基于核主成分分析的地震属性优化方法及应用
|
摘要
传统的基于线性变换的主成分分析法(PCA)是一种有效的地震属性降维优化方法。但是,当原始数据中存在非线性属性时,用主成分分析法提取的主成分就不能反映这种非线性属性。而核主成分分析(KPCA)则是一种基于原始数据的非线性变换,它可以提取出数据之间的非线性关系。本文从方法原理概述入手,分析了一般主成分分析在处理非线性问题上存在的不足,阐述了基于核函数的主成分分析方法,并将其首次应用于地震属性的降维优化中。应用结果表明:基于核函数的主成分分析方法具有优秀的特征提取性能。
Traditional principal analysis method(PCA)based on linear transform is effective method of seismic attribute dimension-reducing optimization.However,the principle component detected by PCA method can't reflect the non-linear attributes if there exists non-linear attribute in raw data.The KPCA is non-linear transform based on the raw data,which can detect the non-linear relationship between the data.Starting from the principle description of the method,the paper analyzed the shortcomings existing in handling the non-linear issue by common PCA,expounded the PCA method based on kernel function,and used the method for the dimension-reducing optimization of seismic attribute for the first time.The application showed the PCA method based on kernel function has wonderful character-detected property.
Traditional principal analysis method(PCA)based on linear transform is effective method of seismic attribute dimension-reducing optimization.However,the principle component detected by PCA method can't reflect the non-linear attributes if there exists non-linear attribute in raw data.The KPCA is non-linear transform based on the raw data,which can detect the non-linear relationship between the data.Starting from the principle description of the method,the paper analyzed the shortcomings existing in handling the non-linear issue by common PCA,expounded the PCA method based on kernel function,and used the method for the dimension-reducing optimization of seismic attribute for the first time.The application showed the PCA method based on kernel function has wonderful character-detected property.
引文
[1]何必朱,周杰,汪功怀.利用多元地震属性预测储层信息.石油地球物理勘探,2003,38(3):258～262
[2]赵加凡,陈小宏.基于灰关联分析与K-L变换的双重属性优化方法.石油地球物理勘探,2004,39(6):661～665
[3]王永刚,刘伟,黄国平.地震属性的GA-BP优化方法.石油地球物理勘探,2002,37(6):606～611
[4]Carreira-Perpian M A.A Review of Dimension Re-duction Techniques.Technical Report CS-96-09,De-part ment of Computer Science,University of Shef-field,1997,5～20
[5]Vladimir N V.The nature ofstatistical learningthe-ory.NewYork:Springer-Verlag,1995,20～30
[6]Scholkopf B et al.Nonlinear component analysis as a kernel eigenvalue problem.Neural Computation,1998,(10):1299～1319
[7]Scholkopf B et al.Kernel principal component anal-ysis.In Proceedings ICANN,LNCS,Springer,1997,583～589
[8]王新峰,邱静,刘冠军.基于核最优K-L变换的故障特征提取方法研究.机械科学与技术,2006,25(3):288～291
[9]印兴耀,周静毅.地震属性优化方法综述.石油地球物理勘探,2005,40(4):482～489
[10]Vladimir N V.Statistical learning theory.New York:John Wiley,1998,115～163
[11]肖建华,吴今培.基于核的特征提取技术及应用研究.计算机工程,2002,28(10):36～38
[12]黄国宏,邵惠鹤.核主元分析及其在人脸识别中的应用.计算机工程,2004,30(13):13～14
[2]赵加凡,陈小宏.基于灰关联分析与K-L变换的双重属性优化方法.石油地球物理勘探,2004,39(6):661～665
[3]王永刚,刘伟,黄国平.地震属性的GA-BP优化方法.石油地球物理勘探,2002,37(6):606～611
[4]Carreira-Perpian M A.A Review of Dimension Re-duction Techniques.Technical Report CS-96-09,De-part ment of Computer Science,University of Shef-field,1997,5～20
[5]Vladimir N V.The nature ofstatistical learningthe-ory.NewYork:Springer-Verlag,1995,20～30
[6]Scholkopf B et al.Nonlinear component analysis as a kernel eigenvalue problem.Neural Computation,1998,(10):1299～1319
[7]Scholkopf B et al.Kernel principal component anal-ysis.In Proceedings ICANN,LNCS,Springer,1997,583～589
[8]王新峰,邱静,刘冠军.基于核最优K-L变换的故障特征提取方法研究.机械科学与技术,2006,25(3):288～291
[9]印兴耀,周静毅.地震属性优化方法综述.石油地球物理勘探,2005,40(4):482～489
[10]Vladimir N V.Statistical learning theory.New York:John Wiley,1998,115～163
[11]肖建华,吴今培.基于核的特征提取技术及应用研究.计算机工程,2002,28(10):36～38
[12]黄国宏,邵惠鹤.核主元分析及其在人脸识别中的应用.计算机工程,2004,30(13):13～14
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心