基于神经网络的地震剖面反褶积新方法

地震剖面的反褶积经常是地震勘探资料处理阶段的一个重要步骤，它是公认的提高剖面分辨率的有效方法．常规的反褶积方法已经很多，但如何在提高剖面分辨率的同时又有效地抑制噪声，提高剖面的信噪比，始终是困扰人们的问题．针对这一问题，提出了一种基于反馈神经网络及其衍生物（ＴＨ网络）的地震剖面反褶积方法．实验结果表明，本方法效果明显，尤其在地震剖面中存在一些相隔较远的反射层时有很好的效果，充分显示了神经网络方法处理地震剖面反褶积问题的有效性．

Seismic Deconvolution Based on Neural Networks

Seismic deconvolution, which is generally recognized as an effective measure to raise the resolution of a seismic section, often plays an important part in seismic data processing. There have been many conventional deconvolution methods, but how to improve the SNR of a seismic section as well as to raise its resolution is still a problem. In this paper, a new deconvolution method based upon feedback neural network models and its derivative TH network is presented. Computer simulating results illustrate that the new method has good performances, especially when there exist rather seperated reflecting layers in the section.