Simulated annealing inversion solving method, device and equipment for terrain and medium

Abstract

The invention discloses a terrain simulated annealing inversion solving method, device and equipment and a medium, through innovative combination of multi-dimensional initial search and a genetic algorithm optimization mechanism, dual improvement of prediction precision and efficiency is realized, and the terrain simulated annealing inversion solving method has remarkable technical advancement and wide application prospects. The method has the beneficial effects that the prediction performance is remarkably improved, the multi-dimensional initial search and a genetic algorithm optimization mechanism are fused, so that the scheme shows excellent performance in complex and variable prediction tasks, and the prediction precision and stability are greatly improved. And calculation efficiency optimization: a dynamic mutation probability adjustment mechanism effectively reduces unnecessary calculation overhead and accelerates a convergence process, so that the scheme has relatively high calculation efficiency while maintaining high-precision prediction. The embodiment of the invention can efficiently and accurately carry out simulated annealing inversion solution of the terrain, and can be widely applied to the technical field of data processing.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a simulated annealing inversion solution method, device, equipment and medium for terrain. Background Art

[0002] The Parker-Oldenburg inversion method occupies a core position in the study of crustal structure, but its technical limitations and shortcomings are also significant, mainly as follows:

[0003] First, the primary flaw is that traditional technical methods adopt simplified linear assumptions when dealing with the key parameters of the inversion process, namely density difference and depth, that is, these parameters are regarded as constant values. This treatment method ignores the complex nonlinear interactions of underground media, which in turn limits the accuracy of the seafloor topography data generated by the inversion, making it difficult to accurately reproduce the details of the real geological structure.

[0004] Second, when searching for the optimal parameters for Parker-Oldenburg inversion, the current mainstream approach relies on genetic algorithms. However, the crossover and mutation mechanism designs in these algorithms do not fully match the unique nature of the inversion problem, resulting in an inefficient solution process and significant fluctuations in solution accuracy under different circumstances, making it difficult to stably achieve high-quality inversion results.

[0005] Third, from the perspective of computing performance and automated operation, existing technical solutions also face severe challenges. Since the entire computing process involves a huge amount of data processing and lacks an efficient and integrated execution platform to optimize and coordinate the computing process, the automation process of terrain inversion is difficult. This not only increases the cost of manual operation, but also weakens the accuracy and practicality of the inversion results, resulting in a relatively rough terrain model that is difficult to meet the needs of high-precision geological exploration and research. Summary of the invention

[0006] The present invention aims to solve the problem of related technical limitations at least to a certain extent. To this end, the present invention proposes a simulated annealing inversion solution method, device, equipment and medium for terrain, which can efficiently and accurately perform simulated annealing inversion solution for terrain.

[0007] On the one hand, an embodiment of the present invention provides a simulated annealing inversion solution method for terrain, comprising:

[0008] Initializing a vector set, a first fitness value and an initial temperature; using the initial temperature as an annealing temperature; wherein the vector set includes a preset number of prediction vectors;

[0009] Traverse all prediction vectors in the vector set to perform inversion solution and fitness evaluation in the prediction area, and obtain the fitness function value of each prediction vector for inversion solution in the prediction area;

[0010] The result of the inversion solution is the predicted terrain data of the predicted area; the value of the fitness function is negatively correlated with the adaptability of the prediction vector for the inversion solution of the predicted area;

[0011] When there is a prediction vector in the vector set whose corresponding fitness function value is less than the iterative optimal threshold, the corresponding prediction vector is output as the optimal prediction vector; otherwise,

[0012] The excellence of each prediction vector is determined according to the fitness function value, and based on the excellence, a random probability selection operation is performed on the prediction vectors in the vector set, and the vector set is updated according to the selection result of the selection operation; and then the minimum fitness function value corresponding to the vector set is used as the second fitness value; the excellence is negatively correlated with the numerical value of the fitness function value;

[0013] When the second fitness value is greater than or equal to the first fitness value, based on the annealing temperature, a random probability mutation operation is performed on the prediction vector in the vector set, and the vector set is updated according to the mutation result of the mutation operation; otherwise, the first fitness value is updated by the second fitness value;

[0014] Based on the second fitness value, a crossover operation with random probability is performed on the prediction vectors in the vector set, and the vector set is updated according to the crossover result of the crossover operation;

[0015] Performing temperature decay processing on the annealing temperature based on a preset temperature decay rate, and updating the annealing temperature according to the decay result of the temperature decay processing;

[0016] When the annealing temperature is greater than or equal to the preset termination temperature, the process returns to execute the steps of traversing all prediction vectors in the vector set to perform inversion solution and fitness evaluation of the prediction area until the annealing temperature is less than the termination temperature, and the optimal prediction vector is output according to the last updated vector set;

[0017] The predicted terrain data obtained by inversion according to the optimal prediction vector is used as the terrain solution result of the prediction area.

