Summary of the invention
It is an object of the invention to provide a kind of method obtaining Value of residual static correction, it can improve residueThe tempo of evolution of static correction value solution procedure and effect.
The present invention provides a kind of method obtaining Value of residual static correction, including: being total to after a) reading in dynamic(al) correctionCentral point road collection data, add up number and the number of geophone station of shot point in described data；B) according to big gunThe number of point, the number of geophone station, the hunting zone of default Value of residual static correction generate initial disaggregation；C)Utilize Poisson dish sampling algorithm to obtain, according to initial disaggregation, disaggregation of sampling, sampling disaggregation is passed through real-valued codingIt is mapped in heredity space and constitutes initial population；D) initial population carrying out genetic manipulation, acquisition meets pre-The individuality of fixed condition is as Value of residual static correction.
Alternatively, step d) includes: D1) calculate the fitness value of each individuality in population, will adapt toThe highest individuality of angle value is as the optimum individual of current iteration；D2) fitness value in population is retained the highestThe individuality of predetermined quantity, is selected by the mode of roulette remaining individuality；D3) after for selectingPopulation, continue to retain the individuality of the highest predetermined quantity of described fitness value, remaining individuality carried out2 exchanges；D4) for the population after exchange, continue to retain the predetermined quantity that described fitness value is the highestIndividuality, remaining individuality is carried out 2 variations；D5) determine whether current iteration reaches default changingWhether generation number and the difference determining the optimum individual of current iteration and the optimum individual of last iteration are less than by mistakeDifference limen value；Wherein, when the iterations and the optimum of current iteration determining that current iteration is not reaching to presetThe difference of the individual optimum individual with last iteration, more than or equal to error threshold, returns and performs step D1)；D6) when determining that current iteration reaches default iterations or the optimum individual of current iteration and last iterationOptimum individual difference less than error threshold time, using the optimum individual of current iteration as residual static correctionAmount.
Alternatively, the step utilizing Poisson dish sampling algorithm to obtain sampling disaggregation according to initial disaggregation includes:C1) equation below is utilized to calculate size threshold G,Wherein, NSFor big gunThe number of point, NrFor the number of geophone station, γiFor the disaggregation that formed after the initial disaggregation of i & lt disturbanceBig Norm Solution, κiThe least-norm solution of the disaggregation for being formed after the initial disaggregation of i & lt disturbance, NcFor initiallySolve the number concentrating solution vector；C2) the initial disaggregation of disturbance forms the n-th disaggregation, and wherein, n is greater thanIt is 1 in the integer of 1 and the initial value of n；C3) equation below is utilized to calculate scale value G' of the n-th disaggregation,Wherein, rp,i∈Rp, rq,i∈Rq, Rp,RqIt it is the n-th disaggregationSolution vector；C4) when scale value G' of the n-th disaggregation is less than or equal to G so that n=n+1, disturbance (n-1)thDisaggregation forms the n-th disaggregation, returns step C3)；C5) when scale value G' of the n-th disaggregation is more than G,Using the n-th disaggregation as sampling disaggregation.
Alternatively, step D4) including: D41) for individuality remaining in population, generate each individualityRandom value m, will there is the individuality less than the random value m of the mutation probability preset as to be made a variationBody, wherein, m ∈ [0,1]；D42) randomly generate two variable position, by the institute in individuality to be made a variationState the gene at two variable position and become the random value in preset range at random, wherein, utilize followingFormula obtains preset range, [Rj,k-c,Rj,k+ c] ∩ [a, b], wherein, Rj,kRepresent the kth that jth is individualIndividual variable position, [a, b] is the hunting zone of default Value of residual static correction,L isThe constant preset, E0For the meansigma methods of the fitness value of all individualities in initial population, EiterFor thisThe fitness value of the optimum individual of iteration.
Alternatively, predetermined quantity is 2.
According to the method obtaining Value of residual static correction of the present invention, on the one hand Poisson dish sampling algorithm is utilized to replaceThe method obtaining initial population for the stochastical sampling in traditional genetic algorithm, increases the equal of initial population distributionEven degree, thus increase evolutionary rate and seek stability of solution.On the other hand, introduce elite and retain planSlightly, population will not carry out genetic manipulation and is directly entered down by optimum a number of individuality as eliteA generation, improves defect individual effect in population, maximally utilises defect individual in populationGene information, effectively prevents defect individual to be destroyed, improves Evolution of Population speed.
By aspect other for part elaboration present general inventive concept in following description and/or advantage, alsoSome be will be apparent from by description, or can learn through the enforcement of present general inventive concept.
Detailed description of the invention
Being described in detail the embodiment of the present invention now, its example represents in the accompanying drawings.Below byIt is described to explain the present invention to embodiment referring to the drawings.
Fig. 1 illustrates the flow chart of the method obtaining Value of residual static correction according to an embodiment of the invention.
As it is shown in figure 1, in step 101, read in the common midpoint gather data after dynamic(al) correction, add up instituteState number and the number of geophone station of shot point in data.
