基于奇异值分解的时变子波提取准确性评价方法

子波提取准确性的评价在地震数据处理中占有重要地位，但是传统的评价准则受噪声影响较大。为此，提出一种基于奇异值分解(SVD)的时变子波提取准确性评价方法，考虑非平稳地震记录的子波提取准确性评价方法中Parsimony准则、丰度准则和绝对峰度准则对噪声环境的承受能力较强，选用Parsimony准则与奇异值分解技术结合，构造了一种抗噪/容噪能力更强的时变子波提取准确性评价准则SVD_P。将SVD_P准则、Parsimony准则和丰度准则同时应用于仿真数据和实际资料处理，对比分析了时频域时变子波提取方法与自适应分段时变子波提取方法的准确性，结果表明：SVD_P准则、Parsimony准则和丰度准则都能对两类子波提取方法进行正确的评价，时频域子波提取法提取子波的准确性高于自适应分段法提取的子波，但是SVD_P准则评价的结果相对误差最小，评价精度最高。

The accuracy evaluation of time-varying wavelet extraction plays an important role in seismic data processing.However,the conditional evaluation criterion is influenced seriously by noise.Therefore,we propose a time-varying wavelet accuracy criterion based on singular value decomposition (SVD).Since the Parsimony criterion,Kurtosis criterion and Absolute kurtosis criterion have good tolerability to noisy environment among the existing evaluation criteria for the non-stationary seismic wavelet extraction accuracy,the Parsimony criterion and SVD technology are combined to construct a SVD_P criterion which has better noise-tolerant ability; and the spectrum division is employed as the deconvolution method.The Parsimony criterion,Kurtosis criterion and SVD_P criterion are applied to the simulation experiment and field data processing to compare the precision of time-frequency domain time-varying wavelet extraction method and adaptive segmentation time-varying wavelet extraction method.The results show that all three criteria could provide valid evaluation of these two wavelet extraction method while the time-frequency domain wavelet extraction method is more accurate than the adaptive segmentation method.Additionally,the evaluation result of SVD_P criterion owns smallest error and highest evaluation precision.

国家自然科学基金项目(40974072)和中国石油大学(华东)研究生创新工程资助项目(YCX2015050)联合资助。