Seismic Image Analysis for Broad and Flat Geologic Feature Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning technologies, such as convolutional neural networks, struggle to accurately identify geologic features of interest in seismic data due to their differences from common visual features, often missing broad and flat features.
Innovation Solution
A deep neural network architecture, such as a U-net, is optimized to detect broad and flat features in seismic images through aspect ratio compression and simultaneous comparison across multiple scales, using 1×2 max pooling operations and 2D convolution transpose operations to analyze seismic data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional convolutional neural networks are used to detect geologic features in seismic data, then the network structure is simple and easy to implement, but broad and flat geologic features are missed due to their difference from common visual features
Solution Approach 1:
The patent segments the detection task by introducing multiple specialized detection heads, each optimized for specific feature types (e.g., broad/flat features vs. compact features). This allows the network to handle different geologic feature characteristics separately, improving overall detection accuracy without requiring a complete redesign of the entire network architecture.
Solution Approach 2:
The patent transforms the input seismic data by applying multiple aspect ratio transformations (e.g., 1:2, 2:1, 1:4, 4:1) to the image dimensions. This dimensional transformation allows the network to detect broad and flat features that would be missed in the original aspect ratio, effectively adding dimensional variability to enhance feature detection capability.
2Adaptability or versatility
If image analysis techniques from non-seismic data are applied to seismic data, then existing machine learning methods can be utilized, but geologic features are missed because they differ from common visual features like faces and objects
Solution Approach 1:
The patent applies local quality by creating specialized detection components tailored to specific geologic feature characteristics. Different detection heads are designed with specific kernels and parameters optimized for detecting particular feature types (e.g., one head for broad features, another for flat features), allowing each part of the network to excel at detecting specific local patterns in seismic data.
Solution Approach 2:
The patent changes key parameters of the neural network architecture to adapt to seismic data characteristics. This includes modifying convolutional kernel sizes, aspect ratios, and detection thresholds to match the unique properties of geologic features in seismic images, which differ fundamentally from common visual features in photographs.
3Measurement precision
If multiple aspect ratio scales are used to detect broad and flat features, then detection accuracy improves, but computational complexity and processing time increase
Solution Approach 1:
The patent performs preliminary action by pre-calculating and storing aspect ratio transformation matrices and detection parameters for different scale levels. During inference, these pre-computed values are directly applied without real-time calculation, significantly reducing processing time while maintaining the benefits of multi-scale analysis for detecting broad and flat geologic features.
Data Source
AI summary
Aspects of the technology described herein identify geologic features within seismic data using modern computer analysis. An initial step is the development of training data for the machine classifier. The training data comprises an image of seismic data paired with a label identifying points of interest that the classifier should identify within raw data. Once the training data is generated, a classifier can be trained to identify areas of interest in unlabeled seismic images. The classifier can take the form of a deep neural network, such as a U-net. Aspects of the technology described herein utilize a deep neural network architecture that is optimized to detect broad and flat features in seismic images that may go undetected by typical neural networks in use. The architecture can include a group of layers that perform aspect ratio compression and simultaneous comparison of images across multiple aspect ratio scales.