基于高阶累积量拟合及混合蚁群算法的地震子波估计

Seismic wavelet estimation using high-order cumulant matching method and hybrid ant colony algorithm.

假设地层反射系数是平稳非高斯独立同分布的序列,利用地震道褶积模型的高阶累积量构造目标函数,通过优化算法来进行地震子波提取。针对该方法的非线性优化问题,利用遗传算法的快速性、随机性、全局收敛性,再结合蚁群算法的并行性、正反馈机制以及求解效率高等特性,来提高求解效率。通过对合成地震记录进行子波估计的理论模型试验,结果表明,该方法对高斯有色噪声有很好的压制效果,且能保留地震记录的相位信息,使子波提取具有较高的精度。将该方法运用于实际地震资料处理,利用估计出的较精确的地震子波来进行反褶积,能有效地拓宽地震剖面的频谱,使其信噪比和分辨率都有所提高。理论模型分析与实例计算验证了该方法的可行性。

Assuming that reflectivity series is stationary,non-Gaussian,and statistically independent,this paper proposes to estimate seismic wavelet by minimizing the nonlinear objective function constructed from high-order cumulant of the seismic convolution model.We used a hybrid ant colony algorithm to solve the problem.Tests on synthetic seismic data show that the proposed method effectively suppresses Gaussian colored noises and keeps phase information of the seismic records,which makes estimated wavelet more accurate.Deconvolution with the estimated wavelet on real data yielded satisfactory results.The spectrum of the resulted data is widened,and signal-to-noise ratio and resolution are improved.