CN102981182B - 2D seismic data all-horizon automatic tracking method based on unsupervised classification - Google Patents

Abstract

The invention provides a 2D seismic data all-horizon automatic tracking method based on unsupervised classification. The method includes reading in 2D seismic data and control horizon data; searching an extreme point of a 2D seismic data waveform; fitting the seismic waveform based on the searched extreme point and according to the Chebyshev polynomials, and taking a fitting coefficient as the eigenvector; and performing eigenvector-based unsupervised classification on the fit seismic waveform to obtain the all-horizon automatic tracking result of the 2D seismic data. The method further includes polishing discontinuous seismic horizon through relevant searching; obtaining a complete horizon line through a horizon segment merging method; and eliminating overlapping through relevant searching. According to the method, a manual intervention mechanism is eliminated and complete 2D seismic data all-horizon automatic tracking is achieved.

Description

Based on the two-dimension earthquake data holostrome position method for automatic tracking of unsupervised segmentation

Technical field

The invention belongs to seismic prospecting data and explain field, specifically, relate to a kind of tracing of horizons method of two-dimension earthquake data.

Background technology

Tracing of horizons (that is, layer position is explained) is the important step that geologic information is explained.Tracing of horizons is exactly analyzed by the geological data that seismic prospecting obtains to obtain the end, ground structure.Seismic horizon is followed the trail of for a long time is all rely on artificial extraction.But there is larger problem in artificial extraction: first, and artificial extraction depends on the working experience for a long time of explanation personnel, and subjective factor impact is very large; Next is that manual interpretation exists very large efficiency, can only make an explanation, cannot realize the tracking to all layers, be usually difficult to for meticulous seismic data analysis (as Seismic Stratigraphic Interpretation) provides basic data position, minority order stratum.Along with continuous research and the improvement of tracing of horizons method, depend on computer implemented tracing of horizons algorithm and constantly propose, efficiency and the effect of tracking all improve to some extent.

Existing tracing of horizons method comprised: (1) P.Alberts etc. proposed a kind of based on artificial neural network tracing of horizons algorithm in 2002, main tracing of horizons pattern-recognition being incorporated into the discontinuous geologic structures such as Cross-fault leveling, then adopts neural network to carry out pattern-recognition; (2) Reda Benbernou etc. adopt fuzzy mehtod to adjudicate on the working foundation of P.Alberts, the automatic formation trace method of composition mixing; (3) M.Aurnhammer etc. proposed a kind of genetic algorithm tracing of horizons algorithm in 2002, its concrete thought is the problem adopting the method based on model to process crossover fault, then problem is converted into the problem of constrained optimization, then uses genetic algorithm to solve; (4) F.Admasu etc. have employed the method for simulated annealing to separate constraint optimal problem in 2004, discuss bayes method solve constrained optimization problem in 2006; (5) wavelet transform introduced in tracing of horizons problem in 2006 by F.Admasu etc., first by geological data Multiscale Wavelet Decomposition, and then carried out tracing of horizons, still adopted bayes method; (6) Yingwei Yu etc. used directional vector field to obtain layer position extreme value information in 2011, destination layer position were modeled as Connected undigraph, then use minimum spanning tree to obtain destination layer position, but thinking can not carry out full tracing of horizons like this; (7) Hilde G.Borgos etc. introduced the holostrome position tracing algorithm based on finite mixtures Gauss in 2005, owing to tracing of horizons to be converted into classification (including supervised classification and semisupervised classification) problem, so there is not Cross-fault leveling problem, complicated geologic media can be applicable to.

But, there is following problem in existing tracing of horizons method: the neural net method of (1) P.Alberts and RedaBenbernou cross over complicated geological environment ability and training sample situation closely related, if training sample comprises this complicated geological situation, then can follow the trail of very well, otherwise can not, but for changeable complicated geological environment, training sample often cannot comprise whole situations, and the acquisition of training sample needs a large amount of manual intervention to indicate; (2) model method of M.Aurnhammer and F.Admasu exists the problem of the approximation ratio of complicated geological situation and the solving precision of method for solving and suboptimal solution problem; (3) Supervised classification mainly contains maximum likelihood and Bayes's classification, the data of a large amount of manual identification are used to train, then its performance is tested by the data do not identified, sorter has very high nicety of grading, it is a lot of that semisupervised classification has needed the sample of mark to lack relative to Supervised classification, but the precision of corresponding classification can decline to some extent, but needs manually to intervene based on the sorting technique of Supervised classification and semisupervised classification, full tracing of horizons automatically can not be realized completely.

