项目名称: 基于压缩感知的地震数据重建理论研究
项目编号: No.41304097
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 天文学、地球科学
项目作者: 张华
作者单位: 东华理工大学
项目金额: 25万元
中文摘要: 传统地震数据采集必须遵循奈奎斯特采样定理,后续的处理也需要较密且完整的地震数据,但野外数据采集可能由于地震道缺失或者勘探成本限制,不一定能满足采样定理要求,因此存在数据重建的问题。本项目基于压缩感知理论,突破传统采样定理的限制,利用随机欠采样方法将传统规则欠采样所带来的互相干假频转化成较低幅度的不相干噪声,从而将数据重建问题转为更简单的去噪问题。数据重建过程采用曲波变换作为稀疏表示基,在L1范数约束下,使用稀疏促进反演策略,形成基于曲波变换的二维和三维地震数据重建理论,在此基础上,引入非均匀快速傅立叶变换,实现基于非均匀曲波变换的二维和三维地震数据重建理论与方法。同时针对单纯随机采样的不足,发展其他采样方式,在保持随机采样优点的同时能够灵活调整,控制采样间隔,进一步提高数据重建质量。该项研究对于指导复杂地区数据采集、缺失地震道重建及压缩数据采集量等方面具有重要的理论意义和实用价值。
中文关键词: 压缩感知;数据重建;非均匀;凸集投影;曲波变换
英文摘要: Traditional seismic data acquisition must follow the Nyquist sampling theorem, and the subsequent processing need more densely and integrated seismic data, but field data acquisition perhaps cann't meet the Nyquist sampling theorem as missing traces and exploration cost limit, so there exit data reconstruction problem. In this item, we introduce the compressed sensing theory, breaking through the traditional Nyquist sampling theorem, rendering coherent aliases of regular undersampling into harmless incoherent random noise using the random undersampling, effectively turning the reconstruction problem into a much simpler denoising problem.We choose curvelet transform as sparse basis during the process of reconstruction, and form a curvelet-based reconstruction theory of 2D and 3D the seismic data by sparsity-promoting inversion with the smallest L1 norm. Based on this, we introduce the nonuniform Fourier transform and achieve the a nonuniform curvelet-based reconstruction theory of 2D and 3D the seismic data by sparsity-promoting inversion. At the same time, Aiming at the deficiency of simple random undersampling, we develop other undersampling,which shares the benefits of random sampling and can be flexibly adjusted and controls the maximum gap size, and further improves the quality of data reconstruction. This r
英文关键词: Compressive sensing;Data reconstruction;Non-uniform;Projections onto convex sets;Curvelet transform