基于改进有监督局部线性嵌入的地震属性降维优化

摘要

为克服直接从地震资料中获取的地震属性具有维数高、数据量大、信息冗余、高度非线性,且数据样本缺乏先验知识分类等缺陷,用基于等价关系的模糊聚类方法对有监督局部线性嵌入算法进行改进,将其应用于地震属性的非线性降维优化,实例表明其地震属性降维优化效果优于常规线性降维方法,为地震资料解释处理、储层预测及物性参数反演提供了良好可靠的数据资料.%In order to overcome the defects of the seismic attribute which is extracted directly from seismic data, such as high dimension, enormous data, redundant information, high nonlinearity and sample data's lack of priori knowledge classification, the fuzzy cluster method based on equivalence relation is used to supervise local linear embedding. The cases of study indicate that the result of nonlinear reducing dimension and optimization of seismic attribute based on the developed SLLE is better than that of the other conventional linear reduce dimension methods. The algorithm can provide good and reliable data for seismic data interpretation, reservoir prediction and material parameters inversion.