Robust CSEM data processing by unsupervised machine learning
| Autor: | Liu Xiaoqiong, Juzhi Deng, Guang Li, Changming Shen, Jingtian Tang, Zhushi He, Youyao Fu |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Data processing
010504 meteorology & atmospheric sciences Series (mathematics) business.industry Computer science Ambient noise level Pattern recognition 010502 geochemistry & geophysics 01 natural sciences Geophysics Operator (computer programming) Distortion Unsupervised learning Artificial intelligence business Reliability (statistics) Data selection 0105 earth and related environmental sciences |
| Zdroj: | Journal of Applied Geophysics. 186:104262 |
| ISSN: | 0926-9851 |
| Popis: | The ambient noise in controlled-source electromagnetic (CSEM) data seriously affects the accuracy and reliability of the exploration result. Traditional correlation-based data selection method requires manually setting the threshold. To overcome the deficiency, we analyze the typical noises in CSEM data and find that normalized cross-correlation (NCC), absolute maximum value of the amplitude (Max), and detrend fluctuation analysis (DFA) can be used to accurately identify high-quality time series. Based on this discovery, we replace traditional manually intervention with unsupervised machine learning and propose a novel CSEM data processing method. We applied the newly proposed method to synthetic and measured CSEM data to verify the feasibility and effectiveness. Experimental results demonstrate that the newly proposed method is superior to the conventional data selection method because it accurately selects the best data fragments from noisy data automatically. The newly proposed method requires no human intervention which makes the results obtained free of subjective distortion caused by the operator. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect