CN105182419A - Pre-stack seismic signal event alignment method based on global optimization - Google Patents

Abstract

The invention discloses a pre-stack seismic signal event alignment method based on global optimization. A time window central point is used as a seed point, a time window is made to continuously slide to obtain a movement amount when the similarity among trace gathers is the greatest as the movement amount of the seed point, global optimization of the movement amount of the seed point is performed, and finally the aligned pre-stack trace gather amplitude is obtained by using the movement amount of the seed point as original data for interpolation through a cubic spline function. The method is advantageous in that, a reference seismic trace is selected through affinity similarity propagation, trace gathers with high similarity can be automatically selected as the reference traces, manual intervention is not needed, and accuracy is higher compared with accuracy obtained through directly and randomly selecting one trace from the higher similarity trace gathers in a similarity matrix. After the movement amount of the seed point is obtained, a new target function and a constraint condition are designed for global optimization of the movement amount of the seed point, and therefore, the problems of sudden jump in the alignment process and waveform distortion are solved.

Description

A kind of pre-stack seismic signal lineups based on global optimization even up method

Technical field

The invention belongs to seismic exploration technique field, be specifically related to the design that a kind of pre-stack seismic signal lineups based on global optimization even up method.

Background technology

More and more meticulousr along with seismic prospecting, migration before stack and AVO inverting have very important effect at engineering construction system and the complex lithology reservoir field of prediction.For improving the precision of lithology imaging, the prestack elastic parameter inversion method in conjunction with Prestack Migration Technology is a good selection.Prestack road collection data, as the basis of inverting, have vital effect to efficiency of inverse process.Owing to being subject to the anisotropic impact of stratum media, prestack road is concentrated exists a large amount of residual move out time, and cause prestack road to concentrate lineups uneven, lineups out-of-flatness can cause imaging effect inaccurate, thus affects the effect of inverting.

At present conventional prestack road collection method of evening up mainly is divided into two large classes: the method for evening up adjusted based on speed and Corpus--based Method effect even up method.The method of evening up of Corpus--based Method effect refers to the objective function first setting up L2 norm, objective function represents primarily of AVOI or III class, then utilize the sliding window of time orientation to produce the mobile of per pass to separate, finally minimize objective function, corresponding solution is optimum solution.The method of evening up based on speed adjustment is that supposition concentrates the seismic event of not evening up to cause mainly due to residual move out time (RMO) in original pre-stack seismic road, therefore uses the high precision velocity of second order or quadravalence RMS velocity field to estimate to even up the lineups of collection.

What 1, adjust based on speed evens up method

In pre-stack seismic road after supposing original noise reduction based on the road collection method of evening up of speed adjustment, the road collection of not evening up causes mainly due to residual move out time (RMO), therefore uses the high precision velocity of the RMS velocity field of second order or quadravalence to estimate to even up the lineups of collection.RMO corrects with the formula based on (1):

$\mathrm{\τ}(x,t)=x^2(V^{-2}-V_{ref}^{-2})/2t---\left(1\right)$

Wherein, τ is dynamic correction value, and x is offset distance, and t is the time at zero-offset place, V _refbe reference velocity function, V is renewal speed.Then, seismic prospecting development makes the distance between seismic shoot point and acceptance point more and more far away.This reason makes offset distance a long way off, uses the more and more difficult description rate pattern of RMO curve.Use Alkhalifah time difference model through improving, model formation is as formula (2):

$t^2=t_0^2+\frac{x^2}{V^2}-\frac{2{\mathrm{\ηx}}^4}{V^2(V^2{t}_0^2+{\left(1+2\mathrm{\η}\right)}^2)}---\left(2\right)$

For high-order residual move out time Negotiation speed residual quantity δ V and hourage δ η determine time difference δ t, as shown in formula (3):

