基于自动编码器的地震随机噪声压制方法

Seismic Random Noise Suppression Method Based on Autoencoder

近些年,随着我国经济飞速发展,地震勘探在油气资源、煤矿资源和铀矿资源勘探中的重要性日益凸显。地震资料处理是地震勘探中的关键环节,而此环节好坏与地震资料的信噪比、分辨率和保真度息息相关。其中,地震随机噪声的压制被认为是提高信噪比步骤中的重要一步。提出一种深度去噪自动编码器的方法去压制地震随机噪声,该方法将地震数据作为输入,然后将其编码成多级抽象,然后将它们重建成没有噪声的地震数据。通过合成数据和实地数据的实验结果可以看出,深度去噪自动编码器比其他传统方法具有更好的去噪效果。

In recent years,with the rapid development of China’s economy,the importance of seismic exploration in the exploration of oil and gas resources,coal resources and uranium resources have been more important.Seismic data processing is a key step of seismic exploration,which is closely related to the signal-to-noise ratio,resolution and fidelity of seismic data.Among them,the suppression of seismic random noise is considered as an important step in the process of improving SNR.This paper proposed a deep denoising autoencoder method to suppress seismic random noise.This method taked seismic data as input,encoded it into multi-level abstractions,and then reconstructed them into seismic data without noise.Experimental results on synthetic data and field data showed that the deep denoising autoencoder had better denoising effect than other traditional methods.