CN102736109B - Method for de-noising, correcting and superposing CRP (Common Reflection Point) gather - Google Patents

Abstract

The invention discloses a method for de-noising, correcting and superposing a CRP (Common Reflection Point) gather, comprising the following steps of: a, performing superposing on an initial CRP gather to obtain a reference trace, alternatively, performing earlier-stage de-noising treatment on the initial CRP gather, wherein the earlier-stage de-noising treatment comprises frequency-domain filtering, wavelet transform, median filtering or mean filtering; b, opening a sliding time window and resolving a correcting value in the time window according to the reference trace, and then correcting the initial CRP gather according to the correcting value to obtain a primary CRP gather; c, orderly repeating the steps a and b by taking the primary CRP gather as the initial CRP gather, so as to obtain a secondary CRP gather; d, de-noising the secondary CRP gather by employing SVD (Singular Value Decomposition), so as to obtain a tertiary CRP gather; and e, superposing the data of the tertiary CRP gather, so as to obtain an achievement profile. The method is capable of correcting the event of the CRP gather to a level and greatly increasing the signal-to-noise ratio of the CRP gather, and also capable of enhancing the continuity of the event of the superposed profile, increasing the signal-to-noise ratio and enhancing a weak signal.

Description

The collection denoising of a kind of CRP road, correction and the method superposed

Technical field

The invention belongs to field of seismic exploration.Seismic prospecting comprises collection, process, explains three large technology, and the core technology wherein processed is deconvolution, superposition and skew, and skew at present mainly develops to Prestack Migration Technology.What more specifically the present invention CRP related in a kind of migration before stack superposed improves one's methods.

Background technology

CRP(common reflection point) road collection be extract in migration before stack process be different from CMP(common midpoint) data acquisition of road collection, the extraction of such road collection be at present mainly used in underground medium rate pattern correction, to be directly added to offset data, prestack inversion work.But the reflective information that can be relatively easy to pick out underground medium same point is concentrated in CRP road, even if be not especially accurately make CRP road concentrate lineups can not complete matching in migration before stack medium velocity information, such CRP road collection directly superposition can cause post-stack migration section resolution step-down or containing fict tectonic information.

Summary of the invention

Task of the present invention is to provide the collection denoising of a kind of CRP road, correct and the method superposed.Its objective is and adopt specific denoising method to improve CRP road collection and to concentrate the correction of lineups to improve offset effect in conjunction with CRP road, the signal to noise ratio (S/N ratio) of migrated section and the fidelity of tectonic information can be improved.

Its technical solution is:

The collection denoising of CRP road, correction and the method superposed, comprise the following steps:

A, initial CRP road collection is carried out superposition ask for library track, and can do denoising in early stage to initial CRP road collection, denoising in early stage comprises frequency domain filtering, wavelet transformation, medium filtering or mean filter; Initial CRP road collection is set as G _cRP(t, i), i=1,2 ..., N, wherein N is that number of channels, the t concentrated is the seismic trace time, i represents the i-th road concentrated, and is set as S with reference to road _r(t),

$S_r\left(t\right)=\left[\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{G}_{\mathrm{CRP}}(t,i)\right]/N$

B, offer sliding window, and time window in ask for correcting value according to above-mentioned library track, then correct initial CRP road collection according to correcting value; Ask in process at correcting value, set up objective function f (t, i),

$f(t,i)=\underset{\mathrm{\τ}=t-T/2}{\overset{\mathrm{\τ}+T/2}{\mathrm{\Σ}}}\left[W\left(\mathrm{\τ}\right)S_r\left(\mathrm{\τ}\right)\right]\·\left[W(\mathrm{\τ}+\mathrm{\Δt}(t,i))G_{\mathrm{CRP}}(\mathrm{\τ}+\mathrm{\Δt}(t,i),i)\right]\→\max$

Wherein W (t) is window function, by obtaining correction amount delta t (t, i) to the optimization of this objective function, utilizing correction amount delta t (t, i) to correct initial CRP road collection and obtaining one-level CRP road collection, being set as by one-level CRP road collection

${\tilde{S}}_{\mathrm{CRP}}^1(t,i)=G_{\mathrm{CRP}}(t+\mathrm{\Δt}(t,i),i)$

C, using one-level CRP road collection as initial CRP road collection, repeat step a, b successively, multiplicity is 1 ~ 2 time, obtains secondary CRP road collection, is set as by secondary CRP road collection

D, by secondary CRP road collection application SVD denoising, obtain three grades of CRP road collection; In SVD denoising process, first secondary CRP road collection is carried out svd, svd expression formula is,