Therefore, a kind of two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation is needed.

Summary of the invention

According to an aspect of the present invention, provide a kind of two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation, described method uses unsupervised segmentation to realize the full automatic formation trace of two-dimension earthquake data.

According to a further aspect in the invention, provide a kind of two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation, described method uses associative search or layer bit slice to break fusion to realize layer position polishing or to eliminate overlapping.

According to an aspect of the present invention, provide a kind of two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation, described method comprises: read in two-dimension earthquake data and key-course bit data; Search the extreme point of two-dimension earthquake data waveform; Based on the extreme point searched, carry out matching seismic waveshape, using fitting coefficient as proper vector by Chebyshev polynomials; The seismic waveshape of matching is carried out to the unsupervised segmentation of feature based vector, thus obtain the full automatic formation trace result of two-dimension earthquake data.

Described extreme point can comprise maximum value or minimum value or can search the zero crossing of two-dimension earthquake data waveform and carry out matching seismic waveshape based on the zero crossing searched by Chebyshev polynomials.

The step of the seismic waveshape of matching being carried out to the unsupervised segmentation of feature based vector can comprise: the probability density function according to proper vector sets up statistical model:

$f\left(a\right|c,u,\mathrm{\Σ})=\underset{k\∈\mathrm{\ϵ}}{\mathrm{\Π}}f\left(a_k\right|c_k,u_{c_k},{\mathrm{\Σ}}_{c_k})$

$=\underset{k\∈\mathrm{\ϵ}}{\mathrm{\Π}}{\left(2\mathrm{\π}\right)}^{-n_A/2}{\left|{\mathrm{\Σ}}_{c_k}\right|}^{-1/2}\exp \{{-(a_k-u_{c_k})}^{\′}{{\mathrm{\Σ}}^{-1}}_{c_k}(a_k-u_{c_k})/2\}---\left(3\right),$ Wherein, proper vector A _kmarginal distribution, c _k∈ 1,2 ..., n _cpresentation class class mark, be always divided into n _cclass, represent the average of Gaussian distribution, represent corresponding variance; Maximum likelihood method is adopted to carry out cluster analysis to proper vector:

$\mathrm{ML}=\underset{\mathrm{\θ}}{\max }\left(f\left(A\right|\mathrm{\θ})\right)=\underset{\mathrm{\θ}}{\max }\left(\underset{k\∈\mathrm{\ϵ}}{\mathrm{\Π}}f\left(a_k\right|c_k,u_{c_k},{\mathrm{\Σ}}_{c_k})\right)---\left(4\right),$ Obtain by carrying out differentiate to formula (4):

$\widehat{u_c}=\frac{1}{\left|{\mathrm{\ϵ}}_c\right|}\underset{k\∈{\mathrm{\ϵ}}_c}{\mathrm{\Σ}}{A}_k---\left(5\right),$

$\widehat{{\mathrm{\Σ}}_c}=\frac{1}{\left|{\mathrm{\ϵ}}_c\right|}\underset{k\∈{\mathrm{\ϵ}}_c}{\mathrm{\Σ}}{(A_k-\widehat{u_c})}^2---\left(6\right),$

$\widehat{c_k}=\arg \underset{c_k}{\max }f\left(a_k\right|c_k,u_c,{\mathrm{\Σ}}_c)---\left(7\right),$ Wherein, with be respectively u, the estimated value of ∑ and c; Corresponding estimated value is obtained by the continuous iteration of EM EM algorithm:

$\widehat{u_c}=\frac{1}{\left|\mathrm{\ϵ}\right|}\underset{k\∈\mathrm{\ϵ}}{\mathrm{\Σ}}f\left(c\right|A_k,\widehat{u_c},\widehat{{\mathrm{\Σ}}_c})A_k---\left(8\right)$

$\widehat{{\mathrm{\Σ}}_c}=\frac{1}{\left|\mathrm{\ϵ}\right|}\underset{k\∈\mathrm{\ϵ}}{\mathrm{\Σ}}f\left(c\right|A_k,\widehat{u_c},\widehat{{\mathrm{\Σ}}_c}){(A_k-\widehat{u_c})}^2---\left(9\right),$ Average and the covariance matrix of each cluster group is obtained by formula (8) and (9); Each classification samples point is mapped in each self-corresponding cluster group, be assigned in different types by each extreme point, these types correspond to each layer of bit slice section initially obtaining the layer site of process, thus obtain the full automatic formation trace result of two-dimension earthquake data.

