中文摘要：

地震记录的初至拾取是地震资料处理的基础工作,特别是静校正、层析成像等后续处理对初至拾取的精度要求较高。当初至波到达时,其信号的振幅、频率、相位均会发生变化,基于此,提出了一种基于时频分析的初至拾取算法,该算法在噪声判别系数的约束下,利用小波变换时频分析提取信号的瞬时相位谱,然后利用瞬时相位谱的过零点属性拾取初至。为了提高计算效率,在给定时窗内部计算信号的瞬时相位谱,并且利用多个不同的噪声判别系数初至拾取的统计结果确定最终的初至时间。理论信号和实际地震资料处理结果表明,该方法具有较好的抗噪能力,初至拾取的精度比较高。

英文摘要：

The first-break picking is a basic work of seismic data processing, especially the static correction and tomography imaging requires accurate first-break time. When the primary wave arrives, the amplitude, frequency and phase of the signals will change. Based on the above recognition, a new first-break picking algorithm is put forward based on time-frequency anal ysis. Under the constraint of noise discrimination coefficient, time-frequency analysis is conducted by wavelet transform to ex- tract the instantaneous phase spectrum of signals, and then pick up the first-break using the zero-crossing point attribute of instantaneous phase spectrum. In order to improve the computational efficiency, the instantaneous phase spectrum of signals is calculated within a pre-defined time window and the statistical results of several first-break results of different noise dis- crimination coefficients are used to determine the final first-break time. Theory signals and real seismic data examples show that the approach owns higher noise-resistance and precision.