oa A Novel Approach towards Semi-Automated Lithofacies Identification from Image Logs

Visualization is an important aspect of modern hydrocarbon borehole geophysical measurements.<br>Downhole tools are now able to acquire high-resolution 2D and 3D maps of the acoustic and electrical<br>properties of the borehole wall and display them in real time as false-colour images of the formations<br>encountered during drilling. These data now form a huge industry database. However, the<br>interpretation of these images under-utilizes the data.<br>To date, the only regularly used quantitative methodology applied to image log interpretation is for the<br>derivation of orientation data (dip and azimuth). Other, occasional quantitative methods use the<br>resistivity measurements themselves, and not the images. However, from the images themselves,<br>much additional information can be extracted, by using advanced object based image analysis software<br>which is widely available and is successfully employed for analyzing digital images at all scales, from<br>microscopic cell structures to satellite pictures.<br>We present a method for identifying lithofacies from image logs employing image analysis methods<br>used in remote sensing and medical science. The new technique presented synthesizes expert<br>knowledge and digital image analysis, to recognize physically and/or chemically consistent objects<br>within an image and relate these to geologically meaningful groups, such as lithofacies. Filters are used<br>to mark bed boundaries and are created from a derivative log extracted from neutron and density logs<br>and from bed orientation calculated using automated sinusoid fitting at every pixel depth in image log<br>within a formal uncertainty framework. The resultant lithofacies classification is then validated through<br>the interpretation of cored intervals by a geologist.<br>The image interpretation calibrated to core ensures the accuracy in the result obtained and the good<br>match between the two gives the confidence to extrapolate the automated image analysis result from<br>areas with core control to areas with poor to no core recovery. The developed method can be quickly<br>adapted to other wells or applied field wide by defining the lithofacies in each case and by appropriate<br>sample selection for each lithofacies. In addition, the methodology is applicable to several kinds of<br>borehole images, for example wireline electrical borehole wall images, core photographs and the more<br>specialized LWD images.