Described method also can comprise: when the layer position obtained after carrying out the unsupervised segmentation of feature based vector to the seismic waveshape of matching exists interruption, carried out the seismic horizon of polishing interruption by associative search.

The step being carried out the seismic horizon that polishing is interrupted by associative search can be comprised: from the layer site of gap, according to the instruction at inclination angle, find down the indication point together, search for the rectangular window of this indication point two ends certain length, related coefficient in window is found to be greater than the extreme point of setting thresholding, and the nearest extreme point of chosen distance adds this layer of site to be concentrated, until there is not the extreme point that related coefficient is greater than thresholding in polishing breach or search window, wherein, described related coefficient represents the amplitude related coefficient of the extreme point in indication point to the vector and window in the layer site of gap to the vector in the layer site of gap.

Described method also can comprise: when related coefficient cannot be found in window to be greater than the extreme point of setting thresholding, obtains complete layer bit line by using layer bit slice section fusion method.

Can be comprised by the step using layer bit slice section fusion method to obtain complete layer bit line: centered by last layer of site of each layer of bit slice section, determine a rectangular window successively, then the initial layers site that whether there is other layer of bit slice section in this window is checked, if exist, these two layer bit slice sections are connected to form new layer bit slice section, in this way constantly detect, until there is not the initial layers site of another one layer bit slice section in the window in all layer bit slice section last layer of sites, finally obtain complete layer bit line.

Described method also can comprise: when the layer position obtained after carrying out the unsupervised segmentation of feature based vector to the seismic waveshape of matching exists overlapping phenomenon, eliminates overlapping phenomenon by searching nearest similar layer site in predetermined window.

The step eliminating overlapping phenomenon by searching nearest similar layer site in predetermined window can comprise: preset a window, from the layer site of the start channel of this class, this layer of site is moved in neighboring track, search same class on this road and the nearest layer site of point of distance translation is divided into this layer of site concentrates, then continue to search down the layer site together, until look for complete layer position, remaining point set is as another cluster group, the i.e. data point of another one layer position.

Method according to the two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation of the present invention eliminates manual intervention mechanism, realizes the full automatic formation trace of two-dimension earthquake data completely.The precision of described method trace horizon is high, speed is fast, efficiently solves the inefficient problem existed in seismic interpretation, can be screen work when sequence stratigraphy study provides geology etc.Described method can be widely used in oil seismic exploration, quality ore deposit and construction work description of locality.

Accompanying drawing explanation

In conjunction with the drawings, from the following describes of embodiment, the present invention these and/or other side and advantage will become clear, and are easier to understand, wherein:

Fig. 1 is the process flow diagram according to the two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation of the present invention;

Fig. 2 is the schematic diagram according to seismic waveshape extreme point of the present invention;

Fig. 3 is according to the schematic diagram being carried out matching seismic waveshape by Chebyshev polynomials of the present invention;

Fig. 4 is that the seismic horizon obtained based on unsupervised segmentation according to the present invention exists the schematic diagram be interrupted;

Fig. 5 is according to the schematic diagram being carried out the seismic horizon of the interruption shown in polishing Fig. 4 by associative search of the present invention;

Fig. 6 is according to the schematic diagram being carried out polishing seismic horizon by layer position segment composition of the present invention;

Fig. 7 is the schematic diagram that overlapping phenomenon appears in the seismic horizon obtained based on unsupervised segmentation according to the present invention;

Fig. 8 is according to the schematic diagram eliminating the overlapping phenomenon shown in Fig. 7 by searching nearest similar layer site in predetermined window of the present invention;

Fig. 9 is the design sketch of the two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation of the application of the invention and the example of the full tracing of horizons obtained;

Figure 10 is the design sketch of the two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation of the application of the invention and another example of the full tracing of horizons obtained.

