【摘要】 本文应用近年来出现的自适应LSL算法(Adaptive least squares Lattice al gorithm)进行地震信号预测反褶积,求取地层反射系数,进而由地层反射系数抽取特征向量,应用统计模式识别技术辅助沉积相的推断。文中,提出了一种用一维聚类和分类抽取地层反射系数的方法,并且给出了两种具体实现方式。计算机实验结果表明,本文方法是有效的。更多还原

【Abstract】 Because of its many advantages, the adaptive least square lattice (LSL for short)algorithm, developed in recent years, has been Widely used in time varying signal processing. In this paper, the LSL algorithm in AR model is applied to the predictive deconvolution of seismic signals for obtaining the reflection coefficient sequences of underground wavelets, it is shown by experiments that LSL algorithm is superior to commonly used Levision algorithm and Burg algorithm. After deconvolution, the reflection coefficient sequences are extracted into feature vectors which, in turn. are classified by statistical pattern recognition technique to aid the inference of the underground sedimentary faces. Due to the special characteristics of reflection coefficient sequences the common used feature extraction methods in pattern recognition literature are not suitable. In this paper, through analysing the special characteristics of reflection coeffici ent sequences, a feature extraction method by one dimensional clustering and classification is proposed, and two practical ways for realizing the method are also given. The results of computer experiments on seismic signals have shown that the method is effective.更多还原