基于时空约束压缩感知的地震数据重建

Seismic data reconstruction based on space-time constraint compressed sensing

在实际勘探中,由于环境、设备或人为因素的影响,采集的地震数据中有很多丢失的数据,严重影响了数据的解释工作。针对这一问题,根据地震数据的时空相关性,提出了一种基于时空约束压缩感知的地震数据重建方法。该方法使用内核奇异值分解(KSVD)字典学习算法训练超完备字典作为稀疏变换基,进而利用改进的稀疏自适应匹配追踪算法(SAMP)完成重建。通过初始稀疏性估计和变步长策略,减少了SAMP中收敛所需的迭代次数。利用真实的地震数据和微电阻率成像数据进行实验,将所提出的方法与压缩感知重建算法进行了比较,不仅提高了重建数据的准确性,而且缩短了执行时间。

In actual exploration, due to the influence of environment, equipment or human factors, there are a lot of missing data in the seismic data collected, which seriously affects the data interpretation work. Aiming at this problem, according to the space-time correlation of seismic data, a method of seismic data reconstruction based on space-time constrained compressed sensing is proposed. In this method, an over-complete dictionary as a sparse transform basis is trained using kernel singular value decomposition(K-SVD) dictionary learning algorithm. The reconstruction is accomplished using an improved sparsity of adaptive matching pursuit(SAMP). By incorporating an initial sparsity estimation step and adopting a variable step size strategy, the number of iterations needed for convergence in SAMP can be significantly reduced. Using real seismic data and micro-resistivity imaging data, the proposed novel method is compared with state-of-the-art compressive sensing reconstruction algorithms. The experimental results show that the accuracy of the reconstructed data is significantly improved, and the execution time is also reduced.