${\tilde{G}}_{\mathrm{CRP}}^2={\mathrm{UEV}}^T$

Wherein E is diagonal matrix, and singular value is distributed on diagonal line, chooses singular value corresponding to advantage energy and forms new diagonal matrix E _p, U and V is unitary matrix, constant in this process.Restructuring secondary CRP road collection obtains three grades of CRP road collection, is set as by three grades of CRP road collection reconstruct singular value expression formula is,

${\tilde{G}}_{\mathrm{CRP}}^3={\mathrm{UE}}_p{V}^T$

Wherein reconstruct singular value number selection range from [1 ~ 1] to [1 ~ N/2];

E, three grades of CRP road collection data to be superposed, obtain achievement section, achievement section is set as S _{stack, j}(t),

$S_{\mathrm{stack},j}\left(t\right)=\left[\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\tilde{G}}_{\mathrm{CRP},j}^3(t,i)\right]/N,j=\mathrm{1,2},...,M$

Wherein j represents jGe road collection, and M is the number of collection.

In above-mentioned steps a, initial CRP road collection signal to noise ratio (S/N ratio) being greater than to 0.5 does not carry out denoising in early stage.

In above-mentioned steps b, the step size settings of sliding window is 1 ~ 16ms, and the length setting of window is 40 ~ 120ms.

The present invention has following Advantageous Effects:

To the present invention is directed in prior art that CRP road collection random noise disturbance is serious, lineups do not have completely more flat feature, propose the lineups partial correction technology for CRP road collection, and utilize svd (SVD) technology it to be carried out to the method for random noise compacting.The present invention not only can make CRP road collection lineups be corrected to level, and its signal to noise ratio (S/N ratio) is improved greatly, and after can making superposition, section lineups continuity strengthens, and signal to noise ratio (S/N ratio) improves, weak signal is strengthened.

Accompanying drawing explanation

Below in conjunction with accompanying drawing and embodiment, the present invention is illustrated further:

Fig. 1 is the FB(flow block) of one embodiment of the present invention.

Fig. 2 shows several variable condition of model data in a kind of processing procedure of the present invention, wherein Fig. 2 a shows original CRP road collection and library track, Fig. 2 b shows the model data after frequency domain filtering denoising, Fig. 2 c shows the model data through overcorrect, and Fig. 2 d shows the model data (achievement section) through svd denoising.

Fig. 3 shows several variable condition of model data when the relevant key element of sliding window in the present invention changes, situation when namely sliding window length of window is set as 100ms, iteration 2 times; Wherein Fig. 3 a shows original CRP road collection and library track, Fig. 3 b shows model data when sliding window step size settings is 2ms after denoising, Fig. 3 c shows model data when sliding window step size settings is 4ms after denoising, Fig. 3 d shows model data when sliding window step size settings is 8ms after denoising, Fig. 3 e shows model data when sliding window step size settings is 16ms after denoising, and Fig. 3 f shows model data when sliding window step size settings is 20ms after denoising.

Fig. 4 shows several variable condition of model data when the relevant key element of sliding window in the present invention changes, situation when namely sliding window step size settings is 16ms, iteration 2 times; Wherein Fig. 4 a shows model data when sliding window length of window is set as 40ms after denoising, Fig. 4 b shows model data when sliding window length of window is set as 80ms after denoising, Fig. 4 c shows model data when sliding window length of window is set as 120ms after denoising, and Fig. 4 d shows model data when sliding window length of window is set as 160ms after denoising.

Fig. 5 shows several variable condition of model data during other relevant key elements changes of sliding window in the present invention, situation when namely sliding window step size settings is 16ms, length of window is set as 80ms; Model data when wherein Fig. 5 a shows iteration one time, Fig. 5 b shows model data during iteration twice, and Fig. 5 c shows model data during iteration three times, and Fig. 5 d shows model data during iteration six times.

Fig. 6 shows the present invention and is applied in the comparatively flat and result (shallow-layer) of CRP road, a certain work area collection.

Fig. 7 shows the present invention and is applied in the comparatively flat and result (middle level) of CRP road, a certain work area collection.

Fig. 8 shows a kind of CRP road collection stack result taking existing denoising method to obtain.

Fig. 9 shows a kind of CRP road collection stack result obtained by the present invention.