Embodiment

There is provided the following description of reference accompanying drawing to help the complete understanding to the embodiments of the invention by claim and equivalents thereof.Comprise various specific detail to help to understand, but these details are only considered to be exemplary.Therefore, those of ordinary skill in the art will recognize without departing from the scope and spirit of the present invention, can make various changes and modifications embodiment described herein.In addition, for clarity and brevity, the description to known function and structure is omitted.

Fig. 1 is the process flow diagram according to the two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation of the present invention.

With reference to Fig. 1, in step 101, read in two-dimension earthquake data and key-course bit data.Wherein, two-dimension earthquake data can comprise the initial CDP of seismic section and stop the amplitude of CDP and each degree of depth, and an available two-dimensional array represents.Key-course bit data can be the depth value of key-course under each CDP correspondence of seismic section.

In step 102, search the extreme point of two-dimension earthquake data waveform.

Specifically, because tracing of horizons is normally in the maximum value of two-dimension earthquake data, minimal value and zero crossing face are carried out, and the first step of therefore carrying out tracing of horizons needs to find these maximum value minimal value or zero crossings.The maximum value of two-dimension earthquake data waveform and minimal value are called as earthquake extreme value.As shown in Figure 2, Fig. 2 illustrates the example of seismic waveshape extreme point.Earthquake extreme value can be defined as:

$e\left(x\right)=\{t:\frac{\mathrm{dD}(x,t)}{\mathrm{dt}}=0\}---\left(1\right).$

Wherein, D={D (x, t) } represent seismic section, x is No. CDP or wire size, and t is two-way time or the degree of depth, then D (x ₀, a seismic trace t) can be expressed as.In one embodiment of the invention, main use earthquake maximum value is as the basis of automatic formation trace, but should be appreciated that, the present invention is not limited to the basis using earthquake maximum value as automatic formation trace, in other embodiments of the invention, also can use earthquake minimal value, basis that earthquake zero crossing is used as automatic formation trace.

Referring back to Fig. 1, in step 103, based on the extreme point searched, carry out matching seismic waveshape by Chebyshev polynomials.

Specifically, seismic trace is rebuild by chebyshev approximating polynomial, obtains corresponding fitting coefficient, and the coefficient that the waveform around the extreme point of therefore two-dimension earthquake data waveform can be obtained by matching represents.The maximum value two ends local earthquake waveform obtained by chebyshev approximating polynomial can be expressed as:

S(z)=b ₀p ₀(z)+b ₁p ₁(z)+…+b _Np _N(z) （2）

Wherein, N represents matching exponent number, $\left\{\begin{array}{c}{p}_0\left(z\right)=1\\ p_1\left(z\right)=z\\ p_{n+1}\left(z\right)=2\·z\·p_n\left(z\right)-p_{n-1}\left(z\right)\end{array}\right.$ For Chebyshev polynomials, b _j(j=0,1 .., N) is matching gained coefficient,

$b_0=\frac{1}{N+1}\underset{k=0}{\overset{N}{\mathrm{\Σ}}}f\left(z_k\right)*p_0\left(z_k\right)=\frac{1}{N+1}\underset{k=0}{\overset{N}{\mathrm{\Σ}}}f\left(z_k\right),$

$b_j=\frac{2}{N+1}\underset{k=0}{\overset{N}{\mathrm{\Σ}}}f\left(z_k\right)*p_j\left(z_k\right)=\frac{2}{N+1}\underset{k=0}{\overset{N}{\mathrm{\Σ}}}f\left(z_k\right)*\cos \left(\frac{\mathrm{j\π}(2k+1)}{2N+2}\right)j=\mathrm{1,2},...,N.$

In seismic waveshape fit procedure, S (z) represents certain one the preceding paragraph waveform, and z is the degree of depth, and the t namely in equation (1), S (z) are corresponding amplitude.Rebuild seismic trace by the method for chebyshev approximating polynomial, obtain fitting coefficient, with these fitting coefficient characterization waveforms, and using fitting coefficient as proper vector.As shown in Figure 3, Fig. 3 is illustrated according to the example being carried out matching seismic waveshape by Chebyshev polynomials of the present invention.

Referring back to Fig. 1, in step 104, the seismic waveshape of matching is carried out to the unsupervised segmentation of feature based vector, thus obtain the full automatic formation trace result of two-dimension earthquake data.

