论文详情
奇异值分解(Singular Value Decomposition,SVD)是近几年发展起来的一种地震数据随机噪声压制方法。基于频率域奇异值分解矩阵降秩运算,利用凸集投影(Projection Onto Convex Sets,POCS)迭代算法,实现了地震数据去噪和插值的同步处理,给出了方法的实现步骤。实际资料处理时，采用分窗处理方式减少了算法对内存的需求,降低了插值和去噪处理的运算量,同时使有效信号的同相轴线性化,满足方法的假定条件。模拟数据和实际资料测试均表明,频率域奇异值分解方法可以在压制地震数据噪声的同时进行插值处理,具有广阔的应用前景。
Singular value decomposition (SVD) is a seismic random noise attenuation method developed in recent years.Using the rank reduction character of frequency singular value decomposition,and by a projection onto convex sets (POCS) iterative algorithm,a method that can simultaneously process seismic data interpolation and random noise reduction is realized,and the anti-diagonal averaging process is characterized in detail.Sliding window in real data processing can linearize the seismic events in the localized window,reduce the memory storage requirement,and decrease the computation cost.The method is tested with synthetic data and real datasets,and the results illustrate that the rank reduction method using singular value decomposition in the frequency domain can suppress seismic noise while interpolate seismic data well,which shows a broad application prospect.
国家自然科学基金(41404099)、中国石油大学(北京)科研基金项目(2462015YQ0512)、海洋石油勘探国家工程实验室“斜缆采集地震数据分析与处理技术研究”项目联合资助。