[0018] In some embodiments, initializing the vector set comprises at least one of the following steps:

[0019] Obtaining a preset standard vector value, and randomly generating a preset number of prediction vectors near the standard vector value;

[0020] A preset number of prediction vectors are randomly generated within a target interval; the target interval represents a value range of the prediction vectors;

[0021] A preset number of prediction vectors are directly obtained in response to a manual setting instruction.

[0022] In some embodiments, traversing all prediction vectors in the vector set to perform inversion solution and fitness evaluation on the prediction area, and obtaining the fitness function value of each prediction vector for inversion solution of the prediction area, includes the following steps:

[0023] The prediction vectors in the vector set are traversed in sequence, and the prediction terrain data of the prediction area corresponding to each prediction vector is obtained by inversion solution using the Parker-Oldenburg inversion method;

[0024] Based on the predicted terrain data, the root mean square calculation is performed in combination with the seismic depth matrix of the predicted area, and the root mean square value is obtained as the fitness function value of the corresponding prediction vector for the adaptability of the inversion solution in the predicted area.

[0025] In some embodiments, outputting the corresponding prediction vector as the optimal prediction vector includes the following steps:

[0026] When there is a prediction vector in the vector set whose corresponding fitness function value is less than the iterative optimal threshold, the prediction vector is directly output as the optimal prediction vector;

[0027] When there are multiple prediction vectors in the vector set whose corresponding fitness function values ​​are less than the iterative optimal threshold, the prediction vector corresponding to the smallest fitness function value is output as the optimal prediction vector.

[0028] In some embodiments, based on the degree of excellence, a random probability selection operation is performed on the prediction vector in the vector set, including the following steps:

[0029] Normalize the excellence levels corresponding to all prediction vectors in the vector set to obtain the normalized excellence level corresponding to each prediction vector;

[0030] The expression of excellence is:

[0031]

[0032] In the formula, e represents the natural constant; RMS represents the fitness function value; minIterDelt represents the iterative optimal threshold; nowTemp represents the annealing temperature;

[0033] Based on the normalized excellence levels corresponding to all prediction vectors in the vector set, an excellence level array is obtained;

[0034] The value of the nth element in the excellence array is the cumulative sum of all normalized excellences corresponding to the first prediction vector to the nth prediction vector in the vector set;

[0035] Generate a random number in the first interval as the selection probability;

[0036] Sequentially traverse the excellence array to find the first element whose probability is greater than the selection probability as the target element;

[0037] Selecting a corresponding prediction vector from the vector set as an update vector according to the element number of the target element, and returning to execute the step of generating a random number in the first interval as a selection probability until a preset number of update vectors are selected from the vector set;

[0038] A preset number of update vectors are used as selection results of the selection operation.

[0039] In some embodiments, the prediction vector includes two types of inversion parameters: density difference and depth; based on the annealing temperature, a random probability mutation operation is performed on the prediction vector in the vector set, including the following steps:

[0040] Based on the annealing temperature, the mutation probability is calculated by combining the first fitness value and the second fitness value;

[0041] Among them, the expression of mutation probability is:

[0042]

[0043] In the formula, mutationRate represents the mutation probability; e represents the natural constant; RMS1 represents the first fitness value; RMS2 represents the second fitness value; nowTemp represents the annealing temperature;

[0044] Generate a random number in the second interval as the mutation determination probability; add 1 to the number of mutation cycles;

[0045] When the mutation determination probability is less than the preset mutation probability, a prediction vector is randomly selected from the vector set as the mutation prediction vector;

[0046] Perform numerical mutation on any inversion parameter in the mutated prediction vector within the target interval, and put the prediction vector after numerical mutation back into the vector set; the target interval represents the value range of the prediction vector;

[0047] Return to the step of generating a random number in the second interval as a mutation determination probability until the number of mutation cycles is equal to a preset number, and obtain a mutation result of the mutation operation;

[0048] Among them, the number of mutation cycles is initialized to 0.

[0049] In some embodiments, the prediction vector includes two types of inversion parameters: density difference and depth; based on the second fitness value, performing a crossover operation with random probability on the prediction vector in the vector set includes the following steps:

[0050] Based on the second fitness value, the number of crossovers is calculated by combining the iterative optimal threshold and the preset number;

[0051] The expression of the number of crossovers is:

[0052]

[0053] In the formula, crossCnt represents the number of crossovers; random(x) represents a random value generated in [0,x]; RMS2 represents the second fitness value; minIterDelt represents the iterative optimal threshold; MaxPredCnt represents the preset number;

[0054] Generate a random number in the third interval as the crossover determination probability; add 1 to the number of crossover cycles;

[0055] When the crossover determination probability is less than the preset crossover operator probability, two prediction vectors are randomly obtained from the vector set as crossover vectors; the crossover vectors include the first vector and the second vector;

[0056] Performing numerical weighted crossover on at least one type of inversion parameters in the first vector and the second vector, and placing the predicted vector after the numerical weighted crossover back into the vector set;

[0057] Return to the step of generating a random number in the third interval as the crossover determination probability until the number of crossover cycles is equal to the number of crossovers, and obtain a crossover result of the crossover operation;

[0058] Among them, the number of crossover cycles is initialized to 0.

