Geophysical Potential Field Gridding With Diffusion-Based Denoising
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for gridding and denoising geophysical potential field data suffer from low accuracy due to irregular spatial distribution and noise, which affects the precision of data interpretation.
Innovation Solution
A method involving stochastic forward modeling to generate noisy data sets, training a diffusion model, and using a noise level selection network to achieve accurate gridding and denoising through Euclidean distance-based scattered point data generation and iterative denoising processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional numerical computation methods or equivalent source methods are used for gridding, then the processing can be completed, but the gridding accuracy is low
Solution Approach 1:
The patent replaces conventional numerical computation methods and equivalent source methods with a deep learning-based neural network model. The system uses a diffusion model trained on synthetic potential field data to perform gridding, substituting traditional mathematical computation with an intelligent algorithm that achieves superior accuracy in reconstructing potential field values at grid points from irregularly distributed measurement data
Solution Approach 2:
The patent transforms the gridding problem by changing the approach from direct numerical computation to a probabilistic diffusion process. The diffusion model learns the underlying data distribution and generates gridded data through iterative denoising steps, where the number of denoising steps is dynamically determined based on noise level classification, thereby improving gridding accuracy
2Measurement precision
If traditional denoising methods are used, then processing can be performed, but the denoising effect is poor when there is a huge difference between noise distribution of measured data and synthetic data sets
Solution Approach 1:
The patent performs preliminary action by training the diffusion model on a large dataset of synthetic potential field data with various noise levels before actual denoising. This pre-training enables the model to learn robust denoising patterns that generalize well to real measured data with different noise characteristics, solving the adaptability problem without requiring retraining
Solution Approach 2:
The patent introduces a noise level classification network as an intermediary that assesses the noise characteristics of input data and determines the appropriate number of denoising steps. This intermediary component bridges the gap between diverse noise distributions and the fixed diffusion model, enabling adaptive denoising performance across different data types and noise conditions
Data Source
AI summary
A method for gridding and denoising of geophysical potential field data includes: first, establishing an original geophysical potential field data and a noisy data set of a geophysical potential field; inputting specified original geophysical potential field data into a diffusion model for training to obtain a qualified diffusion model; second, obtaining a number of denoising steps corresponding to the noisy data set, and training a noise level selection network by using the number of denoising steps together with the noisy data set; then, determining positions of scattered point data according to coordinates and anomalies of the geophysical potential field data to be processed, and assigning the anomalies at corresponding positions to generate the scattered point data; and finally, gridding the scattered point data by using the diffusion model, and denoising the gridded data by using a noise level selection network model in combination with the diffusion model.