Machine learning techniques for noise attenuation in geophysical surveys
AssigneePGS Geophysical AS
InventorMaiza Bekara
Life of this patent
Estimating the term…
Abstract
Techniques are disclosed relating to machine learning in the context of noise filters for sensor data, e.g., as produced by geophysical surveys. In some embodiments, one or more filters are applied to sensor data, such a harsh filter determined to cause a threshold level of distortion in measured reflections, a mild filter determined to leave a threshold level of remaining noise signals, or an acceptable filter. In some embodiments, the system trains a machine learning classifier based on outputs of the filtering procedures and uses the classifier to determine whether other filtered sensor data from the same survey exhibits acceptable filtering. This may improve accuracy or performance in detecting unacceptable filtering, in some embodiments.
Claims25
In a technological process for generating an image of subsurface geological features based on recorded seismic data, in which at least one stage of a processing chain includes applying a filter to the seismic data and determining if application of the filter acceptably removed noise from the seismic data while preserving desired signal information present in the seismic data, the specific improvement comprising automating the determining step by causing one or more computer devices to perform actions comprising:
- accessing stored geophysical data;
- filtering at least a first subset of the geophysical data with at least first and second filters corresponding to first and second different filter properties, respectively, the first filter producing a first filter output and the second filter producing a second filter output;
- associating a first classification with the first filter output and a second classification with the second filter output;
- training a computer-implemented classifier with training data comprising at least the first subset of the geophysical data, the first and second filter outputs, and the first and second classifications;
- filtering a second subset of the geophysical data with a third filter to produce a third filter output;
- using the trained, computer-implemented classifier to generate a third classification based on the second subset of the geophysical data and the third filter output, wherein at least one of the training and using steps comprises generating at least one similarity attribute that quantifies a degree of similarity between filter output values and difference values, wherein the difference values represent differences between the filter output values and corresponding filter input values;
- and
- making a quality control determination, based on the third classification, regarding acceptability of the third filter;
- if the third filter is accepted, storing at least one output of the third filter in a noise-attenuated geophysical dataset, otherwise not storing the at least one output of the third filter in the noise-attenuated geophysical dataset, wherein the at least one output of the third filter comprises a result of applying the third filter to at least one subset of the stored geophysical data;
- generating an image of a sub-surface feature based on the noise-attenuated geophysical dataset;
- and
- storing the image, or data representative thereof, in one or more non-transitory computer-readable media, thereby completing the manufacture of a geophysical data product.
One or more non-transitory computer readable media having instructions stored therein that, when executed by one or more processors, cause the one or more processors to perform a quality control method using geophysical data, the method comprising:
- filtering at least a first subset of the geophysical data with at least first and second filters corresponding to first and second different filter properties, respectively, the first filter producing a first filter output and the second filter producing a second filter output;
- associating a first classification with the first filter output and a second classification with the second filter output;
- training a computer-implemented classifier with training data comprising at least the first subset of the geophysical data, the first and second filter outputs, and the first and second classifications;
- filtering a second subset of the geophysical data with a third filter to produce a third filter output;
- using the trained, computer-implemented classifier to generate a third classification based on the second subset of the geophysical data and the third filter output, wherein at least one of the training and using steps comprises generating at least one similarity attribute that quantifies a degree of similarity between filter output values and difference values, wherein the difference values represent differences between the filter output values and corresponding filter input values;
- making a quality control determination based on the third classification;
- and
- if the third filter is accepted, storing at least one output of the third filter in a noise-attenuated geophysical dataset, otherwise not storing the at least one output of the third filter in the noise-attenuated geophysical dataset, wherein the at least one output of the third filter comprises a result of applying the third filter to at least one subset of the stored geophysical data;
- and
- generating an image of a sub-surface feature based on the noise-attenuated geophysical dataset.
A system for automating filtration quality determinations in a processing chain that generates an image of subsurface geological features based on recorded geophysical data, comprising:
- one or more non-transitory computer readable media having at least a subset of the geophysical data stored thereon;
- a filter;
- an attribute determination module configured to generate at least one similarity attribute that quantifies a degree of similarity between filter output values and difference values, wherein the difference values represent differences between the filter output values and corresponding filter input values;
- and
- a trained classifier configured to classify, based at least in part on an output of the attribute determination module, results of applying the filter to at least the subset of the geophysical data;
- wherein the system is configured to store at least one output of the filter in a noise-attenuated geophysical dataset if the filter is accepted by the classifier, otherwise not to store the at least one output of the filter in the noise-attenuated geophysical dataset, wherein the at least one output of the filter comprises a result of applying the filter to at least one subset of the stored geophysical data;
- and
- an image generation module configured to generate an image of a sub-surface feature based on the noise-attenuated geophysical dataset.
Dependencies are inferred from the claim text.