Existing algorithm can be used to add up the number of the shot point in the common midpoint gather data after dynamic(al) correctionWith the number of geophone station, therefore do not repeat them here.
In step 102, according to the number of shot point, the number of geophone station, default Value of residual static correctionHunting zone generates initial disaggregation.
In one example, initial solution collection can be generated according to equation below (1):
In formula (1), RiFor the i-th solution vector of initial disaggregation, ri,jObey being uniformly distributed between [a, b],Represent the Value of residual static correction of shot point,Represent the Value of residual static correction of geophone station,NcThe number of solution vector, N is concentrated for the number of individuals in default population, i.e. initial solutionsFor the number of shot point,NrFor the number of geophone station, [a, b] is the hunting zone of default Value of residual static correction, and unit is millisecond.
In step 103, Poisson dish sampling algorithm is utilized to obtain, according to initial disaggregation, disaggregation of sampling, will samplingDisaggregation constitutes initial population by real-valued coding mapping in heredity space.
Specifically, utilizing Poisson dish sampling algorithm to obtain, according to initial disaggregation, disaggregation of sampling, obtain adoptsThe solution vector of sample disaggregation spreads all over the uniformity of solution space and is improved, and then, passes through real by sampling disaggregationValue coding mapping constitutes initial population in heredity space, therefore enhances the uniform journey of initial population distributionDegree.
Initial population matrix such as below equation (2):
In formula (2),The solution vector concentrated is solved, i.e. the individuality in population, N for samplingcForNumber of individuals in the population preset, NsFor the number of shot point, NrNumber for geophone station.
In one embodiment, the method shown in Fig. 2 is utilized to utilize Poisson dish sampling algorithm according to initial solutionCollection obtains disaggregation of sampling.
In step 104, initial population is carried out genetic manipulation, obtain the individual conduct conformed to a predetermined conditionValue of residual static correction.
Specifically, available existing genetic algorithm obtain based on initial population conform to a predetermined conditionBody is as Value of residual static correction.The individuality conformed to a predetermined condition can be when this iteration reaches default changingOptimum individual during generation number, in this iteration.The individuality conformed to a predetermined condition can also be when this time changesWhen the difference of the optimum individual that the optimum individual that generation obtains obtains with last iteration is less than error threshold value, shouldOptimum individual in secondary iteration.It is not limited to the above-mentioned condition enumerated it should be understood that pre-conditioned.
In a preferred embodiment, utilize the method shown in Fig. 3 that initial population is carried out genetic manipulation,Obtain the individuality conformed to a predetermined condition as Value of residual static correction.
Fig. 2 illustrates and utilizes Poisson dish sampling algorithm to obtain according to initial disaggregation according to an embodiment of the inventionThe flow chart of the method for sampling disaggregation.The method shown in Fig. 2 can be performed when performing step 103.
As in figure 2 it is shown, in step 201, utilize equation below (3) to calculate size threshold G,
In formula (3), NsFor the number of shot point, NrFor the number of geophone station, γiInitial for i & lt disturbanceThe maximum norm solution of the disaggregation formed after disaggregation, κiFor the disaggregation that formed after the initial disaggregation of i & lt disturbanceLittle Norm Solution, NcThe number of solution vector is concentrated for the number of individuals in default population, i.e. initial solution.
It should be understood that as i=1, the initial disaggregation of i & lt disturbance is that initial disaggregation is carried out disturbance, works as iWhen >=2, the initial disaggregation of i & lt disturbance is to disturb on the basis of the disaggregation of generation after the i-th-1 time disturbanceDynamic.
In step 202, the initial disaggregation of disturbance forms the n-th disaggregation, and wherein, n is greater than the integer equal to 1And the initial value of n is 1.
In step 203, equation below (4) is utilized to calculate scale value G' of the n-th disaggregation,
In equation (4), rp,i∈Rp, rq,i∈Rq, Rp,RqIt it is the solution vector of the n-th disaggregation.
In step 204, determine that whether scale value G' of the n-th disaggregation is more than G.
When scale value G' determining the n-th disaggregation in step 204 is not more than G, in step 205 so thatN=n+1, disturbance the (n-1)th disaggregation forms the n-th disaggregation, returns step 203.
When scale value G' determining the n-th disaggregation in step 204 is more than G, in step 206, by n-thDisaggregation is as sampling disaggregation.
Fig. 3 illustrates and initial population carries out genetic manipulation according to an embodiment of the invention, and acquisition meets pre-The individuality of fixed condition is as the flow chart of the method for Value of residual static correction.Can perform when performing step 104Method shown in Fig. 3.
As it is shown on figure 3, in step 301, calculate the fitness value of each individuality in population, by fitnessIt is worth the highest individuality optimum individual as current iteration.
That is, each individual corresponding target function value in population is calculated.
