Double-Windowed Seismic Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing seismic data analysis methods often miss subtle or unexpected hydrocarbon anomalies due to their reliance on template-based approaches that are not adaptive enough, leading to biased detection towards prominent anomalies and suppression of weaker ones.
Innovation Solution
A double-windowed statistical analysis method using a pattern window and a sampling window of user-selected size and shape to compute statistical distributions and outlier probabilities across multiple spatial scales, allowing for the detection of hydrocarbon presence without prior training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single moving window is used to gather statistics for anomaly detection, then the detection process is simple and fast, but the results are biased towards prominent anomalies and subtle anomalies are suppressed
Solution Approach 1:
The single window approach is segmented into two distinct windows: a pattern window for defining local seismic patterns and a sampling window for gathering statistics. This segmentation allows the system to separately handle pattern recognition and statistical analysis, preventing dominant patterns from overwhelming subtle anomalies while maintaining computational efficiency.
Solution Approach 2:
The invention adds a hierarchical dimension to the analysis by nesting the pattern window within the sampling window. This creates a multi-scale analysis framework where statistics are gathered at different spatial resolutions, enabling detection of both prominent and subtle anomalies without sacrificing detection speed.
2Ease of operation
If template-based or model-based approaches are used to search for known patterns, then the search is directed and efficient, but subtle or unexpected anomalies that do not conform to specifications are missed
Solution Approach 1:
The system performs self-service by automatically learning local seismic patterns directly from the data itself through the pattern window, rather than relying on pre-defined templates or models. This enables the system to adapt to any anomaly type present in the data, including unexpected or subtle features, while maintaining search efficiency through automated pattern recognition.
Solution Approach 2:
The invention changes the fundamental parameter from fixed template characteristics to dynamically learned pattern characteristics. By allowing patterns to be defined by the data itself through statistical analysis in the pattern window, the system becomes adaptable to any anomaly type while maintaining operational efficiency.
3Reliability
If a larger sampling window is used to improve statistical reliability, then the statistical distribution is more reliable, but local subtle anomalies are masked by broader regional trends
Solution Approach 1:
The invention applies local quality by using a nested window structure where the pattern window focuses on local seismic patterns while the sampling window provides broader statistical context. This allows statistics to be gathered reliably from a larger area while maintaining sensitivity to local subtle anomalies through the smaller pattern window that defines the specific pattern of interest.
Data Source
AI summary
Method for identifying geologic features from seismic data (11) using seismic anomaly detection by a double-windowed statistical analysis. Subtle features that may be obscured using a single window on the data are made identifiable using two moving windows of user-selected size and shape: a pattern window located within a sampling window larger than the pattern window (12). If Gaussian statistics are assumed, the statistical analysis may be performed by computing mean and covariance matrices for the data within the pattern window in its various positions within the sampling window (13). Then a specific measure of degree of anomaly for each voxel such as a residue value may be computed for each sampling window using its own mean and covariance matrix (14), and finally the resulting residue volume may be analyzed, with or without thresholding, for physical features indicative of hydrocarbon potential (15).