Seismic Image Alignment via Optical Flow and Gaussian Pyramids
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Seismic survey images of subsurface regions over time show changes due to hydrocarbon extraction, making it difficult to accurately align and analyze the data for efficient hydrocarbon extraction operations.
Innovation Solution
A computing system processes seismic images by determining a multi-dimensional displacement volume using optical flow algorithms and Gaussian pyramids to align seismic images acquired at different times, overcoming cycle-skipping issues and improving signal-to-noise ratio, thereby identifying changes in subsurface properties like hydrocarbon reservoir characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If seismic images are directly aligned at full resolution, then alignment precision is improved, but computational complexity and processing time increase significantly
Solution Approach 1:
The patent divides the alignment process into multiple resolution levels using a pyramid structure. Images are processed at coarse resolutions first to capture large-scale displacements, then progressively refined at finer resolutions. This segmentation allows the system to achieve high alignment precision without the computational burden of processing full-resolution images from the start.
Solution Approach 2:
The patent performs preliminary alignment at lower resolutions before processing at full resolution. By pre-computing displacement fields at coarse levels and using them as initial estimates for finer levels, the system reduces the computational search space and accelerates convergence, thereby reducing overall computational complexity while maintaining precision.
2Measurement precision
If optical flow algorithms are applied at full resolution, then displacement measurement accuracy is improved, but computational time increases
Solution Approach 1:
The patent segments the computational domain into a pyramid of resolutions. Optical flow algorithms are applied sequentially from coarse to fine levels, with each level contributing to the final accurate displacement measurement. This approach achieves full-resolution accuracy while reducing total computational time by distributing the workload across multiple resolution levels.
Solution Approach 2:
The patent computes displacement fields at lower resolutions as preliminary steps before applying optical flow at full resolution. These preliminary computations provide initial estimates that guide the full-resolution optical flow algorithm, reducing its convergence time and overall computational burden while maintaining measurement accuracy.
3Reliability
If large shifts between seismic images are handled directly, then alignment robustness is improved, but cycle-skipping issues worsen
Solution Approach 1:
The patent segments the displacement estimation into hierarchical levels. At coarse resolutions, large shifts are captured robustly without cycle-skipping. These estimates are then refined at finer resolutions, combining the robustness of coarse-level handling with the precision of fine-level measurement, thereby resolving both requirements simultaneously.
Solution Approach 2:
The patent introduces a resolution dimension to the alignment problem. By transforming the problem from a single-resolution challenge to a multi-resolution hierarchy, the system can handle large shifts robustly at coarse levels while achieving precise measurements at fine levels, effectively resolving the contradiction between robustness and precision.
4Loss of information
If multi-dimensional displacement volume is determined using traditional methods, then subsurface change detection is achieved, but signal-to-noise ratio deteriorates
Solution Approach 1:
The patent segments the image processing into a pyramid structure, processing images at multiple resolutions before combining results. This segmentation allows noise to be filtered at each level while preserving genuine subsurface changes, improving the signal-to-noise ratio without sacrificing the precision needed for accurate change detection.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
A method for aligning a plurality of seismic images associated with a subsurface region of the Earth may include receiving the seismic images and determining a first respective relative shift volume between a first seismic image and a second seismic image, a second respective relative shift volume between the first seismic image and a third seismic image, and a third respective relative shift volume between the second seismic image and the third seismic image. The method may include determining a first shift volume associated with the first seismic image and a second shift volume associated with the second seismic image based on the first, second, and third respective relative shift volumes. The method may then apply the first shift volume to the first seismic image and the second shift volume to the second seismic image.