$\begin{array}{c}{(t+\mathrm{\δ}t)}^2-t^2=x^2(\frac{1}{{(V+\mathrm{\δ}V)}^2}-\frac{1}{V^2})\\ -\[\frac{2\left(\mathrm{\η}+\mathrm{\δ}\mathrm{\η}\right)x^4}{{\left(V+\mathrm{\δ}V\right)}^2\left({\left(V+\mathrm{\δ}V\right)}^2{t}_0^2+\left(1+2\left(\mathrm{\η}+\mathrm{\δ}\mathrm{\η}\right)\right)x^2\right)}-\frac{2{\mathrm{\ηx}}^4}{V^2\left(V^2{t}_0^2+\left(1+2\mathrm{\η}\right)x^2\right)}\]\end{array}---\left(3\right)$

In order to improve dimensionless offset distance, make ζ=x/Vt ₀with dimensionless speed Δ=δ V/V, substituting into formula (3) can obtain:

$\begin{array}{c}\left(\frac{2t}{t_0}+\frac{\mathrm{\δ}t}{t_0}\right)\frac{\mathrm{\δ}t}{t_0}=\frac{-\left(2+\mathrm{\Δ}\right)\mathrm{\Δ}}{{\left(1+\mathrm{\Δ}\right)}^2}{\mathrm{\ξ}}^2-2{\mathrm{\ξ}}^4\\ \×\[\frac{\mathrm{\η}+\mathrm{\δ}\mathrm{\η}}{{(1+\mathrm{\Δ})}^2({\left(1+\mathrm{\Δ}\right)}^2+(1+2\left(\mathrm{\η}+\mathrm{\δ}\mathrm{\η}\right)\left){\mathrm{\ξ}}^2\right)}-\frac{\mathrm{\η}}{(1+(1+2\mathrm{\η}\left){\mathrm{\ξ}}^2\right)}\]\end{array}---\left(4\right)$

Try to achieve δ V and δ η by the error minimized between input data and time shift amount δ t, iterations can oneself setting.

2, Corpus--based Method effect even up method

Hinkley proposed a kind of dynamic road collection and evens up method (DGF) in 2004, it is that method evened up by a kind of road collection of statistics, first this method maps one by one in processing procedure, namely each output sample point data obtains through process by with the input data on same time point, can be expressed more easily by formula (5):

D _a(t,x)＝D _b{(t+m(t,x)),x}(5)

Wherein, x is offset distance, evens up in method the road collection numbering that also can regard as and sort from small to large at this road collection; T is the time, a and b represents collection respectively and even up rear and before evening up data.Window during by opening lateral offset and longitudinal time, moves by road and makes 2 Norm minimums between twice, namely solve formula (6):

${\∫}_{t_{\min }}^{t_{\max }}F\left(t,\mathrm{\τ}\right)={\∫}_{t_{\min }}^{t_{\max }}{\left(a\left(t+\mathrm{\τ}\right)-b\left(t-\mathrm{\τ}\right)\right)}^{2}dt\→\min ---\left(6\right)$

The time shift τ of any twice can be obtained _ij.Taking up an official post in offset distance direction, to get 5 roads be one group, wherein T ₁represent the amount of movement between first and second, T ₂represent the 3rd amount of movement between road and first, T ₃represent the amount of movement of the 4th road and first, T ₄represent the amount of movement of the 5th road and first.5 track datas can ask for ten amount of movements, and namely all there is an amount of movement between twice arbitrarily, under least squares sense, try to achieve above 4 amount of movements, computing formula is as shown in formula (7):

T ₁＝(2T _1,2+T _1,3+T _1,4+T _1,5-T _2,3-T _2,4-T _2,5)/5

T ₂＝(T _1,2+2T _1,3+T _1,4+T _1,5+T _2,3-T _3,4-T _3,5)/5

T ₃＝(T _1,2+T _1,3+2T _1,4+T _1,5+T _2,4+T _3,4-T _4,5)/5

T ₄＝(T _1,2+T _1,3+T _1,4+2T _1,5+T _2,5-T _3,5-T _4,5)/5(7)

In prestack road collection is optimized, although can ensure that lineups are evened up substantially by normal moveout correction, but due to some factors, such as because just the rise and fall normal moveout correction that causes of earth's surface is inaccurate, because of the error of the traveltime-distance equation that horizontal layer isotropic medium causes.Existence due to these impacts makes prestack road collection lineups still uneven, needs to do meticulous evening up further.

