Welcome to GeoPatterns

Pattern Recognition Technology

The hunt for hydrocarbons is driven by pattern recognition. Subsurface geology is imaged using seismic data and well logs. Both seismic data and well logs are analyzed by using pattern recognition. Wiggles, wavelets and trace excursions are "recognized" to infer something about the subsurface geology.

The most effective computational engine developed to date to handle this complex problem is the human mind. The oil finder becomes an "expert" at recognizing patterns in seismic and well data through years of training and experience. But the oil finder's expertise is a perishable commodity.

 In addition to the unique pattern engine, GeoPatterns has developed a voxel visualization system that utilizes false color imagery in a 32-bit color environment unlike any other seismic interpretation tool on the market. This proprietary 3-D visualization tool greatly enhances the explorationist's ability to see the important patterns in the data and enhances his oil finding capabilities.

The GeoPatterns toolkit allows our explorationists to quickly and intuitively identify the expression of the play concept in the data, and to move from concept to qualified lead to drillable prospect in the shortest possible time without sacrificing either attention to detail or completeness. Subtle anomalies and even more subtle reservoir details are unlocked from the patterns as never before.
 

Pattern Recognition Hierarchy

GeoPatterns' approach to analyzing complex data sets, like 3-D seismic, adheres to the established pattern recognition principles. The basic pattern recognition hierarchical units are features, patterns and textures. Features are the basic fractal unit. At the next level, Patterns are a collection of features; at the next level in the hierarchy, Textures are a collection of Patterns and Features. Finally, Features, Patterns and Textures combine to represent petrophysically calibrated objects, which we call geobodies. Ultimately, it is the interpreter who assigns the interpretation of the observed anomalies in pattern-space based on the geobody's shape, thickness, depositional environment, size and prospectivity.

GeoPatterns uses standard pattern recognition vocabulary that distinguishes features, patterns and textures – from the smallest fractal unit to higher order units, in hierarchal order. "Attribute" is used to describe all fragment-based calculations. A template is the combination of several attributes that highlights and differentiates the zone of interest from the surrounding geology. A palette constitutes the unique cut-off parameters (decision boundaries) of each attribute used during the calibration and extraction proce

Pattern Recognition Methodology

After the petrophysical data has been time-tied to the inverted seismic, the first step in GeoPatterns' process is to generate a pattern database. Depending upon the data quality and project objectives, attributes are generated from either a band limited inversion or a broadband seismic inversion. GeoPatterns creates a database of fractal based features, patterns, and textures that are geologic based patterns.

The scope and objective of many studies is to design a template and palette based on certain well control with sonic and density information. The accuracy of the generated template is dependent upon quality well control. Well control serves as a constraint so that precise cut-offs can be established to uniquely identify the region of interest. Most individual features, patterns and textures are examined. The more difficult challenge is to intuit those combination of attributes, which, when combined in certain way, uniquely identify and highlight the desired geologic zone of interest.

The next step in the interpretive process is to create various visual realizations of the data in pattern-space by using false color imagery to view select portions of the GeoPatterns' pattern. This can be done either deterministically or heuristically, depending upon the data quality and problem to be solved. The interpreter scans the data cube in pattern space for geologic features of interest, identifying and isolating them with variations of the palette. Once this visualization template has been generated, a slight variant in the palette is made in order to extract only that portion of the data that matches the desired criteria.

Geobodies

Extracted geobodies represent sets of voxels of similar interference pattern resulting from a combination of GeoPattern's attributes. These extracted geobodies have similar multi-loop acoustic signatures. Each of the extracted geobodies is the product of a unique multi-attribute combination. The geobody extraction process involves applying the palette to the data set, which probes each voxel with the desired search criteria. Once each voxel in the data cube has been detected, each voxel identifies adjacent voxels on six sides and begins to build itself into connected geobodies. This data mining exercise can generate up to several hundred unqualified leads, ranked by voxel count.