Quantitative Characterization of River Depositional System Employing Detailed Automated Dip Picking Based on Optimized Ridge Filtering

摘要

Interpretation of different types of reservoir genesis in fluvial settings, both from core and logs exposes significant challenges. Currently within industry the electrical wireline imagers are the most detailed logging methods available and provide unique information making solutions possible to resolve this puzzle. However imager data is rarely used to the full extent in geologic modelling due to its ambiguity and subjectivity in dip and azimuth processing. Current image interpretation most frequently employs a manual dip picking process. This procedure requires significant man hours and potentially introduces human bias. The latter, from our experience, leads to smoothing of the dip and azimuth patterns as well as selecting only the most obvious bed boundary contrasts. Indeed it is critical to capture the entire range of dips for confident sedimentological interpretation. The majority of existing automated dip picking routines are based on pad-to-pad correlation, edge detection and Hough transform. These methods are considerably affected by noise and/or pad depth mismatch and often produce unreliable results. The newly developed automated dip picking routine adopts the ridge filtering methodology to the borehole image data. Data preparation involves overlapping sliding window, stripes removal, pads equalization and morphological filtering. The ridge filter methodology is applied with multiple scale factors creating multiple filtered images to ensure successful boundary picking in noisy or partially noisy datasets. The subsequent contouring routine is the key part of the bed boundary definition process. Apparent dip and azimuth are calculated using minimum least squares fitting of the sinusoid of the contour points. The appropriate bed boundary is selected by statistical analysis of the image area separated by each boundary created from the set of multiple filtered images. The method presented in this paper allows for characterization of every bed boundary resolved by the high resolution resistivity image. The results show a 4.5 refinement, hence increased resolution in bed boundary definition as compared to the conventional, manual approach provided by logging contractors. The matching of the sinusoids to bed boundaries from both methods is also provided to highlight the improved accuracy of this automated method. The thorough and robust characterization of the bedding data now permits recognition of sedimentary bodies of different genesis: braided rivers, meandering rivers and crevasse splays/deltas. The detailed dip and azimuth information allowed estimation of bedsets thicknesses and flow direction. The latter information and its distribution were directly used in the geologic sector model providing a valuable constraint on channel belt width and characteristic geometry.