Seismic Data Interpolation via Structured Dictionary Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Marine seismic surveys face challenges in generating accurate high-resolution images of subterranean formations due to spatial aliasing caused by coarse spacing of seismic receivers, which results in image artifacts that do not accurately represent structural features, and existing interpolation techniques often rely on a priori information or assumptions about seismic signal morphology.
Innovation Solution
Structured dictionary learning is employed to interpolate seismic data by training a set of basis vectors (atoms) and sparse coefficients on patches of recorded seismic data, constraining atoms to represent linear or nonlinear reflection events, allowing for regularization and interpolation over a finer receiver-coordinate grid, thereby reducing aliasing and noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If coarse spacing of seismic receivers is used, then the cost and complexity of the survey system is reduced, but spatial aliasing occurs causing image artifacts that do not accurately represent structural features
Solution Approach 1:
The patent introduces an intermediary process (interpolation algorithm) that operates between the coarsely spaced receiver measurements and the final seismic image. This intermediary step estimates and fills in the missing fine-scale spatial information that would otherwise require densely spaced receivers, thereby achieving high-resolution images without the associated cost and complexity increases
Solution Approach 2:
The patent replaces the mechanical solution of using densely spaced physical receivers with a computational solution (interpolation algorithms including kriging, inverse distance weighting, and neural networks). This substitution achieves the same goal of reducing spatial aliasing and improving image accuracy without the physical constraints and costs of dense receiver deployment
2Measurement precision
If existing interpolation techniques are used, then some spatial aliasing correction is achieved, but the techniques rely on a priori information or assumptions about seismic signal morphology which limit their applicability
Solution Approach 1:
The patent employs multiple interpolation techniques with different parameters and assumptions (kriging with various variogram models, inverse distance weighting with different power laws, neural networks with different architectures). By changing these parameters and selecting the most appropriate technique for each specific dataset and geological context, the system achieves both accurate spatial aliasing correction and broad adaptability to different survey conditions
Solution Approach 2:
The patent implements a dynamic selection process where the interpolation technique and its parameters are adjusted based on the characteristics of the input seismic data. The system evaluates data quality, noise levels, and geological features to dynamically choose the most suitable interpolation method, thereby achieving both precision in correction and versatility in applicability
Data Source
AI summary
This disclosure is directed to processes and systems that generate enhanced-resolution seismic images by interpolating sparsely recorded seismic data. Structured dictionary learning is employed to train a set of basis vectors, called “atoms,” and corresponding sparse coefficients on patches of the recorded seismic data. The atoms are constrained to represent the geometric structure of reflection events in the recorded seismic data gather. Linear combinations of the atoms are used to compute interpolated patches over a finer receiver-coordinate grid. The interpolated patches replace the original patches in the recorded seismic data to obtain interpolated seismic data that can be used to generate an image of the subterranean formation.