CN102305945A  Linear noise eliminating method  Google Patents
Linear noise eliminating method Download PDFInfo
 Publication number
 CN102305945A CN102305945A CN201110165506A CN201110165506A CN102305945A CN 102305945 A CN102305945 A CN 102305945A CN 201110165506 A CN201110165506 A CN 201110165506A CN 201110165506 A CN201110165506 A CN 201110165506A CN 102305945 A CN102305945 A CN 102305945A
 Authority
 CN
 China
 Prior art keywords
 line noise
 check point
 signal
 related coefficient
 offset
 Prior art date
Links
 238000001914 filtration Methods 0.000 claims abstract description 34
 238000005070 sampling Methods 0.000 claims abstract description 7
 230000001131 transforming Effects 0.000 claims abstract description 4
 238000000034 methods Methods 0.000 claims description 6
 230000000694 effects Effects 0.000 abstract description 4
 238000004458 analytical methods Methods 0.000 description 1
Abstract
Description
Technical field
The present invention relates to data processing technique, the line noise removing method in the particularly seismic analysis data.
Background technology
It is the important evidence and the assurance of seismic data interpretation, oil and gas reservoir prediction that highfidelity, high resolving power and high s/n ratio seismic data are handled.Therefore, the noise remove of prestack, stack seismic signal and decay are important steps during seismic data is handled.The noise source of seismic signal is very complicated, like sound wave, ground roll, industrial electro interference, ghosting, repeatedly reflection, repetitive shock, side wave, end ripple, reverberation and singing and various random disturbance.According to the characteristic of various noises, seismic noise can be divided into two big types: coherent noise and random noise.The original section of original single big gun seismologic record as shown in Figure 1.Transverse axis is the Taoist monastic name (TraceNum) in a corresponding road (Trace), each monitoring point among the figure, from TraceNum1 to TraceNum300; The longitudinal axis is a monitoring time, from 0 to 2350 millisecond (millisecond).Have tangible line noise to disturb among Fig. 1, these line noises show as its lineups in image be linear basically, and different with the direction (direction of useful signal is vertical) of useful signal, and shown in arrow among the figure is exactly a route property noise wherein.The method of existing elimination line noise (coherent noise a kind of) has in time domain carries out, and also has at frequency domain.Main disposal route has: excision method, fk etc.The excision method is that the zone that line noise is arranged is excised, make the data sampling value regulation the time be zero in the window.The method significant wave in the excision coherent interference has also been cut away.In addition, when seismic trace makes in the window timesampling value be zero will cause that the spectrum estimation value distorts.Fk filtering is a kind of method of removal coherent interference commonly used, and this method is to carry out filtering at frequencywavenumber domain.The excision of frequency field often causes the alias effect, makes seismologic record false lineups, significant wave wave form distortion occur, brings difficulty to explanation.Prestack timespace domain method is carried out in time domain, also is the more also effective method that goes line noise to use at present.But, singlely based on time domain or frequency domain filtering, because that useful signal and noise still are at frequency domain that time domain all has is overlapping, if eliminate noise fully, the loss of useful signal will inevitably appear.That is, all can there be the contradiction of signal to noise ratio (S/N ratio) and resolution in existing filtering method, and noise removing ground is complete more, and the useful signal loss is many more, and the resolution of seismologic record sectional view is low more.
Summary of the invention
Technical matters to be solved by this invention is, provides a kind of when guaranteeing denoising effect, useful signal do not caused the line noise removing method of damage.
The present invention solves the problems of the technologies described above the technical scheme that is adopted to be, a kind of line noise removing method may further comprise the steps:
(1) in sensing range; Choose a check point successively, calculate the pairing related coefficient of each direction except that vertical direction, get wherein maximal value as the related coefficient of this check point through this check point; Whether the related coefficient of judging this check point is more than or equal to preset maximum value; In this way, represent that then there is line noise in the pairing direction of this related coefficient, get into step (3); As not, then get into step (2);
(2) judge in sensing range whether also have check point, in this way, return step (1), as denying the denoising completion;
(3) institute samples a little on this line noise direction of confirming; Adopt Fourier transform that sampled data is transformed to score field earlier based on fractional order; The sampled data that is converted into score field is being carried out predictive filtering; With the useful signal of line noise, obtain the score field signal value of line noise after the predictive filtering as predictive filtering;
(4) with the score field signal value inverse transformation of line noise to time domain, obtain the timedomain signal value of line noise;
(5) from the section signal, deduct the timedomain signal value of line noise, return step (1).
The present invention at first will utilize certain road in the section to confirm the direction of the line noise of certain some coherent signal, samples in this direction then.For the signal after the sampling; Line noise just can be regarded as useful signal; And useful signal is just regarded random noise, can utilize the score field predictive filtering to dope line noise like this, deducts the line noise that dopes with the original section signal at last and has just obtained the result after the denoising.Fourier transform (FrFT based on fractional order; Fractional FourierTransform) as a kind of generalized form of Fourier (Fourier) conversion: the expression of signal on the Fractional Fourier territory; Merged the information of signal simultaneously at time domain and frequency domain; Can combine timedomain and frequencydomain that signal is divided, distinguish useful signal and line noise effectively.Therefore, the line noise that the present invention is based on score field filtering is eliminated, and is different from existing singlely based on time domain or frequency domain filtering, can solve the contradiction of signal to noise ratio (S/N ratio) and seismic resolution preferably.
Concrete, the method for calculating the pairing related coefficient of direction of this check point of process is:
Cof＝(sum1*sum1)/(sum2*ncc)
Wherein, Cof representes the pairing related coefficient of current detection direction through check point; Sum1 is the sampled value sum of each point on the current detection direction; Sum2 be each point on the current detection direction sampled value square with; Ncc is that current detection direction upsampling value is not the number of 0 point.
