Hydrocarbon exploration and extraction can be facilitated using machine-learning models. For example, a system described herein can receive seismic data indicating locations of geological bodies in a target area of a subterranean formation. The system can provide the seismic data as input to a trained machine-learning model for determining whether the target area of the subterranean formation includes one or more types of geological bodies. The system can receive an output from the trained machine-learning model indicating whether or not the target area of the subterranean formation includes the one or more types of geological bodies. The system can then execute one or more processing operations for facilitating hydrocarbon exploration or extraction based on the seismic data and the output from the trained machine-learning model.

Methods and apparatus are described for reducing noise in measurements made by one or more pressure sensors disposed in a cable having a generally longitudinal axis. Estimated axial vibration noise at a location along the cable is determined based at least in part on measurements from one or more motion sensors disposed along the cable. The estimated axial vibration noise is subtracted from pressure sensor measurements corresponding to the location. The result is noise-attenuated pressure sensor measurements corresponding to the location.

The present disclosure describes methods and systems, including computer-implemented methods, computer program products, and computer systems, for smoothing seismic data. One computer-implemented method includes obtaining, by a hardware data processing apparatus, a plurality of seismic data samples; forming, by the hardware data processing apparatus, guiding vectors using the plurality of seismic data samples and a plurality of guiding structure attributes; generating, by the hardware data processing apparatus, a structure guided directional weighted vector filter using the guiding vectors and a plurality of weighting factors; filtering, by the hardware data processing apparatus, the seismic data samples using the structure guided directional weighted vector filter to generate smoothed seismic data; and initiating output of the smoothed seismic data.

Methods and devices according to various embodiments perform full-wave inversion jointly for datasets acquired at different times over the same underground formation using a time-lag cost function with target regularization terms. This approach improves the 4D signal within reservoirs and suppresses 4D noise outside.

Methods and devices according to various embodiments perform full-wave inversion jointly for datasets acquired at different times over the same underground formation using a time-lag cost function with target regularization terms. This approach improves the 4D signal within reservoirs and suppresses 4D noise outside.

Systems, media, and methods for processing seismic data are disclosed. For example, in one embodiment, the method may include receiving a plurality of partial image partitions of a migrated seismic image, and stacking the plurality of partial image partitions such that a first image is generated. The method may also include aligning the plurality of partial image partitions based at least partially on the first image. Aligning may include adjusting at least one of the plurality of partial image partitions and generating a displacement field. The method may also include, based at least in part on the displacement field, stacking the plurality of aligned partial image partitions to generate a second image. The method may further include based at least in part on the second image, realigning the plurality of aligned partial image partitions.

Imperfect separation at the higher frequencies has been observed and was eventually was tracked down to the poor GFE signal that is normally used in the inversion. The invention thus uses a “derived GFE” for each source, obtained by comparing the shot records and remove the differences, instead of the prior estimated GFE signal put out by the controller, thus accurately maximizing the separation of the data.

Imperfect separation at the higher frequencies has been observed and was eventually was tracked down to the poor GFE signal that is normally used in the inversion. The invention thus uses a “derived GFE” for each source, obtained by comparing the shot records and remove the differences, instead of the prior estimated GFE signal put out by the controller, thus accurately maximizing the separation of the data.

Methods, systems, and computer-readable medium to perform operations including: generating a first time-frequency spectrum of a first seismic trace from an original seismic dataset; generating a second time-frequency spectrum of a second seismic trace from an enhanced seismic dataset, where the second seismic trace; calculating a difference between the first time-frequency spectrum and the second time-frequency spectrum to generate a noise estimate in the first seismic trace; constructing, based on (i) the noise estimate, (ii) the first time-frequency spectrum, and (iii) the second time-frequency spectrum, a time-frequency mask (TFM); and using the constructed TFM to generate a third time-frequency spectrum of an output trace that corresponds to the first and second seismic traces.

In some embodiments, a time migration diffraction imaging method includes computing a pseudo-depth characterizing a subsurface seismic event by scaling a vertical traveltime using a scaling velocity. A specularity value for a subsurface seismic event is determined according to the pseudo-depth, and a contribution weight for a corresponding seismic trace amplitude is determined according to the specularity value. The specularity value may be determined according to an angle between a traveltime gradient and a normal to a local reflector surface. A diffraction image is generated according to a weighted sum of seismic trace amplitudes. The weighted sum attenuates the contribution of specular events relative to diffraction events.