0800h
AN: IN21A-1171    [Abstracts]
TI: General Approach for Rock Classification Based on Digital Image Analysis of Electrical Borehole Wall Images
AU: * Linek, M
EM: m.linek@geophysik.rwth-aachen.de
AF: Applied Geophysics RWTH Aachen University, Lochnerstr. 4-20, Aachen, 52056 Germany
AU: Jungmann, M
IN21A-1171 AF: Fraunhofer Institute for Applied Information Technology, Schloss Birlinghoven, Sankt Augustin, 53754 Germany
AU: Berlage, T
IN21A-1171 AF: Fraunhofer Institute for Applied Information Technology, Schloss Birlinghoven, Sankt Augustin, 53754 Germany
AU: Berlage, T
IN21A-1171 AF: Informatics V RWTH Aachen University, Ahornstr. 55, Aachen, 52056 Germany
AU: Clauser, C
IN21A-1171 AF: Applied Geophysics RWTH Aachen University, Lochnerstr. 4-20, Aachen, 52056 Germany
AB: Within the Ocean Drilling Program (ODP), image logging tools have been routinely deployed such as the Formation MicroScanner (FMS) or the Resistivity-At-Bit (RAB) tools. Both logging methods are based on resistivity measurements at the borehole wall and therefore are sensitive to conductivity contrasts, which are mapped in color scale images. These images are commonly used to study the structure of the sedimentary rocks and the oceanic crust (petrologic fabric, fractures, veins, etc.). So far, mapping of lithology from electrical images is purely based on visual inspection and subjective interpretation. We apply digital image analysis on electrical borehole wall images in order to develop a method, which augments objective rock identification. We focus on supervised textural pattern recognition which studies the spatial gray level distribution with respect to certain rock types. FMS image intervals of rock classes known from core data are taken in order to train textural characteristics for each class. A so-called gray level co-occurrence matrix is computed by counting the occurrence of a pair of gray levels that are a certain distant apart. Once the matrix for an image interval is computed, we calculate the image contrast, homogeneity, energy, and entropy. We assign characteristic textural features to different rock types by reducing the image information into a small set of descriptive features. Once a discriminating set of texture features for each rock type is found, we are able to discriminate the entire FMS images regarding the trained rock type classification. A rock classification based on texture features enables quantitative lithology mapping and is characterized by a high repeatability, in contrast to a purely visual subjective image interpretation. We show examples for the rock classification between breccias, pillows, massive units, and horizontally bedded tuffs based on ODP image data.
DE: 3045 Seafloor morphology, geology, and geophysics
DE: 3280 Wavelet transform (3255, 4455)
DE: 3299 General or miscellaneous
DE: 3625 Petrography, microstructures, and textures
DE: 3699 General or miscellaneous
SC: Earth and Space Science Informatics [IN]
MN: Fall Meeting 2005