Summary of the invention

The object of the invention is to solve in prior art owing to being subject to the anisotropic impact of stratum media, prestack road is concentrated exists a large amount of residual move out time, prestack road is caused to concentrate lineups uneven, and then cause imaging effect inaccurate, thus affect the problem of efficiency of inverse process, propose a kind of pre-stack seismic signal lineups based on global optimization and even up method.

Technical scheme of the present invention is: a kind of pre-stack seismic signal lineups based on global optimization even up method, comprise the following steps:

S1, initialization road collection flattening parameters;

S2, choose benchmark road;

S3, calculating Seed Points amount of movement;

S4, lineups are evened up.

Further, window size, window amount of movement, search radius and similarity matrix fractile threshold value when in step S1, road collection flattening parameters comprises prestack road collection.

Further, step S2 comprises step by step following:

S21, calculate the similarity matrix C of any Liang Ge road collection;

S22, based on similarity matrix initialization Attraction Degree matrix and degree of membership matrix;

S23, iteration upgrade Attraction Degree matrix and degree of membership matrix;

S24, calculate and make Attraction Degree matrix and the maximum road collection k of degree of membership matrix sum;

S25, judge whether iterations reaches predetermined number of times, if then enter step S3, otherwise enter step S26;

S26, judge that whether collection k is consistent with result during last iteration, if then enter step S3, otherwise return step S23.

Further, step S3 comprises step by step following:

S31, ask for the maximum similarity Matrix C of Liang Ge road collection _maxand define optimum amount of movement matrix S;

S32, compute matrix C _maxthe value c at corresponding fractile threshold value place _m;

S33, statistics are greater than c _mthe road collection number of value, selects the maximum row of number, the amount of movement of window Seed Points when amount of movement corresponding to this row similarity is current;

The amount of movement of window Seed Points when S34, interpolation obtain all the other;

S35, global optimization is carried out to window Seed Points amount of movement time each.

Further, step S35 comprises step by step following:

The amount of movement matrix that between S351, calculating road collection, similarity is maximum;

The displacement difference sub matrix of S352, calculated level direction and vertical direction;

Whether the displacement difference sub matrix of S353, determined level direction and vertical direction meets constraint condition, if then enter step S4, otherwise selects the amount of movement of similarity suboptimum form new amount of movement matrix and return step S352.

Further, adopt cubic spline interpolation to realize lineups to the method that amount of movement matrix carries out interpolation in step S4 to even up.

The invention has the beneficial effects as follows: when the present invention utilizes, window center point is as Seed Points, amount of movement when continuous sliding window tries to achieve that between collection, similarity is maximum is as the amount of movement of Seed Points, global optimization is carried out to Seed Points amount of movement, obtain the prestack road collection amplitude after evening up finally by cubic spline function using Seed Points amount of movement as raw data interpolation, following beneficial effect can be reached:

(1) choose benchmark seismic trace by neighbour's similar propagation, automatically can choose the high road collection of similarity as benchmark road, not need manual intervention, concentrate optional one accuracy higher compared to the direct road that similarity is higher from similarity matrix.

(2) after obtaining Seed Points amount of movement, design new objective function and constraint condition carries out global optimization to Seed Points amount of movement, solve the problem of " kick " and the waveform distortion of evening up in process.

Accompanying drawing explanation

Fig. 1 is that a kind of pre-stack seismic signal lineups based on global optimization provided by the invention even up method flow diagram.

Fig. 2 is the process flow diagram step by step of step S2 of the present invention.

Fig. 3 is the process flow diagram step by step of step S3 of the present invention.

Fig. 4 is similarity and amount of movement graph of a relation.