Embodiment

Composition graphs 1 and Fig. 2, the collection denoising of a kind of CRP road, correction and the method superposed, comprise the following steps:

A, initial CRP road collection is carried out superposition ask for library track, and can do denoising in early stage to initial CRP road collection, denoising in early stage comprises frequency domain filtering, wavelet transformation, medium filtering or mean filter; Initial CRP road collection is set as G _cRP(t, i), i=1,2 ..., N, wherein N is that number of channels, the t concentrated is the seismic trace time, i represents and concentrate the i-th road, is set as S with reference to road _r(t),

$S_r\left(t\right)=\left[\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{G}_{\mathrm{CRP}}(t,i)\right]/N---\left(1\right)$

In this step, higher for signal to noise ratio (S/N ratio), the initial CRP road collection being such as greater than 0.5 can not carry out denoising in early stage.For setting up model data, the result of denoising in early stage is see Fig. 2 b, and library track is see Fig. 2 a right diagram part.

B, offer sliding window, and time window in ask for correcting value according to above-mentioned library track, then correct initial CRP road collection according to correcting value; Ask in process at correcting value, set up objective function f (t, i),

$f(t,i)=\underset{\mathrm{\τ}=t-T/2}{\overset{\mathrm{\τ}+T/2}{\mathrm{\Σ}}}\left[W\left(\mathrm{\τ}\right)S_r\left(\mathrm{\τ}\right)\right]\·\left[W(\mathrm{\τ}+\mathrm{\Δt}(t,i))G_{\mathrm{CRP}}(\mathrm{\τ}+\mathrm{\Δt}(t,i),i)\right]\→\max ---\left(2\right)$

Wherein W (t) is window function, by obtaining correction amount delta t (t, i) to the optimization of this objective function, utilizing correction amount delta t (t, i) to correct initial CRP road collection and obtaining one-level CRP road collection, being set as by one-level CRP road collection

${\tilde{S}}_{\mathrm{CRP}}^1(t,i)=G_{\mathrm{CRP}}(t+\mathrm{\Δt}(t,i),i)---\left(3\right)$

In this step, the step-length of sliding window can be set as 1 ~ 16ms, and the length of window can be set as 40 ~ 120ms.The result that this step obtains is see Fig. 2 c.

C, using one-level CRP road collection as initial CRP road collection, repeat step a, b successively, multiplicity is 1 ~ 2 time, obtains secondary CRP road collection, is set as by secondary CRP road collection

D, by secondary CRP road collection application SVD denoising, obtain three grades of CRP road collection; In SVD denoising process, first secondary CRP road collection is carried out svd, svd formula,

${\tilde{G}}_{\mathrm{CRP}}^2={\mathrm{UEV}}^T---\left(4\right)$

Wherein E is diagonal matrix, and singular value is distributed on diagonal line, chooses singular value corresponding to advantage energy and forms new diagonal matrix E _p, U and V is unitary matrix, constant in this process.Restructuring secondary CRP road collection obtains three grades of CRP road collection, is set as by three grades of CRP road collection reconstruct singular value expression formula is,

${\tilde{G}}_{\mathrm{CRP}}^3={\mathrm{UE}}_p{V}^T---\left(5\right)$

Wherein reconstruct singular value number selection range from [1 ~ 1] to [1 ~ N/2].The result that this step obtains is see Fig. 2 d.

E, three grades of CRP road collection data to be superposed, obtain achievement section, achievement section is set as S _{stack, j}(t),

$S_{\mathrm{stack},j}\left(t\right)=\left[\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\tilde{G}}_{\mathrm{CRP},j}^3(t,i)\right]/N,j=\mathrm{1,2},...,M---\left(6\right)$

Wherein j represents jGe road collection, and M is the number of collection.

The data investigation result that the present invention obtains can see Fig. 9, and the data investigation result that existing denoising method obtains is as Fig. 8.Just can find through contrast, the present invention can improve the signal to noise ratio (S/N ratio) of migrated section greatly, and improves the fidelity of tectonic information.

Further, inventor is also for when in the present invention, certain some key element changes, and these changes have done more deep research to the impact of data stack result (model data).Such as:

One, several variable condition of model data when the relevant key element of sliding window changes in the present invention, situation when namely sliding window length of window is set as 100ms, iteration 2 times; See Fig. 3.Note: correct the distortion of result lineups when step-length is chosen as 20ms.

Two, several variable condition of model data when the relevant key element of sliding window changes in the present invention, situation when namely sliding window step size settings is 16ms, iteration 2 times; See Fig. 4.Note: when window is less, pseudo-axle is more, lineups distortion or correct not thorough when window is larger.

Three, several variable condition of model data when other relevant key elements of sliding window change in the present invention, situation when namely sliding window step size settings is 16ms, length of window is set as 80ms; See Fig. 5.Note: the increase of iterations is not obvious on the impact of calibration result, but more iterations can ensure the stability that calculates.

