Pattern recognition in hydrocarbon exploration

The problem of pattern recognition is, simply stated, that of assigning a classification to members of a set of objects based on measurements made on the objects. In a typical application problem, the relationship between the various possible classifications of a given object and the measurements is unknown or is very complex. In such instances, we turn to machines to examine the measurements and assist us in making the classification or to help us unravel the complex relationships between the classifications and measurements on a set of objects whose classifications are known a priori. In the context of hydrocarbon exploration, we have tried to answer the question: Is a particular portion of the earth's subsurface in a favorable situation for hydrocarbons to occur? The measurements to be used must be made on geophysical data observable at the surface of the earth. We have restricted the study to the use of only seismic reflection data for measurements.