Fig. 5 is the process flow diagram step by step of step S35 of the present invention.

Fig. 6 is the road composite section figure do not evened up.

Fig. 7 is the road composite section figure evened up not carrying out the optimization of Seed Points amount of movement.

Fig. 8 carries out the road collection flattened section figure after global optimization to Seed Points amount of movement.

Embodiment

Below in conjunction with accompanying drawing, embodiments of the invention are further described.

The invention provides a kind of pre-stack seismic signal lineups based on global optimization and even up method, as shown in Figure 1, comprise the following steps:

S1, initialization road collection flattening parameters.

Window size N when road collection flattening parameters comprises prestack road collection _w, window amount of movement N _s, search radius (being less than window amount of movement) N _rwith similarity matrix fractile threshold value Toi.

The usual prestack road data sample number of collection on time orientation is several thousand sampling points, window when time orientation is got, and window can move up and down, time window in the number of data points that comprises window size N when being designated as _w.N _wsize by way of two cycles of collection waveform, can arbitrary extracting road collection do Fourier transform together, number of data points in estimation dominant frequency determination one-period.

The scope that window moves up and down is no more than window amount of movement N _s.Window amount of movement N _susually one-period data volume size is got.

Search radius (being less than window amount of movement) N _rmaximum amount of movement when window slides up and down when referring to.

Similarity matrix fractile threshold value Toi, fractile under first simple introduction before how introduction arranges this parameter, fractile, from the point every equal intervals in data acquisition, all needed the number in data acquisition to carry out descending sequence before asking fractile.As 2-fractile, exactly data are divided into two parts, the data volume being less than and being greater than data point corresponding to 2-fractile respectively accounts for the half of whole data acquisition, in like manner known, corresponding 3 data points of 4-fractile, data acquisition is divided into four equal parts by these 3 data points, makes every partial data represent 1/4th of whole Data distribution8.Similarity matrix fractile threshold value Toi gets the 3rd 4-fractile, i.e. value 0.75 of similarity matrix usually, and for the prestack road collection that road collection is second-rate, Toi can get between 0.8-0.9.

S2, choose benchmark road.

Before carrying out collection and evening up, first need the road collection determining reference, concentrate in section in prestack road, a section has tens roads usually, needs to select together as benchmark road from this tens road.

The calculating relying on similarity between Dao Ji and road collection is chosen in benchmark road, adopts neighbour's similar propagation iteration to ask for benchmark road, as shown in Figure 2, comprises the following steps:

S21, every together from time 0ms to prestack road composite section, with time window size N _w.for length intercepts size, calculate the similarity between any Liang Ge road collection according to formula (8):

$C(j_1,j_2)=\frac{f_{j_2}\·f_{j_2}}{\sqrt{f_{j_1}\·f_{j_1}\sqrt{f_{j_2}\·f_{j_2}}}}---\left(8\right)$

Wherein j ₁and j ₂represent the label of not people having a common goal, span is 1 ... N, N are the number of channels of collection.The Matrix C calculated is exactly similarity matrix.

S22, based on similarity matrix C, definition Attraction Degree matrix R (i, k) with degree of membership matrix A (i, k) two category informations between road collection are represented, wherein R (i, k) collection i in Shi Cong road points to candidate reference road k, it reflects the appropriate level that collection k is suitable as the benchmark road collection of collection i; A (i, k) points to road collection i from candidate reference road k, it reflects collection i and selects k as the appropriate level in its benchmark road.

