Seismic Signal Reconstruction Method Based on Hankel Tensor Decomposition

Abstract

The invention discloses a seismic signal reconstruction method based on Hankel tensor decomposition, which includes constructing a Hankelized objective function, converting the Hankelized objective function into a matrix completion objective function, and adopting an alternate minimization method to alternately solve. The present invention effectively combines the Hankel structure and low-rank tensor by introducing low-rank tensor decomposition, and solves the problem that only relying on tensor low-rank decomposition for regularization requires too high low-rank data; at the same time, By means of low-rank tensor decomposition, the process of singular value decomposition of tensor is avoided, and the solution speed is greatly improved.

Description

technical field [0001] The technical field of seismic data processing of the present invention particularly relates to a seismic signal reconstruction method based on Hankel tensor decomposition. Background technique [0002] With entering the 21st century, the global economy has entered a new period of development, and the demand for energy from all walks of life has become increasingly strong. However, new energy has not been able to replace the core position of oil and natural gas in the energy system. With the rapid development of the global economy, all countries are heavily dependent on oil and natural gas, and their demand is increasing. The domestic consumption of oil and natural gas, a resource that is related to people's livelihood, has reached hundreds of millions of tons. [0003] Although our country has a large land area, compared with the Middle East, our country is not on the oil belt. Only some parts of Tibet are on the tail of the global rocky belt. Insuff...

AI Technical Summary

Problems solved by technology

[0003] Although our country has a large land area, compared with the Middle East, our country is not on the oil belt. Only some parts of Tibet are on the tail of the global rocky belt. Insufficient gas: With continuous exploitation, the rock reserves of Daqing Oilfield, which we are proud of, can only be used for decades

However, due to the increase in the dimension of high-dimensional data, the general low-rank performance will be worse than that of two-dimensional data

If the low rank of the data itself is not very good, the recovery accuracy of TNN will also be greatly affected, and because the core of TNN is based on the tensor singular value decomposition algorithm, this leads to a very high computational cost of the algorithm

Images