基于优化fastICA盲源分离算法的地震属性融合方法研究

地震属性的种类很多,但存在冗余性问题。提出了基于盲信号理论,以负熵为目标函数,采取初值降敏感性和5阶收敛速度改进迭代公式对快速独立分量分析进行优化的算法,进而实现盲源分离。本算法采用贝叶斯方法构造满足非高斯分布的阈值函数,在变换域中进行信噪分离。针对属性中存在信息冗余的情况,设计了4种图像融合规则进行属性融合,实现了地震属性数据去噪和融合的一体化处理。理论模型及实际资料试验结果表明,该方法能够有效压制地震数据中的随机噪声,可获得比常规阈值去噪方法更高的峰值信噪比,使融合结果中的地质信息更加清晰,从而有利于相关地质特征的解释。

There are many kinds of seismic attributes,but there is redundancy among these.An improved fast independent component analysis (fastICA) algorithm,to perform blind source separation,is proposed.Based on blind signal theory,the algorithm uses negative entropy as the objective function,adopting low initial sensitivity and the 5th order convergence rate to speed up the iterative formula.This algorithm uses the Bayesian method to construct a threshold function that satisfies the non-Gaussian distribution,and performs signal and noise separation in ICA domain.In view of the information redundancy of attributes,four different image fusion rules are designed to fuse the seismic attributes,to realize the integrated processes of denoising and fusion.Test results on synthetic and field data have shown that this seismic attribute fusion approach can effectively suppress random noise.Compared to the conventional threshold denoising method,the proposed approach could help obtain a higher peak SNR,and keep more geologic information in the fusion results,which is conducive to the improved interpretation of relevant geological features.

国家科技重大专项(2016ZX05006-002)资助。