Search

Diffraction wavefield contains valuable information on subsurface composition through velocity extraction and sometimes anisotropy estimation. It can also be used for the delineation of geological features, such as faults, fractures and mineral deposits. Diffraction recognition is, therefore, crucial for improved interpretation of seismic data. To date, many workflows for diffraction denoising, including deep-learning applications, have been provided, however, with a major focus on sedimentary settings or for ground-penetrating radar data. In this study, we have developed a workflow for a self-supervised learning technique, an autoencoder, for diffraction denoising on synthetic seismic, ground-penetrating radar and hardrock seismic datasets. The autoencoder provides promising results especially for the ground-penetrating radar data. Depending on the target of the studies, diffraction signals can be tackled using the autoencoder both as the signal and/or noise when, for example, a reflection is a target. The real hardrock seismic data required additional pre- and post-autoencoder image processing steps to improve automatic delineation of the diffraction. Here, we also coupled the autoencoder with Hough transform and pixel edge detection filters. Along inlines and crosslines, diffraction signals have sometimes a similar character as the reflection and may spatially be correlated making the denoising workflow unsuccessful. Coupled with additional image processing steps, we successfully isolated diffraction that is generated from a known volcanogenic massive sulphide deposit. These encouraging results suggest that the self-supervised learning techniques such as the autoencoder can be used also for seismic mineral exploration purposes and are worthy to be implemented as additional tools for data processing and target detections.

Diffraction imaging can help better understand small-scale geological structures. Due to their often-weak signal, in order to image them, it is necessary to separate diffraction signals from the rest of the wavefield. Many different methods have been developed for diffraction wavefield separation, and the newest trend includes the application of artificial neural networks and deep learning. Available case studies with a deep-learning approach for diffraction separation show good results when applied to synthetic and sedimentary setting datasets where diffraction signals are either strong or have pronounced characteristics. Examples, however, are missing from crystalline or hardrock geological settings where the signal-to-noise ratio is by far lower and diffraction signals are usually within a complex reflectivity medium, have steep tails and are usually incomplete. In this study, we showcase the application of a deep semantic segmentation model on synthetic seismic, real ground-penetrating radar, and hardrock seismic datasets. Synthetic seismic sections were generated using different random noise levels and coherent noise resembling a complex reflectivity pattern interfering with diffraction tails. For the real GPR dataset, diffraction signals were successfully delineated, although in some locations reflections were picked up because of their similar pixel values as the apex of the diffractions. As for the real seismic dataset, through a number of approaches, we were able to completely delineate a single diffraction within several inlines that was generated from a massive sulphide body. The algorithm also enabled us to recognize an incomplete diffraction, at the edge of the seismic cube, which was never labelled. This diffraction originated from outside of the seismic volume and may be a target for future mineral exploration programmes, thanks to the deep semantic segmentation algorithm providing this possibility.