利用多道相似组稀疏表示方法压制随机噪声
发布时间:2018-12-19 13:11
【摘要】:针对目前地震数据随机噪声压制方法采用的稀疏表示方法中单一的正交基无法根据地震数据特征自适应调整基函数,基于数据分块的自适应超完备学习字典方法通常忽略块之间相似性的问题,提出基于多道相似组稀疏表示模型的去噪算法。由于地震记录波形在邻近记录道存在较强的相似性,首先在训练窗口内计算与目标地震数据块所包含的多道记录的波形相似度,利用相似度最高的一组数据块构造多道相似组;然后采用自适应超完备字典学习算法完成基于多道相似组的字典构建与稀疏编码;最后通过迭代阈值收缩算法求解L1范数最小优化问题,逐步提高编码系数的稀疏程度,保留地震数据主要特征,压制随机噪声。与现有随机噪声压制算法对比,本文算法具有更高的峰值信噪比(PSNR),并且能更好地保持复杂区域地震数据同相轴的局部特征。
[Abstract]:According to the sparse representation method used in random noise suppression method of seismic data at present, the single orthogonal basis can not adaptively adjust the basis function according to the characteristics of seismic data. An adaptive over-complete learning dictionary method based on data partitioning usually ignores the similarity between blocks and proposes a denoising algorithm based on the sparse representation model of multi-channel similarity groups. Due to the strong similarity of seismic waveform in adjacent tracks, the waveform similarity of multi-track records contained in the target seismic data block is first calculated in the training window, and a group of data blocks with the highest similarity is used to construct a multi-track similar group. Then, the self-adaptive super-complete dictionary learning algorithm is used to construct and encode the dictionary based on multi-channel similarity group. Finally, the iterative threshold shrinkage algorithm is used to solve the L1 norm minimum optimization problem, which can gradually improve the sparse degree of coding coefficients, preserve the main characteristics of seismic data, and suppress random noise. Compared with the existing stochastic noise suppression algorithms, the proposed algorithm has a higher PSNR (PSNR), and can better maintain the local characteristics of the cophase axis of seismic data in complex regions.
【作者单位】: 东北石油大学计算机与信息技术学院;东北石油大学电气信息工程学院;
【基金】:国家自然科学基金项目(61502094) 大庆市指导性科技计划项目(zd-2016-009) 东北石油大学科研培育基金项目(NEPUPY-1-22)联合资助
【分类号】:P631.4
本文编号:2386964
[Abstract]:According to the sparse representation method used in random noise suppression method of seismic data at present, the single orthogonal basis can not adaptively adjust the basis function according to the characteristics of seismic data. An adaptive over-complete learning dictionary method based on data partitioning usually ignores the similarity between blocks and proposes a denoising algorithm based on the sparse representation model of multi-channel similarity groups. Due to the strong similarity of seismic waveform in adjacent tracks, the waveform similarity of multi-track records contained in the target seismic data block is first calculated in the training window, and a group of data blocks with the highest similarity is used to construct a multi-track similar group. Then, the self-adaptive super-complete dictionary learning algorithm is used to construct and encode the dictionary based on multi-channel similarity group. Finally, the iterative threshold shrinkage algorithm is used to solve the L1 norm minimum optimization problem, which can gradually improve the sparse degree of coding coefficients, preserve the main characteristics of seismic data, and suppress random noise. Compared with the existing stochastic noise suppression algorithms, the proposed algorithm has a higher PSNR (PSNR), and can better maintain the local characteristics of the cophase axis of seismic data in complex regions.
【作者单位】: 东北石油大学计算机与信息技术学院;东北石油大学电气信息工程学院;
【基金】:国家自然科学基金项目(61502094) 大庆市指导性科技计划项目(zd-2016-009) 东北石油大学科研培育基金项目(NEPUPY-1-22)联合资助
【分类号】:P631.4
【相似文献】
相关博士学位论文 前1条
1 李海山;基于稀疏表示理论的地震信号处理方法研究[D];中国石油大学(华东);2013年
相关硕士学位论文 前2条
1 范弘毅;基于稀疏表示和压缩感知的地震信号处理及应用研究[D];电子科技大学;2014年
2 许德鑫;基于稀疏表示和字典学习的地震数据去噪研究[D];吉林大学;2016年
,本文编号:2386964
本文链接:https://www.wllwen.com/kejilunwen/diqiudizhi/2386964.html