论文部分内容阅读
针对复杂地质结构、陡倾角相干噪声、空间采样不均匀等情况下F-x域反褶积去噪技术的不足,提出首先应用具有时-频聚集性度量准则的广义S变换将时间-空间域的地震数据变换至时间-频率-空间域(t-f-x)的数据,在t-f-x域中对每一个频率切片应用经验模态分解(EMD),移除噪声占主导地位的本征模态函数以压制相干和随机噪声的滤波方法。模型分析表明第一本征模态函数表征的高频信息以噪声为主,移除第一本证模态函数可以达到压制噪声的目的。经广义S变换后形成t-f-x域中EMD滤波方法等效于具有依赖于空间位置、频率、高波数截断特征的自适应f-k滤波。此滤波方法考虑了数据的局部时-频特征,且具有执行简单的特点。与AR预测滤波方法比较,此法滤除的成分包含较少的低波数的信息,滤除的成分非常的局部化,且获得结果没有表现出过度平滑的特征。实际资料的应用表明在经广义S变换后形成t-f-x域中运用EMD滤波方法能够有效地压制随机和陡倾角相干噪声。
In order to overcome the shortcomings of Fx domain deconvolution denoising techniques such as complex geological structure, steep-dip coherent noise and spatial sampling, a generalized S-transform with time-frequency clustering metric is proposed to transmit time-space domain seismic Data is transformed into data in the time-frequency-space domain (tfx), Empirical mode decomposition (EMD) is applied to each frequency slice in the tfx domain, noise-dominant intrinsic mode functions are removed to suppress coherent and random Noise filtering method. The model analysis shows that the high-frequency information characterized by the first eigenmodel function is dominated by noise, and the removal of the first syndrome mode function can achieve the purpose of suppressing noise. The EMD filtering method in the form of t-f-x domain after generalized S transform is equivalent to having adaptive f-k filtering depending on the spatial position, frequency and high wavenumber truncation features. This filtering method takes into account the local time-frequency characteristics of the data and has the characteristics of simple implementation. Compared with the AR prediction filtering method, the components filtered by this method contain less low wave number information, the filtered components are very localized, and the obtained results show no over-smooth features. The application of the actual data shows that the use of EMD filtering in the formation of t-f-x domain after generalized S-transform can effectively suppress random and steep-dip coherent noise.