利用改进的在线字典学习估计时变子波

发布时间：2018-07-11 21:13

本文选题：在线字典学习 + 稀疏表示　；参考：《石油地球物理勘探》2016年05期

【摘要】：为了获得符合实际的混合相位子波,提出了一种基于在线字典学习的时变子波估计方法。将时变子波估计转化为在线字典学习问题,通过过完备字典的在线学习实现冗余字典的自适应更新。字典中的每个原子代表子波的一个分量,通过原子的线性组合实现对时变子波的有效逼近。在线字典学习可以灵活地利用训练数据,改进字典中的原子,提升字典的自适应特性。同时,根据地震数据的特点,对训练数据与稀疏表示的残差项进行滤波处理,改进了在线字典学习方法,降低了对噪声的敏感性。无噪声和含噪声合成数据的实验结果证明了本文方法的有效性,而且对噪声具有一定的鲁棒性。实际子波估计结果以及Wiener滤波的反褶积剖面和频谱分析表明,本文方法得到的结果并未使噪声能量增强,但频带得到拓宽,从而为时变子波估计提供了新思路。
[Abstract]:In order to obtain practical mixed phase wavelet, a method for estimating time-varying wavelet based on online dictionary learning is proposed. The time-varying wavelet estimation is transformed into an online dictionary learning problem, and the redundant dictionary can be updated adaptively through the on-line learning of the over-complete dictionary. Each atom in the dictionary represents a component of the wavelet, and the effective approximation of the time-varying wavelet is realized by the linear combination of atoms. Online dictionary learning can make use of training data flexibly, improve the atom in dictionary, and enhance the adaptability of dictionary. At the same time, according to the characteristics of seismic data, the residual terms of training data and sparse representation are filtered, the online dictionary learning method is improved, and the sensitivity to noise is reduced. The experimental results of noise-free and noise-free synthetic data show that the proposed method is effective and robust to noise. The actual wavelet estimation results and the deconvolution profile and spectrum analysis of Wiener filter show that the proposed method does not increase the noise energy but broadens the frequency band, thus providing a new idea for the estimation of time-varying wavelet.
【作者单位】：电子科技大学光电信息学院;电子科技大学信息地学研究中心;
【基金】：国家自然科学基金(41274127,40874066)资助
【分类号】：P631.4

【相似文献】