Seismic interpretation is now serving as a fundamental tool for depicting subsurface geology and assisting activities in various domains, such as environmental engineering and petroleum exploration. However, most of the existing interpretation techniques are designed for interpreting a certain seismic pattern (e.g., faults and salt domes) in a given seismic dataset at one time; correspondingly, the rest patterns would be ignored. Interpreting all the important seismic patterns becomes feasible with the aid of multiple classification techniques. When implementing them into the seismic domain, however, the major drawback is the low efficiency particularly for a large dataset, since the classification need to be repeated at every seismic sample. To resolve such limitation, this study first present a seismic pattern interpretation dataset (SpiDat), which tentatively categorizes 12 commonly-observed seismic patterns based on their signal intensity and lateral geometry, including these of important geologic implications such as faults, salt domes, gas chimneys, and depositional sequences. Then we propose a seismic pattern interpretation network (SpiNet) based on the state-of-the-art deconvolutional neural network, which is capable of automatically recognizing and annotating the 12 defined seismic patterns in real time. The impacts of the proposed SpiNet come in two folds. First, applying the SpiNet to a seismic cube allows interpreters to quickly identify the important seismic patterns as input to advanced interpretation and modeling. Second, the SpiNet paves the foundation for deriving more task-oriented seismic interpretation networks, such as fault detection. It is concluded that the proposed SpiNet holds great potentials for assisting the major seismic interpretation challenges and advancing it further towards cognitive seismic data analysis.
As a fundamental tool for characterizing subsurface geology, three-dimensional (3D) seismic interpretation plays a crucial role in various disciplines, such as civil engineering, geohazard assessment, and energy exploration. Interpreting a seismic volume is a time-consuming and labor-intensive process and often requires mutual collaborations between geologists, geophysicists, petrophysicists, and more. Manual interpretation has been the most straightforward and effective approach for solving this problem, in which an interpreter visually analyzes the seismic reflection patterns, identifies the important patterns, and labels them by distinct marks and/or colors. However, the dramatically increasing size of 3D seismic surveying is now significantly challenging the efficiency of such manual interpretation.
For accelerating the interpretation process, geoscientists have made great efforts into developing a full suite of computer-aided tools, such as edge detection, geometry estimation, facies analysis, object extraction, and more. However, most of these tools are designed for interpreting one or some certain features by analyzing seismic signals from different perspectives. Correspondingly, the rest features present in a seismic dataset would be undesirably ignored. For example, as the first edge-detection tool, the coherence attribute (Bahorich and Farmer, 1995) estimates the lateral similarity of seismic waveforms and thereby is effective in depicting the faults and stratigraphic features that obviously break the waveform continuity. Since its popularity, a number of variations and schemes have been developed in improving such attribute (e.g., Luo et al., 1996;Marfurt et al., 1998;Gersztenkorn and Marfurt, 1999;Cohen and Coifman, 2002;Tingdahl and de Rooij, 2005;Di and Gao, 2014;Wang et al., 2016). While clearly highlighting the major faults of apparent displacements, however, most of the edgedetection tools are less efficient for subtle structure interpretation, such as fracture characterization and facies analysis, in which the lateral variation of seismic signals is subtle and beyond the resolution of edge detectors. Detailed summaries of the edge detection can be found in Chopra (2002), Kington (2015), and Di and Gao (2017a). For the purpose of detecting the small-scale structures like subtle faults and fractures, geophysicists then turn to evaluating the variation of the geometry of seismic reflectors, which successfully link the fractures with the high-order reflector geometric attributes, such as curvature (Roberts, 2001) and flexure (Gao, 2013). A suite of schemes is also available for such geometry estimation, whose efficiency in identifying planar seismic structures, such as fractures, has been documented in various case studies (e.g., Di andGao, 2014b, 2017b;Gao and Di, 2015;Yu andLi, 2017a, 2017b;Qi and Marfurt, 2017). However, such geometric analysis often fails for stratigraphic features, such as channels, reefs, lobes, and overbanks. Instead, accurate stratigraphic interpretation becomes possible by performing seismic facies analysis, particularly the GLCM analysis that estimates the local arrangement of seismic amplitudes in 3D space (Gao, 1999;Eichkitz et al., 2013;Di and Gao, 2017c). The GLCM tool is based on the fact that rock particles are packed in different ways with the depositional environment varying, and correspondingly, the reflection patterns are locally different in terms of their amplitude, frequency, and/or phase.
While depicting the target seismic pattern from the surrounding ones, however, these techniques fail to extract themas separate objects that can be readily fed into framework construction and modeling. For example, a salt body can be visually depicted as high homogeneity and low contrast in GLCM maps (Gao, 2003), but isolating it from the surrounding patterns requires additional tools used in computer graphic and imaging processing. For example, normalized cuts (Lomask et al., 2007) detects salt domes by solving a global optimization problem. The active-contour-models method (Shafiq et al., 2015) starts with the initial boundary from interpreters and then gradually deform it to fit the salt boundary observed in the attribute image. Wu (2016) incorporates discrete pickings by an interpreter into the detection process to guide accurate delineation of salt boundaries, especially in complicated zones with gaps or outliers. Ramirez et al. (2016) adopt the theory of sparse representation and apply it to automatically segment salt structures from 3D seismic dataset. Similarly, (semi-)automatic fault extraction has been popular in the past years with numbers of algorithms presented in this field, including ant tracking (Pedersen et al., 2002), Hough transform (AlBinhassan andMarfurt, 2003), eigenvector analysis (Barnes, 2006), dynamic time wrapping (Hale, 2013), motion vector (Wang et al., 2014), and more.
However, such object extraction often works for one certain structur
This content is AI-processed based on open access ArXiv data.