Further; In order to improve the fidelity of useful signal; In the step (1) after the related coefficient of judging check point is less than preset maximum value; The further size of the related coefficient of this check point and predetermined minimum relatively is less than or equal to predetermined minimum like the related coefficient of check point, then gets into step (2); Like the related coefficient of check point greater than predetermined minimum; And less than preset maximum value; Then get into step (3); And after step (3) obtains the signal value of line noise, with calculating the signal value that weights come the line noise that weighting obtains, and upgrade the signal value of line noise with the signal value after the weighting;
Said weights are: (related coefficientpredetermined minimum of check point)/(preset maximum valuepredetermined minimum).
The invention has the beneficial effects as follows, can well remove line noise, can guarantee highfidelity, high resolving power and the high s/n ratio requirement in the seismic signal interpretation process simultaneously again.
Description of drawings
Fig. 1 is original single big gun seismologic record;
Fig. 2 is a line noise section of using the present invention's prediction;
Fig. 3 is the predictive filter synoptic diagram of embodiment;
Fig. 4 uses the filtered that the present invention obtains.
Embodiment
Elimination at the enterprising line linearity noise of original section of single big gun seismologic record as shown in Figure 1 is handled, and may further comprise the steps:
(1) confirm the sensing range in the section through two given time parameter T (n/2), T (n), T (0), wherein T (n/2) is the zerotime of seismic section interim orbit, and T (n), T (0) are respectively the zerotimes in road, section the right, road, the left side.Initial detection time is confirmed that through the zerotime T (n/2) of section interim orbit and the zerotime T (1) or the T (n) of section bypass 1≤x≤n, n are total road several 300 of Fig. 1 midship section in the x road in the section:
When 0≤x≤150, the x roads initial detection time of T (x) are:
T(n/2)+[T(1)T(150)]*[offset(x)offset(150)]/[offset(1)offset(150)]；
When 150＜x≤300, the x roads initial detection time of T (x) are:
T(150)+[T(300)T(150)]*[offset(x)offset(150)]/[offset(300)offset(150)]；
Wherein, offset (x) is the offset distance in x road.Offset (x) can directly read from the file of seismologic record;
Filter range just is limited at the point of T (1) to the line of T (150) and T (150) to the part below the line of T (300) like this;
Certainly, as not considering to reduce operand, also can each point in the section be carried out denoising Processing as check point;
Each point in the section is by (TraceNum MilliSecond) confirms;
(2) in sensing range, choose a check point successively, calculate the pairing related coefficient of each direction except that vertical direction through this check point, the computing formula of related coefficient is following:
Cof＝(sum1*sum1)/(sum2*ncc)
Sum1 is the sampled value sum of each point on the current detection direction; Sum2 be each point on the current detection direction sampled value square with; Ncc is that current detection direction upsampling value is not the number of 0 point, and can be known by the computing formula of related coefficient: cof is greater than 0 and less than 1;
Got this check point wherein maximal value as the related coefficient of this check point.Set two threshold value: predetermined minimum mincof and preset maximum value maxcof;
When the related coefficient of check point is less than or equal to mincof, then think not to be no linear noise on this check point, do not carry out filtering, return step (2);
When the related coefficient of check point is just carried out Filtering Processing to it more than or equal to maxcof, promptly next step;
When the related coefficient of check point greater than mincof less than maxcof, equally it is carried out filtering, but its result to multiply by a weights c, the c=(related coefficient of check pointmincof)/(maxcofmincof);
(3) confirmed the line noise direction after, on this line noise direction the institute sample a little, sampled data is converted into score field through FrFT; Have the characteristics of strong correlation according to line noise, filtering factor (F0, F1 in the score field predictive filter are set; Fn), n+1 is the number of filtering factor, rule of thumb; Generally between 4 to 10, the setting principle of filtering factor is the n value: the square error between filtration module output valve and the expectation value is minimum;
Predictive filter is as shown in Figure 2; Comprise input end, expectation value memory module, output terminal, filtration module, wherein filtration module comprises n delayer, a n+1 multiplier, 1 totalizer; Each multiplier stores a filtering factor; Data after data before postponing and the delay input to respectively in the multiplier carries out computing, and the output result of all multipliers carries out additive operation in totalizer, and the result of totalizer output is filtered; It is socalled that filter filtering is actual exactly filtering factor is carried out convolution with the corresponding data of treating denoising;
The input end of predictive filter is used for, and receives the score field sampled data of input;
The expectation value memory module of predictive filter is used for, with storing as expectation value in the value of score field through analyzing line noise in advance;
Filtration module is used for, utilize filtering factor (F0, F1 ... Fn) data of input are carried out filtering;
The output terminal of predictive filter is used for, and exports filtered sampled data;
In the FrFT conversion process, exponent number p can according to the Generalized Time bandwidth product (Generalized TimeBandwidth Product, GTBP) minimum principle is confirmed, the computing formula of GTBP is: So just can in oneperiod, search out the optimal mapping exponent number of the corresponding exponent number of minimum GTBP as us: Wherein, x _{p}(t) be p rank Fourier Transform of Fractional Order, For score field the time wide, Be the bandwidth of score field, a, t are temporary variable.
(4) will be converted into the sampled data input predictive filter of score field, predictive filter is the useful signal of line noise as predictive filtering, and the useful signal of reality is regarded as random noise, obtains the score field signal value of filtered result for line noise; Like the related coefficient of check point in the step (3) greater than mincof less than maxcof, also filtered result is carried out weighting with weights c, upgrade the score field signal value of line noise with the result after the weighting;
(5) with the score field signal value inverse transformation of line noise to time domain;
(6) from original section, deduct the timedomain signal value of the line noise that finally obtains in the step (5); Be the net result of having eliminated line noise; Return step (2) more next check point is carried out the detection of filtering next time, all check points in sensing range dispose.
Fig. 3 is for obtaining the line noise section after carrying out predictive filtering on the original seismic section shown in Figure 1; The section that Fig. 4 obtains for the line noise that deducts prediction with original seismic section.As can be seen from Figure 4, the inventive method can well be removed line noise, can guarantee " three height " requirement in the seismic signal interpretation process simultaneously again.The present invention can solve the contradiction of signal to noise ratio (S/N ratio) and seismic resolution, obtains highfidelity, high resolving power and high s/n ratio seismic data, and then the important evidence and the assurance of seismic data interpretation, oil and gas reservoir prediction are provided.
Claims (6)
Priority Applications (1)
Application Number  Priority Date  Filing Date  Title 

