Magnetotelluric Noise Suppression Based On Sparse Decomposition And Compressive Sensing Reconstruction Algorithm

Due to the weakness and large bandwidth of the signal of the natural magnetotelluric,it is highly susceptible to various noise pollution.With the progress of human civilization,cultural electromagnetic interference caused by all kinds of factors is becoming more and more serious,resulting the magnetotelluric data collected in the field showed a typical near-source effect.Therefore,how to eliminate the strong interference in magnetotelluric data,and improve the quality of MT data have been an important research topic in the field of MT.In this paper,we introduce sparse decomposition and reconstruction algorithm of compressive sensing,which are respectively for suppress the magnetotelluric noise.The main research contents are as follows:(1)Sparse decomposition algorithm has the advantages of fewer restrictions,strong adaptability,flexible and efficient,which provides a new method for the suppression of magnetotelluric interference.MP is the most commonly used sparse decomposition method and the essence is a greedy algorithm.In order to solve the problem that matching pursuit has large amount of calculation and magnetotelluric data processing is low efficiency,a new method for magnetotelluric noise suppression based on matching pursuit and genetic algorithm is proposed in this paper.First of all,we use Gabor atom to build a over-complete dictionary,and divide the over-complete dictionary collection.Then,we use the adaptability of genetic algorithm to search for the optimal matching atom and its location.Finally,we use the optimal matching atom to sparse decompose the signal to be processed and reconstruct the useful signal.Through analysis and processing for the typical strong interference simulated by computer and the measured magnetotelluric data of ore concentration area,we get some important conclusions as follows.Compared with matching pursuit and orthogonal matching pursuit,the method can quickly and effectively removed the large-scale strong interference of magnetotelluric data in time domain sequence.Moreover,the quality of magnetotelluric data for the low frequency band has been improved significantly.(2)Compressive sensing is a new and effective signal sampling theory.Moreover,compressive sensing reconstruction of signal is one of the most important part of the compressed sensing theory.In order to quickly and effectively suppress the large-scale strong interference in ore concentration area,and retain the useful information of low frequency band of magnetotelluric data.First of all,according to the type of magnetotelluric noise,we choose the redundant dictionary of many kinds of atomic structures such as wavelet packets and cosines to sparse decomposition.Then,we use the Gaussian matrix observation sampling.Finally,we use stagewise orthogonal matching pursuit(StOMP)to reconstruct useful signals,so as to achieve the purpose of fast de-noising.The proposed method provides a new research idea for carrying out massive measured magnetotelluric data processing in the future.Therefore,it has a broad application prospect.