Computer vision techniques applied to automatic detection of sinusoids in borehole resistivity imaging – A comparison with the MSD method

Abstract

In this research computer vision techniques are applied to borehole resistivity imaging in order to establish an alternative procedure to the mean square dip (MSD) processing. The MSD is regularly applied to detect sinusoids and dips automatically in borehole imaging and dipmeter logs. The present proposal is based on Gabor’s filters, morphological transformations, Hough’s transform, and clustering techniques. The MSD method and the computer vision proposal were tested in 1012 m of images, showing 7.986% of false positives for the MSD processing and 0.879% for the computer vision approach. This methodology tries to emulate the geologists behavior when they make image interpretation; instead of making correlations between resistivity curves like the MSD does. There are no special computer requirements, and it can be applied directly in the field for quick well-site dip results. This procedure can be easily integrated into log units and most commercial borehole-imaging processing software. The processing workflow was developed in python using standard libraries.