【摘要】 自适应最小平方网格算法(简称 LSL 法)具有比梯度算法更快的收敛速度和参数变化跟踪能力。本文概述了 LSL 滤波器的主要优点和基本原理,并将其用于自适应地震记录反褶积。对理论合成和实际地震记录分别采用 Levinson法、Burg 法及 LSL 法作反褶积,其结果表明,在子波时变情况下,LSL 法反褶积结果优于其它方法。LSL 法的长处在于其时间递推公式中多了一个似然变量参数。此变量较小,意味着地震记录中有效信息较少,反之,表示有效信息较多。似然变量的一阶差分,可以用于检测来自其它概率分布的数据点的位置,改善反褶积的效果。更多还原

【Abstract】 Adaptive least-squares lattice algorithm,abbreviated to LSL algo- rithm,has faster convergence rate and better parameter tracing ability than gradient algorithm does.The paper gives a summarization of ma- jor merits and principle of the LSL filter,which has been applied to the seismic adaptive deconvolution.The deconvolutions of synthetic and field seismic data were respectively made using Levinson method, Burg method and LSL method.The comparision among the deconvolu- tion results shows that in the case of time-variant wavelet,LSL de- convolution result is better than the deconvolution results of other methods.The advantage of LSL method rises from the likelihood variable existing in time recurrence formula.A low likelihood variable indi- cates less useful informations in seismic records,and a high variable more useful informations.The first order difference of likelihood va- riable may be used to locate data points in other probability distribu- tions and to improve deconvolution result then.更多还原