For the ease of calculating, each element in Matrix C is got negative sign, definition Attraction Degree matrix R (i, k) and degree of membership matrix A (i, k) two objective functions, iterative R (i, k) and A (i, k).Size and the similarity matrix C of Attraction Degree and degree of membership matrix are in the same size, all N × N, be 0 matrix by these two matrix initialisation, definition Attraction Degree objective function and degree of membership objective function carry out iteration renewal, and defined function is represented by formula (9), (10), (11), (12):

$R(i,k)=C(i,k)-\underset{k^{\′}\≠k}{max}\{A(i,k^{\′})+C(i,k^{\′})\}---\left(9\right)$

$A(i,k)=\min \{0,R(k,k)+\underset{i\∉\{i,k\}}{\Σ}max\{0,R\left(i^{\′},k\right)\left\}\right\}---\left(10\right)$

R(k,k)＝P(k)-max{A(k,i′)+C(k,i′)}(11)

$A(k,k)=\underset{i^{\′}\∉\{i,k\}}{\Σ}max\{0,R(i^{\′},k)\}---\left(12\right)$

Wherein, P (k) gets the average of Matrix C.

S23, upgrade Attraction Degree matrix and degree of membership matrix by formula (13) iteration:

$\left\{\begin{array}{c}{R}_{new}(i,k)=\mathrm{\λ}\×R_{old}(i,k)+(1-\mathrm{\λ})\×R(i,k)\\ A_{new}(i,k)=\mathrm{\λ}\×A_{old}(i,k)+(1-\mathrm{\λ})\×A(i,k)\end{array}\right.---\left(13\right)$

In formula, λ (0 < λ < 1) is convergence coefficient (ratio of damping), be mainly used in the stability regulating convergence of algorithm speed and iterative process, the larger oscillatory occurences of λ becomes more not obvious, this means that effect is better, but shortcoming be speed of convergence will be slack-off, otherwise, λ reduces, although the very fast reforming phenomena of speed of convergence clearly.Subscript old and new represents net result that is last and this updating message respectively.Stopping criterion for iteration be iterations be greater than preset number of times or number of iterations time after R (i, k) remain unchanged with A (i, k).

S24, to be calculated by formula (14) and make Attraction Degree matrix and the maximum road collection k of degree of membership matrix sum:

$k=\arg \underset{k}{max}\{A(i,k)+R(i,k)\}---\left(14\right)$

If i=k in formula, then collection i in road is the benchmark road collection of this prestack road composite section; If i ≠ k, then collection k in road is the benchmark road collection of collection i.

S25, judge whether iterations reaches predetermined number of times, if then enter step S3, otherwise enter step S26.

S26, judge that whether collection k is consistent with result during last iteration, if then enter step S3, otherwise return step S23.

In order to enable Attraction Degree matrix and degree of membership matrix Fast Convergent, introduce contraction factor ρ, the definition of ρ is as shown in formula (15):

Wherein, for constructed fuction variable.Carry out renewal to R (i, k) with A (i, k) two matrixes to obtain:

$R_{i+1}(i,k)=\mathrm{\&rho;}\&CenterDot;\mathrm{\&lambda;}\&CenterDot;R_i(i,k)+(1-\mathrm{\&lambda;})\&CenterDot;R_{i+1}^{old}(i,k)---\left(16\right)$

$A_{i+1}(i,k)=\mathrm{\&rho;}\&CenterDot;\mathrm{\&lambda;}\&CenterDot;A_i(i,k)+(1-\mathrm{\&lambda;})\&CenterDot;A_{i+1}^{old}(i,k)\mathrm{\&lambda;}\&Element;\&lsqb;0.5,1\&rsqb;---\left(17\right)$

In the embodiment of the present invention, value is 4.1, and the value that therefore can calculate ρ is 0.729.

S3, calculating Seed Points amount of movement.

After selecting benchmark seismic trace, in prestack road composite section, select some data point as Seed Points and calculate the amount of movement of Seed Points.With time window size N _wfor length, window when time orientation is opened, with time window amount of movement N _ssized by sliding window until time window cover time upper all sampling points, using window center point time each as Seed Points.