So visible, if consider from computing velocity, ensureing that, on the basis that correcting algorithm is stablized with outcome quality, step-length should be as far as possible long, window as far as possible short, iterations is tried one's best few.Such as: the step-length of sliding window is preferably 60 ~ 100ms, the length of window is preferably 1 ~ 16ms; Iterations is preferably 2 times.

Here is embody rule of the present invention example:

Inventor applies the present invention to all CRP roads collection of certain work area survey line, is the result of one of them CRP road collection as shown in Figure 6 and Figure 7.Can find out, the effect in signal to noise ratio (S/N ratio) raising and lineups correction is quite obvious.To the section that the net result of its process goes out for the superposition shown in Fig. 9, for the result (shown in Fig. 8) of raw data superposition, the lifting of its treatment effect is reflected on the whole, comprise the various aspects such as the lifting of signal to noise ratio (S/N ratio), the enhancing of weak signal energy, the continuity of lineups, effect is very outstanding.

The relevant technologies content do not addressed in aforesaid way is taked or uses for reference prior art to realize.

It should be noted that, under the instruction of this instructions, those skilled in the art can also make such or such easy variation pattern, such as equivalent way, or obvious mode of texturing.Above-mentioned variation pattern all should within protection scope of the present invention.

Claims (1)

1. CRP road collection denoising, correction and the method superposed, is characterized in that comprising the following steps:

A, initial CRP road collection is carried out superposition ask for library track, and do denoising in early stage to initial CRP road collection, denoising in early stage comprises frequency domain filtering, wavelet transformation, medium filtering or mean filter; Initial CRP road collection is set as G _cRP(t, i), i=1,2 ..., N, wherein N is that number of channels, the t concentrated is the seismic trace time, i represents and concentrate the i-th road, is set as S with reference to road _r(t),

$S_r\left(t\right)=\left[\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{G}_{\mathrm{CRP}}(t,i)\right]/N$

B, offer sliding window, and time window in ask for correcting value according to above-mentioned library track, then correct initial CRP road collection according to correcting value; Ask in process at correcting value, set up objective function f (t, i), wherein τ is time shift amount, and T is fixing window length,

$f(t,i)=\underset{\mathrm{\τ}=t-T/2}{\overset{\mathrm{\τ}+T/2}{\mathrm{\Σ}}}\left[W\left(\mathrm{\τ}\right)S_r\left(\mathrm{\τ}\right)\right]\·\left[W(\mathrm{\τ}+\mathrm{\Δt}(t,i))G_{\mathrm{CRP}}(\mathrm{\τ}+\mathrm{\Δt}(t,i),i)\right]\→\max$

Wherein W (t) is window function, by obtaining correcting value △ t (t, i) to the optimization of this objective function, utilizing correcting value △ t (t, i) to correct initial CRP road collection and obtaining one-level CRP road collection, being set as by one-level CRP road collection

${\tilde{G}}_{\mathrm{CRP}}^1(t,i)=G_{\mathrm{CRP}}(t+\mathrm{\Δ}(t,i),i)$

C, using one-level CRP road collection as initial CRP road collection, repeat step a, b successively, multiplicity is 1 ~ 2 time, obtains secondary CRP road collection, is set as by secondary CRP road collection be abbreviated as

D, by secondary CRP road collection application SVD denoising, obtain three grades of CRP road collection; In SVD denoising process, first secondary CRP road collection is carried out svd, svd expression formula is,

${\tilde{G}}_{\mathrm{CRP}}^2={\mathrm{UEV}}^T$

Wherein U and V is unitary matrix, V ^tfor the transposed matrix of V, E is diagonal matrix, and singular value is distributed on diagonal line, chooses singular value corresponding to advantage energy and forms new diagonal matrix E _p, and secondary CRP road collection of recombinating obtains three grades of CRP road collection, is set as by three grades of CRP road collection be abbreviated as reconstruct singular value expression formula is,

${\tilde{G}}_{\mathrm{CRP}}^3={\mathrm{UE}}_p{V}^T$

Wherein reconstruct singular value number selection range from 1 ~ N/2;

E, three grades of CRP road collection data to be superposed, obtain achievement section, achievement section is set as S _{stack, j}(t),

$S_{\mathrm{stack},j}\left(t\right)=\left[\underset{i=1}{\overset{N}{\text{\Σ}}}{\tilde{G}}_{\mathrm{CRP},j}^3(t,i)\right]/N,j=\mathrm{1,2},...,M$

Wherein j represents jGe road collection, and M is the number of collection;

In described step b, the step size settings of sliding window is 1 ~ 16ms, and the length setting of window is 40 ~ 120ms.