Specifically, because the waveform around the extreme point on identical layer position is usually closely similar, therefore can by proper vector be divided into different classes, the cluster group obtained.Extreme point corresponding to the proper vector in cluster group represents in same class, each cluster group's expression layer bit slice section.

To the classification of proper vector be the statistical model being based upon proper vector basis on.The joint distribution of proper vector is considered to obey Gaussian Mixture distribution.Therefore, statistical model is set up according to the probability density function of these independently proper vectors:

$f\left(a\right|c,u,\mathrm{\Σ})=\underset{k\∈\mathrm{\ϵ}}{\mathrm{\Π}}f\left(a_k\right|c_k,u_{c_k},{\mathrm{\Σ}}_{c_k})$

$=\underset{k\∈\mathrm{\ϵ}}{\mathrm{\Π}}{\left(2\mathrm{\π}\right)}^{-n_A/2}{\left|{\mathrm{\Σ}}_{c_k}\right|}^{-1/2}\exp \{{-(a_k-u_{c_k})}^{\′}{{\mathrm{\Σ}}^{-1}}_{c_k}(a_k-u_{c_k})/2\}---\left(3\right),$

Wherein, proper vector A _kmarginal distribution, c _k∈ 1,2 ..., n _cpresentation class class mark, be always divided into n _cclass, represent the average of Gaussian distribution, represent corresponding variance.

Subsequently, maximum likelihood method can be adopted to carry out cluster analysis to proper vector:

$\mathrm{ML}=\underset{\mathrm{\θ}}{\max }\left(f\left(A\right|\mathrm{\θ})\right)=\underset{\mathrm{\θ}}{\max }\left(\underset{k\∈\mathrm{\ϵ}}{\mathrm{\Π}}f\left(a_k\right|c_k,u_{c_k},{\mathrm{\Σ}}_{c_k})\right)---\left(4\right),$

Can obtain by carrying out differentiate to formula (4):

$\widehat{u_c}=\frac{1}{\left|{\mathrm{\ϵ}}_c\right|}\underset{k\∈{\mathrm{\ϵ}}_c}{\mathrm{\Σ}}{A}_k---\left(5\right),$

$\widehat{{\mathrm{\Σ}}_c}=\frac{1}{\left|{\mathrm{\ϵ}}_c\right|}\underset{k\∈{\mathrm{\ϵ}}_c}{\mathrm{\Σ}}{(A_k-\widehat{u_c})}^2---\left(6\right),$

$\widehat{c_k}=\arg \underset{c_k}{\max }f\left(a_k\right|c_k,u_c,{\mathrm{\Σ}}_c)---\left(7\right),$

Due to with being respectively u, the estimated value of ∑ and c, and being mutually related each other, therefore need by EM(greatest hope) the continuous iteration of algorithm obtains corresponding estimated value:

$\widehat{u_c}=\frac{1}{\left|\mathrm{\ϵ}\right|}\underset{k\∈\mathrm{\ϵ}}{\mathrm{\Σ}}f\left(c\right|A_k,\widehat{u_c},\widehat{{\mathrm{\Σ}}_c})A_k---\left(8\right)$

$\widehat{{\mathrm{\Σ}}_c}=\frac{1}{\left|\mathrm{\ϵ}\right|}\underset{k\∈\mathrm{\ϵ}}{\mathrm{\Σ}}f\left(c\right|A_k,\widehat{u_c},\widehat{{\mathrm{\Σ}}_c}){(A_k-\widehat{u_c})}^2---\left(9\right)$

Average and the relevant information such as covariance matrix of each cluster group can be obtained after classification, then each classification samples point is mapped in each self-corresponding cluster group, be assigned in different types by each extreme point, these types then correspond to each layer of bit slice section initially obtaining the layer site of process, thus the full automatic formation trace result of two-dimension earthquake data can be obtained.

The unsupervised segmentation of feature based vector and point corresponding to the cluster group that obtains represents a seismic horizon in theory, but in fact may be large due to region span, waveform possibility similarity near two points of certain geographical location interval same layer position is not high, therefore can be assigned to different cluster group the insides in the unsupervised segmentation of feature based vector, the layer position obtained like this exists is interrupted.As shown in Figure 4, Fig. 4 illustrates that the seismic horizon obtained based on unsupervised segmentation according to the present invention exists the example of being interrupted.