As shown in Figure 3, calculate Seed Points amount of movement and mainly comprise following step:

Amount of movement when similarity is maximum between S31, recording channel collection, defines optimum amount of movement matrix S, and the calculating of S is as shown in formula (18):

$S(j_1,j_2)=\arg \underset{l\&Element;N_r}{max}\frac{f\&CenterDot;g_l}{\sqrt{f\&CenterDot;f}\sqrt{g_l\&CenterDot;g_l}}---\left(18\right)$

If i _s=(k-1) N _m, definition set $\hat{L}=\left\{-N_s,-N_s+1,\mathrm{......},0,\mathrm{......},N_s-1,N_s\right\},$ G in formula _lfor definition length is N _sthe vector of+1, g _li-th element g _l(i)=D (i _s+ 1+i-1, j ₂), D is the pre-stack seismic road collection data of input.

Fixing road collection j ₁time window, move up and down collection j ₂time window, each movement is spaced apart a data point, and the border of movement is no more than search radius N _r.In search procedure, record the amount of movement of the highest window constantly of similarity, Similarity Measure carries out according to formula (8), tries to achieve the Matrix C of record maximum similarity _max.

S (j ₁, j ₂) be a unsymmetrical matrix, such as: S (2,1) and S (1,2), S (2,1) represents that fixing road collection 2 moves the optimum amount of movement of the 1 road collection that collection 1 obtains, and S (1,2) represents that fixing 1 road collection moves the optimum amount of movement that collection 2 obtains.Diagonal entry is all 0 because on diagonal line element representative, the auto-correlation of road collection self, obviously, time window do not carry out any mobile time similarity maximum.In order to vivider explanation, Fig. 4 illustrates the relation between similarity and amount of movement, and wherein horizontal ordinate represents amount of movement, and being moved to the left amount of movement is negative, and the amount of movement that moves right is positive number, and ordinate is coefficient of similarity.As can be seen from the figure, road collection 2 move up 3 sampling interval time related coefficient reach maximum, record the amount of movement that namely this amount of movement is exactly this Seed Points.

S32, compute matrix C _maxthe value c at corresponding fractile threshold value place _m, such as, suppose have 6 prestack road, road collection to obtain C _maxshown in matrix following table, for the c that this matrix is obtained _m=0.85.

Number of channels	1	2	3	4	5	6	Number
								1	1	0.9	0.71	0.35	0.89	0.87	4
2	0.9	1	0.76	0.43	0.95	0.96	4

3	0.71	0.76	1	0.78	0.72	0.69	1
								4	0.35	0.43	0.78	1	0.64	0.41	1
5	0.89	0.95	0.72	0.64	1	0.93	4
								6	0.87	0.96	0.69	0.41	0.93	1	4

Statistics is greater than c _mthe road collection number of value, selects the row that number is maximum, and this just means that this group road collection similarity is the strongest, very similar between road collection, can by strengthening the similarity between road collection time mobile.The situation not unique for the most multirow of number can select a line from maximum row, the first row can be selected according in upper table, by S (1,1), S (1 in the first row of optimum amount of movement matrix, 2), S (1,5), S (1,6) retain, other are given up.This amount of movement corresponding to row similarity is as the amount of movement of window Seed Points time current, and therefore in upper table, 1,2,5,6 road Seed Points amount of movements are determined.

S34, for not determining that the Seed Points of amount of movement obtains according to formula (19) interpolation:

Wherein, j _brepresent the footnote of selected road collection, j _minwith j _maxwhat represent respectively is the maximum Taoist monastic name of amount of movement and minimum Taoist monastic name. represent selected road collection set.In more than showing, 3,4 roads can be obtained by the amount of movement interpolation in all the other four roads.

The amount of movement of window center point when can obtain each by above-mentioned steps, by Seed Points amount of movement composition matrix m on whole prestack road composite section.The amount of movement now obtained is coarse, if implement the mobile phenomenon obtaining the lineups after evening up and there will be " kick " with these amount of movements to prestack road collection.Why there will be this phenomenon, conclude its reason and have two aspects: on the one hand the calculating of similarity carries out measuring similarity to a whole segment signal, find the strongest amount of movement of similarity and do not mean that the aligned in position of lineups in this section of waveform obviously (amplitude is larger).On the other hand, some Seed Points amount of movement is that interpolation obtains, these amount of movements accurately can not reflect lineups degrees of offset between this road collection and benchmark road, therefore can not directly using these amount of movements as Seed Points amount of movement, need optimize amount of movement obtain amount of movement more accurately.