In one embodiment, object function uses the function calculating stacked section energy, can be according to followingEquation (5) calculate the fitness value of each individuality in population,
In formula (5), E (Ri) it is the individual stacked section energy of i-th, dyhRepresent y-th dynamic(al) correctionAfter common midpoint gather in h track data, t is in the common midpoint gather after y-th dynamic(al) correctionThe sampling time of h track data, nsFor the sequence number of shot point corresponding to i-th individuality, nrIndividual for i-thThe sequence number of corresponding geophone station,
It should be understood that the fitness value of each individuality in first time iterative computation population, i.e. calculate original speciesThe fitness value of each individuality in Qun, for the second time each individuality in iteration and later iterative computation populationThe adaptation of each individuality in the population that fitness value is formed after i.e. calculating the last iteration of this iterationAngle value.
It should be understood that object function is used as existing object function, it is not limited in above-described embodimentObject function.
In step 302, retain the individuality of the predetermined quantity that fitness value is the highest in population, to remainingBody is selected by the mode of roulette.
Specifically, for individuality remaining in population, each individuality is calculated according to equation below (6)Select probability,
In equation (6), P (Ri) it is the individual select probability of i-th, E is object function, and i, p, q areIndividual sequence number, NcFor the number of individuals in population.
Then, the random value χ (χ ∈ [0,1]) that i-th is individual is generated, if P is (Ri-1)≤χ≤P(Ri), then selectI-th parent individuality enters filial generation.
Should be appreciated that, it is possible to use the predetermined quantity that existing selection algorithm is the highest to fitness value in populationIndividuality beyond individuality select, be not limited to the mode of roulette.
In step 303, for the population after selecting, continue to retain the predetermined number that described fitness value is the highestThe individuality of amount, carries out 2 exchanges to remaining individuality.
Specifically, for individuality remaining in population, generate the random value ζ (ζ ∈ [0,1]) of each individuality,To have the individuality less than the random value ξ exchanging probability preset as individuality to be exchanged, remaining is madeFor non-individuality to be exchanged.Randomly generate two exchange positions, described in two individualities to be exchangedGene section between two exchange positions swaps, the most individual to produce filial generation.For non-to be exchangedIndividuality, parent individuality is directly entered filial generation as offspring individual.
Should be appreciated that, it is possible to use existing exchange algorithm is the highest to fitness value in the population after selectingIndividuality beyond the individuality of predetermined quantity swaps, and is not limited to 2 exchanges.
In step 304, for the population after exchange, continue to retain the predetermined number that described fitness value is the highestThe individuality of amount, carries out 2 variations to remaining individuality.
In a preferred embodiment, for individuality remaining in population, the random of each individuality is generatedValue m(m ∈ [0,1]), the individuality with the random value m less than the mutation probability preset is made a variation as waitingIndividuality.
Then, randomly generate two variable position, by the said two variable position in individuality to be made a variationThe gene at place becomes the random value in preset range at random, and wherein, available equation below (7) obtainsPreset range,
[Rj,k-c,Rj,k+ c] ∩ [a, b] (7)
In formula (7), Rj,kRepresenting the kth variable position that jth is individual, [a, b] is default remainingThe hunting zone of remaining static correction value,L is default constant, E0Institute for initial populationThere are the meansigma methods of the fitness value of individuality, EiterFitness value for the optimum individual of current iteration.Therefore,Along with the evolution of the increase of iterations, i.e. algorithm, the said two variable position in individuality to be made a variationWhen the gene at place becomes the random value in preset range at random, preset range is gradually reduced, and improves understandingStability.
Should be appreciated that, it is possible to use existing mutation algorithm is the highest to fitness value in the population after exchangeIndividuality beyond the individuality of predetermined quantity makes a variation, and is not limited to 2 variations.
In step 305, determine whether current iteration reaches default iterations and determine current iterationWhether optimum individual is less than error threshold with the difference of the optimum individual of last iteration.
Wherein, when the iterations and the optimum individual of current iteration determining that current iteration is not reaching to presetStep 301 is performed more than or equal to error threshold, return with the difference of the optimum individual of last iteration.
When determining that current iteration reaches the optimum individual of default iterations or current iteration in step 305When being less than error threshold with the difference of the optimum individual of last iteration, in step 306, by current iterationOptimum individual is as Value of residual static correction.
Additionally, may be implemented as computer program according to the said method of the exemplary embodiment of the present invention,Thus when running this program, it is achieved said method.
According to the method obtaining Value of residual static correction of the present invention, on the one hand Poisson dish sampling algorithm is utilized to replaceThe method obtaining initial population for the stochastical sampling in traditional genetic algorithm, increases the equal of initial population distributionEven degree, thus increase evolutionary rate and seek stability of solution.On the other hand, introduce elite and retain planSlightly, population will not carry out genetic manipulation and is directly entered down by optimum a number of individuality as eliteA generation, improves defect individual effect in population, maximally utilises defect individual in populationGene information, effectively prevents defect individual to be destroyed, improves Evolution of Population speed.
Although having show and described some embodiments of the present invention, it will be understood by those skilled in the art thatWithout departing from the principle of the present invention and the situation of spirit being limited its scope by claim and equivalent thereofUnder, these embodiments can be modified.