CN 201110165506 CN102305945B (en)  20110620  20110620  Linear noise eliminating method 
Applications Claiming Priority (1)
Application Number  Priority Date  Filing Date  Title 

CN 201110165506 CN102305945B (en)  20110620  20110620  Linear noise eliminating method 
Publications (2)
Publication Number  Publication Date 

CN102305945A true CN102305945A (en)  20120104 
CN102305945B CN102305945B (en)  20130925 
Family
ID=45379821
Family Applications (1)
Application Number  Title  Priority Date  Filing Date 

CN 201110165506 CN102305945B (en)  20110620  20110620  Linear noise eliminating method 
Country Status (1)
Country  Link 

CN (1)  CN102305945B (en) 
Cited By (3)
Publication number  Priority date  Publication date  Assignee  Title 

CN103616719A (en) *  20131205  20140305  中国矿业大学(北京)  Microseism acquisition device and method with noise identification and selfadaptive amplification functions 
CN105843779A (en) *  20160607  20160810  华中科技大学  Realtime noise elimination method oriented to POTDR opposite scattered light signals 
CN108957530A (en) *  20180523  20181207  电子科技大学  A kind of crack automatic testing method based on Acceleration Algorithm in Seismic Coherence Cube slice 
Families Citing this family (2)
Publication number  Priority date  Publication date  Assignee  Title 

