Research On Seismic Data Compression Algorithm Based On Distributed Principal Component Analysis

Seismic exploration is an effective technical method in geophysical exploration.With the increasing demand for oil,gas and mineral resources,the scale of seismic exploration keeps expanding,and the amount of seismic data also increases dramatically.However,how to efficiently transmit and store massive seismic data is an urgent engineering problem to be solved.Data compression is an important technical means to solve the above problem.This article takes the current situation of data compression research at home and abroad as a starting point,and uses qualitative and quantitative evaluation methods to compare the Fourier transform,discrete cosine transform,wavelet transform,and principal component analysis compression transform methods.Among them,the principal component analysis method not only completely removes the correlation between variables,but also makes the energy more concentrated after the transformation.More importantly,the mean square error produced by compression is the smallest.These are its advantage in the field of seismic data compression.However,this method only removes the local correlation,but does not consider the global correlation,so the compression ratio needs to be improved.Of course,if the compression ratio is pursued blindly,reconstructed data will not be available.In order to solve these problems,this article designed a seismic data compression algorithm based on distributed principal component analysis for the network topology of non-cable seismic exploration system.The main research contents are as follows:1)Distributed principal component analysis compression algorithm.The seismograph calculates the local covariance matrix and local mean vector of each channel data,then calculates the relevant cumulants,and transmits these cumulants to the central control.Then the central control calculates the global covariance matrix and the global mean vector,and then uses the eigenvalue decomposition and cumulative contribution rate to calculate the global eigenvector matrix,and transmits the global eigenvector matrix along with the global mean vector to all seismographs.Finally,each seismograph projects the centralized original data on the global eigenvector matrix.In the distributed principal component analysis compression algorithm,the cumulative contribution rate determines the waveform fidelity.If the value of the cumulative contribution rate is not appropriate,the reconstructed data may not be able to achieve high waveform fidelity.2)Fidelity-restricted distributed principal component analysis compression algorithm.The central control calculates the cumulative contribution rate based on data from the first shot according to a given waveform fidelity,and take this as the cumulative contribution rate of multi-shot data.Then the central control calculates the global mean vector and global covariance matrix of the multi-shot data according to the relevant cumulative from the seismograph,and then obtains the global eigenvalue and global eigenvector through eigenvalue decomposition,and then determines the retained eigenvectors based on the cumulative contribution rate,and transmits them along with the global mean vector to each seismograph.Finally,each seismograph projects the centralized original data into a global eigenvector matrix composed of retained eigenvectors,and guarantees the desired waveform fidelity.3)According to the simulation results,the effectiveness of the fidelity-restricted distributed principal component analysis seismic data compression algorithm is analyzed,and the influence of the original data signal-to-noise ratio and sampling rate on the compression quality is discussed.Field experiments verify the effectiveness of the algorithm.The proposed seismic data compression method based on distributed principal component analysis can not only meet the compression requirement of data collected by non-cable seismic exploration system,and make seismic data transmission and storage more efficient,but also provide a new method for seismic data compression.