S35, global optimization is carried out to window Seed Points amount of movement time each.

Except above-mentioned " kick " phenomenon, on time orientation, if the Seed Points amount of movement difference of window center exceeds search radius and also to cause finally implementing when collection is evened up waveform compression or stretch too severe time adjacent, make waveform distortion within unacceptable scope.Therefore need to carry out global optimization to window Seed Points amount of movement time each, as shown in Figure 5, concrete steps are as follows:

S351, calculate according to formula (20) the amount of movement matrix that between road collection, similarity is maximum:

$f(x,y)=\arg \underset{\mathrm{\&Delta;}l\&Element;L}{max}\frac{g_x\&CenterDot;(g_y+\mathrm{\&Delta;}l)}{\sqrt{g_x\&CenterDot;g_y}\sqrt{(g_y+\mathrm{\&Delta;}l)\&CenterDot;(g_y+\mathrm{\&Delta;}l)}}---\left(20\right)$

Wherein, g _xwith g _ydifferent prestack road collection waveform in window when being same respectively, Δ l is amount of movement, and arg symbol represents that functional value gets dependent variable.

The displacement difference sub matrix of S352, calculated level direction and vertical direction with

Whether the displacement difference sub matrix of S353, determined level direction and vertical direction meets constraint condition, and constraint condition is as shown in formula (21):

$\begin{array}{ccc}s.t& \left|\right|{\&dtri;}_{y}f(x,y)|{|}_{\mathrm{\&infin;}}\&le;R& \left|\right|{\&dtri;}_{x}f\left(x,y\right)|{|}_{\mathrm{\&infin;}}\&le;R\end{array}---\left(21\right)$

In formula, s.t. represents constraint condition, and R represents threshold value, || || _∞be matrix Infinite Norm symbol, represent the element asking matrix maximum.

If then enter step S4, otherwise the amount of movement of similarity suboptimum is selected to form new amount of movement matrix and return step S352.

Constraint condition ensure that the difference of neighboring seeds point amount of movement is no more than search radius N on horizontal and vertical _r.The free-air correction factor of global optimization can be obtained thus, i.e. the amount of movement of horizontal direction and vertical direction.

Original pre-stack seismic road composite section as shown in Figure 6, adopt road collection to even up algorithm and have very large lifting to the lineups flatness that rear the entire profile evened up by road collection, but there is the phenomenon of " kick " in the part lineups in circle, originally lineups become two lineups, as shown in Figure 7, obtain Fig. 8 after carrying out global optimization to Seed Points amount of movement, in circle, part can find out that lineups problem of " kick " while evening up is resolved.

S4, lineups are evened up.

After calculating Seed Points amount of movement matrix m, when supposing q, window center point coordinate is i _q, the amount of movement of its correspondence is q=1,2 ..., Q, definition set define the two-dimensional array X the same with prestack road collection size of data ^newbe used for storing the amount of movement of prestack road collection each data sample, can be tried to achieve by interpolation, as shown in formula (22):

X ^newstore stretching sample coordinate, by the matrix D of formed objects _newrepresent lineups even up after prestack road collection, represent the coordinate of original prestack road collection with matrix X, D matrix represents the amplitude of original prestack road collection.D _newask for according to D, X, X ^newcarry out cubic spline interpolation to obtain.Be complete one j for any given together during interpolation, arrange as independent variable with the jth of X, arrange as functional value with the jth of D, build cubic spline functions as shown in formula (23)

$D_{new}=f_{X_j,D_j}\left(X^{new}\right(i,j\left)\right)---\left(23\right)$

Cubic spline difference can obtain comparatively level and smooth result, and its Interpolation Principle a given n+1 data point is divided into n interval, and cubic spline equation meets the following conditions:

(1) at each piecewise interval [x _i, x _i+1] (i=0,1 ..., n-1, x increase progressively), S (x)=S _ix () is a cubic polynomial.

