Fuzzy Clustering of Seismic Sequences: Segmentation of Time-Frequency Representations

Time-frequency (TF) representations based on minimum mean cross-entropy (MMCE) solution, Bessel kernel, and generalized marginal page distribution are used as the input of clustering. It is shown that general trends of frequency changed with time are not the same for these transforms. The proposed method is based on the knowledge integration of TF transforms followed by a clustering data matrix and the derivation of fuzzy membership values. Based on the output of cluster membership values, the full bandwidth illumination (FBWI) index is defined as a tool for qualitative seismic interpretation. The method is applicable for studying the frequency behavior of reflectors in a reservoir for detecting fluid migration paths.

Introduction

In exploration seismology, reflectors are defined as the locus of reflection points in a seismic ray path. The acoustic impedance change of two overlaying layers is the reason for their formation in seismic sections. Some characteristics of a reflector are similar amplitude strength range, continuity of phase, and coherent energy with respect to other samples within a seismic data ensemble. Due to the complex structure of a subsurface, even a reflector representing a distinguished geological layer shows different seismic character both in spatial and temporal axes. The seismic interpreter would like to examine and group different patterns of a two-dimensional (2-D)/three-dimensional (3-D) seismic reflector and hopefully find a link to lithological or fluid content changes. The realization of data grouping along a reflector or a sequence of reflectors is needed in seismic stratigraphic interpretation.