In this case, need will occur the layer position polishing be interrupted according to the mode of associative search of the present invention.Therefore, the two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation according to the present invention also can comprise: in step 105, is carried out the seismic horizon of polishing interruption by associative search.Below, specifically describe according to the method being carried out the seismic horizon that polishing is interrupted by associative search of the present invention with reference to Fig. 5.

Fig. 5 is according to the schematic diagram being carried out the seismic horizon of the interruption shown in polishing Fig. 4 by associative search of the present invention.With reference to Fig. 5, for the disconnection problem of same layer position, from the layer site E of gap, according to the instruction at inclination angle, find down the indication point A together, the rectangular window of search A two ends certain length d, related coefficient in window is found to be greater than setting thresholding (such as, Cov) extreme point, and the nearest extreme point of chosen distance (such as, A2) add this layer of site to concentrate, until there is not the extreme point that related coefficient is greater than thresholding in polishing breach or search window, wherein, described related coefficient represents the amplitude related coefficient of the extreme point in indication point A to the vector and window of the layer site E of gap to the vector of the layer site E of gap, such as, vector (A, E) with vector (A1, E) amplitude related coefficient, vector (A, E) with vector (A2, E) amplitude related coefficient etc.

If related coefficient cannot be found in window to be greater than the extreme point of setting thresholding by step 105, then cannot be carried out the seismic horizon of polishing interruption by described associative search, now, still there will be the layer bit slice section of interruption.In this case, layer bit slice section fusion method according to the present invention can be adopted to obtain complete layer bit line.Therefore, the two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation according to the present invention also can comprise: in step 106, carry out polishing seismic horizon by layer position segment composition.Below, specifically describe according to layer bit slice section fusion method of the present invention with reference to Fig. 6.

Fig. 6 is according to the schematic diagram being carried out polishing seismic horizon by layer position segment composition of the present invention.With reference to Fig. 6, certain one deck position has three seismic trace T _x-1, T _xand T _x+1.This layer of position is at T _x-1and T _x+1road exists layer site A and A2, but at T _xextreme point by related coefficient cannot be found during step 105 in window to be greater than setting thresholding on road.In this case, the rectangular window that a length of side is Dw is determined successively centered by last layer of site of each layer of bit slice section, then check the initial layers site that whether there is other layer of bit slice section in this window, if exist, these two layer bit slice sections are connected to form new layer bit slice section.In this way constantly detect, until there is not the initial layers site of another one layer bit slice section in the window in all layer bit slice section last layer of sites, finally obtain complete layer bit line.

In addition, because the unsupervised segmentation of proper vector carries out based on similarity, so be difficult to avoid belonging to two different layers positions in the result of unsupervised segmentation but the very similar layer site of waveform is divided in same class.As shown in Figure 7, Fig. 7 illustrates that the example of overlapping phenomenon appears in the seismic horizon obtained based on unsupervised segmentation according to the present invention.

In this case, the mode by searching nearest similar layer site in predetermined window according to the present invention is needed to eliminate overlapping phenomenon.Therefore, the two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation according to the present invention also can comprise: in step 107, eliminates overlapping phenomenon by searching nearest similar layer site in predetermined window.Below, specifically describe according to the method eliminating overlapping phenomenon by searching nearest similar layer site in predetermined window of the present invention with reference to Fig. 8.

Fig. 8 is according to the schematic diagram eliminating the overlapping phenomenon shown in Fig. 7 by searching nearest similar layer site in predetermined window of the present invention.With reference to Fig. 8, different layers position is divided in same class and occurs overlapping problem, a window can be set, from the layer site S of the start channel of this class, this point is moved to S ' in neighboring track, search same class on this road and the nearest layer site S1 of distance S ' is divided into this layer of site concentrates, then continue to search down the layer site (namely searching down the layer site together along S → S1 direction) together, until look for complete layer position, remaining point set is as another cluster group, the i.e. data point of another one layer position.Remove the overlapping phenomenon in all clusters group according to the method successively.

Fig. 9 is the design sketch of the two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation of the application of the invention and the example of the full tracing of horizons obtained.Figure 10 is the design sketch of the two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation of the application of the invention and another example of the full tracing of horizons obtained.Logical carry out comparing result with the layer position of manual trace and show, the layer position of automatic tracing and the layer position of manual interpretation basically identical, demonstrate the validity of described method.

