A Robust Automatic Dip Picking Technique to Improve Geological Interpretation and Post-Drill Formation Evaluation of Azimuthal Wellbore Image Logs
Shin-Ju Ye · SPE Annual Technical Conference and Exhibition · 2015
Abstract Automatic dip picking avoids the disadvantages of hand picking (subjective, time-consuming, and ergonomically unfriendly); in particular, it reduces dip uncertainty for low-resolution borehole images. Many automatic dip picking methods have been developed since the 1950's. Each of these methods is tied to a specific type of tool (e.g., dipmeter tools or partial-wellbore-coverage pad-type imagers), and they are unreliable to process high-angle and horizontal (HAHz) wellbore azimuthal image data. This article presents a new method to detect dips in HAHz borehole images. Although this method was originally designed for full wellbore azimuthal images, it is applicable to any dipmeter and pad-type image logs. The proposed automatic dip picking method is simple, yet robust. It consists of the following three main steps: first, determine an optimal sinusoidal trend on the image at each depth using a minimum-variance technique; second, compute the contrast at each depth along the optimal sinusoidal line on the gradient (or first-derivative) image; and finally, locate bedding surface boundaries at depths with the highest contrasts. This method emulates the human pattern recognition process by first observing overall global image sinusoidal trends, and then examining detailed local bed contrasts to locate the bed boundaries. Conventional pattern fitting methods using edge detection and Hough transform (or sinewave fitting) can be unreliable and greatly affected by the noise and artifacts on the image, as these methods focus immediately on detailed and localized image features, and often overlook global image patterns. Although other correlation techniques have been tested, the minimum-variance method is the preferred approach, for it generates the most consistent sinewaves. It correlates all azimuthal bins simultaneously and is much more reliable than two-curve correlation methods. Since the variances are calculated along sinusoidal lines, there is no need to perform individual sinewave fitting. The minimum-variance method combines the conventional, complicated, two-step dipmeter processing methodology (first curve correlation, then sinewave fitting) into one single step to determine an optimal local sinewave. With fewer steps, there are fewer scenarios and problems to consider and fewer parameters to adjust. Moreover, at each depth, the minimum-variance method can find the optimal sinewave within a full range of search parameters (sinusoid amplitude and phase). It can identify faults if visible, as well as reversed sinewaves in “bulls-eye” sections where the wellbore axis is almost parallel to the bedding planes. The proposed method has been applied to many logging-while-drilling density images and dipmeter data from fields worldwide. It has substantially reduced the interpretational time in building earth models from image logs for logging tool response modeling in petrophysical evaluation.