(2) S (x is met _i)=y _i(i=0,1 ..., n).

(3) the first order derivative S'(x of S (x)) and second derivative S " (x) is all continuous print in [a, b] interval, and namely S (x) curve is smooth.So n cubic polynomial segmentation can be write:

S _i(x)＝a _i+b _i(x-x _i)+c _i(x-x _i) ²+d _i(x-x _i) ³(24)

Wherein, a _i, b _i, c _i, d _irepresent 4n unknowm coefficient, expression is as follows:

$\begin{array}{c}{a}_i=y_i\\ b_i=\frac{y_{i+1}-y_i}{h_i}-\frac{h_i}{2}{m}_i-\frac{h_i}{6}\left(m_{i+1}-m_i\right)\\ c_i=\frac{m_i}{2}\\ d_i=\frac{m_{i+1}-m_i}{6h_i}\end{array}---\left(25\right)$

M _ifor the curve coefficients of batten.

Those of ordinary skill in the art will appreciate that, embodiment described here is to help reader understanding's principle of the present invention, should be understood to that protection scope of the present invention is not limited to so special statement and embodiment.Those of ordinary skill in the art can make various other various concrete distortion and combination of not departing from essence of the present invention according to these technology enlightenment disclosed by the invention, and these distortion and combination are still in protection scope of the present invention.

Claims (6)

1. the pre-stack seismic signal lineups based on global optimization even up a method, it is characterized in that, comprise the following steps:

S1, initialization road collection flattening parameters;

S2, choose benchmark road;

S3, calculating Seed Points amount of movement;

S4, lineups are evened up.

2. pre-stack seismic signal lineups according to claim 1 even up method, it is characterized in that, window size, window amount of movement, search radius and similarity matrix fractile threshold value when in described step S1, road collection flattening parameters comprises prestack road collection.

3. pre-stack seismic signal lineups according to claim 2 even up method, and it is characterized in that, described step S2 comprises step by step following:

S21, calculate the similarity matrix C of any Liang Ge road collection;

S22, based on similarity matrix initialization Attraction Degree matrix and degree of membership matrix;

S23, iteration upgrade Attraction Degree matrix and degree of membership matrix;

S24, calculate and make Attraction Degree matrix and the maximum road collection k of degree of membership matrix sum;

S25, judge whether iterations reaches predetermined number of times, if then enter step S3, otherwise enter step S26;

S26, judge that whether collection k is consistent with result during last iteration, if then enter step S3, otherwise return step S23.

4. pre-stack seismic signal lineups according to claim 3 even up method, and it is characterized in that, described step S3 comprises step by step following:

S31, ask for the maximum similarity Matrix C of Liang Ge road collection _maxand define optimum amount of movement matrix S;

S32, compute matrix C _maxthe value c at corresponding fractile threshold value place _m;

S33, statistics are greater than c _mthe road collection number of value, selects the maximum row of number, the amount of movement of window Seed Points when amount of movement corresponding to this row similarity is current;

The amount of movement of window Seed Points when S34, interpolation obtain all the other;

S35, global optimization is carried out to window Seed Points amount of movement time each.

5. pre-stack seismic signal lineups according to claim 4 even up method, and it is characterized in that, described step S35 comprises step by step following:

The amount of movement matrix that between S351, calculating road collection, similarity is maximum;

The displacement difference sub matrix of S352, calculated level direction and vertical direction;

Whether the displacement difference sub matrix of S353, determined level direction and vertical direction meets constraint condition, if then enter step S4, otherwise selects the amount of movement of similarity suboptimum form new amount of movement matrix and return step S352.

6. pre-stack seismic signal lineups according to claim 4 even up method, it is characterized in that, adopt cubic spline interpolation to realize lineups to the method that amount of movement matrix carries out interpolation and even up in described step S4.