The invention provides a kind of two-dimension earthquake data holostrome position method for automatic tracking based on unsupervised segmentation.Described method uses unsupervised segmentation to realize the full automatic formation trace of two-dimension earthquake data.In addition, described method also uses associative search or layer bit slice to break fusion to realize layer position polishing or to eliminate overlapping.Described method eliminates manual intervention mechanism, realizes the full automatic formation trace of two-dimension earthquake data completely.The precision of described method trace horizon is high, speed is fast, efficiently solves the inefficient problem existed in seismic interpretation, can be screen work when sequence stratigraphy study provides geology etc.Described method can be widely used in oil seismic exploration, quality ore deposit and construction work description of locality.

Can perform according to said method of the present invention according to computer program instructions.Because these programmed instruction can be included in computing machine, application specific processor or able to programme or specialized hardware, the instruction therefore performed wherein can be conducive to the execution of above-mentioned function.As understood by those skilled in the art, computing machine, processor or programmable hardware comprise the memory device that can store or receive software or computer code, and described software or computer code are in the method described in the present invention by realization when computing machine, processor or hardware access and execution.

Although the present invention is shown with reference to its exemplary embodiment and describes, but it should be appreciated by those skilled in the art, when not departing from the spirit and scope of the present invention by claim and equivalents thereof, various change can be carried out to its form and details.

Claims (4)

1., based on a two-dimension earthquake data holostrome position method for automatic tracking for unsupervised segmentation, described method comprises:

Read in two-dimension earthquake data and key-course bit data;

Search the extreme point of two-dimension earthquake data waveform;

Based on the extreme point searched, carry out matching seismic waveshape, using fitting coefficient as proper vector by Chebyshev polynomials;

The seismic waveshape of matching is carried out to the unsupervised segmentation of feature based vector, thus obtain the full automatic formation trace result of two-dimension earthquake data;

Wherein, also comprise:

When the layer position obtained after carrying out the unsupervised segmentation of feature based vector to the seismic waveshape of matching exists interruption, carried out the seismic horizon of polishing interruption by associative search;

Wherein, the step being carried out the seismic horizon of polishing interruption by associative search is comprised:

From the layer site of gap, according to the instruction at inclination angle, find down the indication point together, search for the rectangular window of this indication point two ends certain length, related coefficient in window is found to be greater than the extreme point of setting thresholding, and the nearest extreme point of chosen distance adds this layer of site to be concentrated, until there is not the extreme point that related coefficient is greater than thresholding in polishing breach or search window, wherein, described related coefficient represents the amplitude related coefficient of the extreme point in indication point to the vector and window in the layer site of gap to the vector in the layer site of gap;

Wherein, also comprise:

When related coefficient cannot be found in window to be greater than the extreme point of setting thresholding, obtain complete layer bit line by using layer bit slice section fusion method;

Wherein, the step by using layer bit slice section fusion method to obtain complete layer bit line comprises:

A rectangular window is determined successively centered by last layer of site of each layer of bit slice section, check the initial layers site that whether there is other layer of bit slice section in this window, if exist, these two layer bit slice sections are connected to form new layer bit slice section, in this way constantly detect, until there is not the initial layers site of another one layer bit slice section in the window in all layer bit slice section last layer of sites, finally obtain complete layer bit line.

2. the method for claim 1, wherein described extreme point comprises maximum value or minimum value or searches the zero crossing of two-dimension earthquake data waveform and carry out matching seismic waveshape based on the zero crossing searched by Chebyshev polynomials.

3. the method for claim 1, also comprises:

When the layer position obtained after carrying out the unsupervised segmentation of feature based vector to the seismic waveshape of matching exists overlapping phenomenon, eliminate overlapping phenomenon by searching nearest similar layer site in predetermined window.

4. method as claimed in claim 3, wherein, the step eliminating overlapping phenomenon by searching nearest similar layer site in predetermined window comprises:

Preset a window, from the layer site of the start channel of this class, this layer of site is moved in neighboring track, search same class on this road and the nearest layer site of point of distance translation is divided into this layer of site concentrates, continue to search down the layer site together, until look for complete layer position, remaining point set is as another cluster group, the i.e. data point of another one layer position.