Diffraction Extraction Using a Low-Rank Matrix Approximation Method

Peng Lin

doi:10.1109/LGRS.2022.3187045

Diffraction Extraction Using a Low-Rank Matrix Approximation Method

2022-01-01

3

PRE

AI

DOI：10.1109/LGRS.2022.3187045

Original

Original request for help

Save

Abstract

En 中文

Subsurface discontinuous features, such as faults, cavities, and pinch-outs, are commonly related to the spatial distribution of hydrocarbon reservoir zones and safe coal mining. Diffractions generated from subsurface discontinuities carry valuable information and thus are capable of accurately revealing these geological structures. Because diffractions behave as weak amplitudes, they are easily covered by specular reflections. The low-rank method performs well for diffraction extraction from specular reflections. However, the separation results are sensitive to the noise levels in traditional rank-reduction methods. The higher the noise level, the weaker the low-rank operator; therefore, the presence of noise affects the subsequent imaging. To improve the separation quality, an improved diffraction-separation method is proposed that uses parameterized nonconvex penalty functions in terms of the low-rank assumption of seismic records. This new algorithm considers arctangent penalty functions as regularization terms for an accurate low-rank matrix approximation. The proposed algorithm is used to extract diffracted energy and eliminate reflected energy from noisy data. We use applications to synthetic and field examples to demonstrate that the new algorithm is able to extract high-quality diffractions, which helps locate and reveal subsurface geological discontinuities.

Keywords:

Diffraction

Reflection

Geology

Matrix decomposition

Data mining

Noise measurement

Imaging

Diffraction extraction

low rank

multichannel singular-spectrum analysis (MSSA)

small-scale discontinuities

Diffraction Extraction Using a Low-Rank Matrix Approximation Method

Abstract

Journal

Organization