Optically Derived Numerical Descriptors for Classification of Carbonate Rock Fabrics

Abstract

Abstract Carbonate rocks, although relatively simple in their basic mineralogy, present a bewildering array of constituents, textures, structures, and diagenetic features, as well as complex pore geometry. For these reasons, detailed quantitative studies of carbonate rocks require a large investment of time on the part of the carbonate petrologist simply for the purpose of data gathering. Conversely, purpose of data gathering. Conversely, if large numbers of samples are to be studied or if time and cost are important factors, relatively little information is extracted from the available samples. Derivation of petrographic data from carbonate rocks may be considered a problem of image analysis. One approach to such problems is to utilize techniques and algorithms developed in the field of pattern recognition. Identification of rock components and characterization of rock fabric basically involve signature analysis and pattern interpretation. In addition to the characterization of rock constituents, it is equally important to define the nature and distribution of rock pore systems, which may be at least as complex as grain fabrics. It is possible to identify individual constituents of carbonate rocks, such as biotic constituents, by a multispectral process. with this procedure, the reaction process. with this procedure, the reaction of different phases to staining agents, differences in surface texture, crystal fabric, and other "differences in appearance" are combined to achieve unique spectral signatures for each class of constituents. Presently available equipment, originally Presently available equipment, originally designed for analysis of multispectral radar and light imagery from satellites, can classify ten different constituents identified by up to five different spectral characteristics. Once rock constituents have been properly identified, individual particles can be counted, their areas particles can be counted, their areas computed, and statistical measures derived from the image. Equipment can be computer-interfaced for immediate data reduction and display.