Full text loading...
-
Seismic Random Noise Attenuation via Unsupervised Sparse Machine Learning
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 82nd EAGE Annual Conference & Exhibition, Oct 2021, Volume 2021, p.1 - 5
Abstract
In the field of exploration geophysics, seismic waves received by near-surface geophones are usually corrupted by random noise, which degrades the performance of the following seismic exploration process, such as imaging and inversion. Therefore, random noise attenuation plays an essential step in seismic data processing. In this research, we propose a denoising autoencoder to remove random noise from seismic records. Different from traditional autoencoders that constrain representations, the denoising autoencoder trys to attain appropriate representations by changing the reconstruction criterion, which allows neural network to capture the true seismic wave composition and then attenuate random noise. Compared with the other methods, real data shows that the proposed method achieves better performance in terms of the weak signal preservation and random noise attenuation.