| The advent of high resolution logging devices now allows us to visualize geologic or natural events in greater detail than previously possible. These logs produce sonic or electrical images of the wellbore and have a vertical resolution to tenths of inches. However, the abundance of information can burden the log interpreter. A program to preprocess the images or highlight features of interest might be worthwhile if it could increase the log interpreter's productivity.;In this study, a program for identifying salient features in wellbore images is developed. Conventional image processing routines convert the raw digital data to symbolic forms. The symbolic forms are manipulated at different levels of visual abstraction to divide the pattern recognition problem into smaller, more manageable tasks. Additional information from relevant sources such as previous geological interpretations, conventional well logs, or well data can be used to constrain interpretations or limit searches. Results from five field cases show that the software system can successfully detect image features related to natural fractures, washouts, and sand-shale facies.;The system is implemented on a microcomputer and consists of four parts: interactive wellbore analysis modules, inference engine, specialized knowledge bases, and pattern matching algorithms. |