CN106443787B (en) *  20150812  20180904  中国石油天然气集团公司  Prestack seismic gather presses method for denoising and its device 
CN106019377B (en) *  20160511  20180112  吉林大学  A kind of twodimensional seismic survey noise remove method based on timespace domain frequency reducing model 
Citations (2)
Publication number  Priority date  Publication date  Assignee  Title 

US5521881A (en) *  19940902  19960528  Exxon Production Research Company  Method of processing seismic data having multiple reflection noise 
CN101655834A (en) *  20090917  20100224  哈尔滨工业大学  Signal separation method based on fractional wavelet transform 

2011
 20110620 CN CN 201110165506 patent/CN102305945B/en not_active IP Right Cessation
Patent Citations (2)
Publication number  Priority date  Publication date  Assignee  Title 

US5521881A (en) *  19940902  19960528  Exxon Production Research Company  Method of processing seismic data having multiple reflection noise 
CN101655834A (en) *  20090917  20100224  哈尔滨工业大学  Signal separation method based on fractional wavelet transform 
NonPatent Citations (1)
Title 

朱全银等: "基于分数阶傅里叶变换的线性调频干扰抑制", 《探测与控制学报》 * 
Cited By (4)
Publication number  Priority date  Publication date  Assignee  Title 

CN103616719A (en) *  20131205  20140305  中国矿业大学(北京)  Microseism acquisition device and method with noise identification and selfadaptive amplification functions 
CN105843779A (en) *  20160607  20160810  华中科技大学  Realtime noise elimination method oriented to POTDR opposite scattered light signals 
CN108957530A (en) *  20180523  20181207  电子科技大学  A kind of crack automatic testing method based on Acceleration Algorithm in Seismic Coherence Cube slice 
CN108957530B (en) *  20180523  20190823  电子科技大学  A kind of crack automatic testing method based on Acceleration Algorithm in Seismic Coherence Cube slice 
Also Published As
Publication number  Publication date 

CN102305945B (en)  20130925 
Similar Documents
Publication  Publication Date  Title 

Randall  A history of cepstrum analysis and its application to mechanical problems  
Chen  Dipseparated structural filtering using seislet transform and adaptive empirical mode decomposition based dip filter  
Wang et al.  The unimportance of phase in speech enhancement  
Kepesi et al.  Adaptive chirpbased time–frequency analysis of speech signals  
Vold et al.  High resolution order tracking at extreme slew rates, using Kalman tracking filters  
CN101647061B (en)  Noise variance estimator for speech enhancement  
JP5227393B2 (en)  Reverberation apparatus, dereverberation method, dereverberation program, and recording medium  
CN102832908B (en)  Wavelet transform and variablestepsize LMS (least mean square) adaptive filtering based signal denoising method  
Liu et al.  A 1D timevarying median filter for seismic random, spikelike noise elimination  
KR100580428B1 (en)  A compensation method for the distorted secondary current of a current transformer  
JP5714180B2 (en)  Detecting parametric audio coding schemes  
KR101184992B1 (en)  Improved coding techniques using estimated spectral magnitude and phase derived from mdct coefficients  
CN103238181B (en)  Method for restoring spectral components attenuated in test denoised speech signal as a result of denoising test speech signal  
Olhede et al.  A generalized demodulation approach to timefrequency projections for multicomponent signals  
US8073634B2 (en)  Method to extract target signals of a known type from raw data containing an unknown number of target signals, interference, and noise  
CN102116868B (en)  Seismic wave decomposition method  
EP1921609A1 (en)  Noise suppressing method and apparatus and computer program  
KR20190005217A (en)  Frequencybased audio analysis using neural networks  
Liu et al.  Seismic data interpolation beyond aliasing using regularized nonstationary autoregression  
Djurovic et al.  Robust Lestimation based forms of signal transforms and timefrequency representations  
US8547105B2 (en)  System and method for detecting a location of fault in a cable  
Liu et al.  Random noise attenuation using fx regularized nonstationary autoregression  
US8751220B2 (en)  Multiple microphone based low complexity pitch detector  
CN101802909B (en)  Speech enhancement with noise level estimation adjustment  
CN100349007C (en)  Method for eliminating single frequency interference in signal of seismological record 
Legal Events
Date  Code  Title  Description 

PB01  Publication  
C06  Publication  
SE01  Entry into force of request for substantive examination  
C10  Entry into substantive examination  
GR01  Patent grant  
C14  Grant of patent or utility model  
CF01  Termination of patent right due to nonpayment of annual fee 
Granted publication date: 20130925 Termination date: 20170620 

CF01  Termination of patent right due to nonpayment of annual fee 