[0059] On the other hand, an embodiment of the present invention provides a simulated annealing inversion solution device for terrain, comprising:

[0060] The first module is used to initialize a vector set, a first fitness value and an initial temperature; the initial temperature is used as an annealing temperature; wherein the vector set includes a preset number of prediction vectors;

[0061] The second module is used to traverse all prediction vectors in the vector set to perform inversion solution and fitness evaluation on the prediction area, and obtain the fitness function value of each prediction vector for inversion solution of the prediction area;

[0062] The result of the inversion solution is the predicted terrain data of the predicted area; the value of the fitness function is negatively correlated with the adaptability of the prediction vector for the inversion solution of the predicted area;

[0063] The third module is used to output the corresponding prediction vector as the optimal prediction vector when the fitness function value corresponding to the prediction vector in the vector set is less than the iterative optimal threshold; otherwise,

[0064] The fourth module is used to determine the excellence of each prediction vector according to the fitness function value, perform a random probability selection operation on the prediction vectors in the vector set based on the excellence, and update the vector set according to the selection result of the selection operation; and then use the minimum fitness function value corresponding to the vector set as the second fitness value; the excellence is negatively correlated with the numerical value of the fitness function value;

[0065] A fifth module is used for performing a random probability mutation operation on the prediction vector in the vector set based on the annealing temperature when the second fitness value is greater than or equal to the first fitness value, and updating the vector set according to the mutation result of the mutation operation; otherwise, updating the first fitness value by the second fitness value;

[0066] A sixth module is used to perform a crossover operation with a random probability on the prediction vectors in the vector set based on the second fitness value, and update the vector set according to the crossover result of the crossover operation;

[0067] A seventh module is used to perform temperature attenuation processing on the annealing temperature based on a preset temperature attenuation rate, and update the annealing temperature according to the attenuation result of the temperature attenuation processing;

[0068] The eighth module is used for returning to the step of performing inversion solution and fitness evaluation of the prediction area by traversing all prediction vectors in the vector set when the annealing temperature is greater than or equal to the preset termination temperature, until the annealing temperature is less than the termination temperature, and outputting the optimal prediction vector according to the last updated vector set;

[0069] The ninth module is used to obtain the corresponding predicted terrain data through inversion according to the optimal prediction vector as the terrain solution result of the prediction area.

[0070] On the other hand, an embodiment of the present invention provides an electronic device, including: a processor and a memory; the memory is used to store programs; the processor executes the program to implement the simulated annealing inversion solution method for the above-mentioned terrain.

[0071] On the other hand, an embodiment of the present invention provides a computer storage medium storing a program executable by a processor, and the program executable by the processor is used to implement the simulated annealing inversion solution method for the above-mentioned terrain when executed by the processor.

[0072] The embodiment of the present invention initializes a vector set, a first fitness value and an initial temperature; uses the initial temperature as an annealing temperature; wherein the vector set includes a preset number of prediction vectors; traverses all prediction vectors in the vector set to perform inversion solution and fitness evaluation on the prediction area, and obtains the fitness function value of each prediction vector for inversion solution of the prediction area; wherein the result of the inversion solution is the predicted terrain data of the prediction area; the numerical value of the fitness function value is negatively correlated with the adaptability of the prediction vector for inversion solution of the prediction area; when there is a prediction vector in the vector set whose corresponding fitness function value is less than the iterative optimal threshold, the corresponding prediction vector is output as the optimal prediction vector; otherwise, the excellence of each prediction vector is determined according to the fitness function value, and based on the excellence, the prediction vector in the vector set is selected with random probability, and the vector set is updated according to the selection result of the selection operation; and then the minimum fitness function value corresponding to the vector set is used as the second fitness value; the optimal The degree of excellence is negatively correlated with the numerical value of the fitness function value; when the second fitness value is greater than or equal to the first fitness value, based on the annealing temperature, a random probability mutation operation is performed on the prediction vector in the vector set, and the vector set is updated according to the mutation result of the mutation operation; otherwise, the first fitness value is updated through the second fitness value; based on the second fitness value, a random probability crossover operation is performed on the prediction vector in the vector set, and the vector set is updated according to the crossover result of the crossover operation; the annealing temperature is subjected to temperature decay processing based on a preset temperature decay rate, and the annealing temperature is updated according to the decay result of the temperature decay processing; when the annealing temperature is greater than or equal to the preset termination temperature, the steps of traversing all the prediction vectors in the vector set to perform inversion solution and fitness evaluation of the prediction area are returned until the annealing temperature is less than the termination temperature, and the optimal prediction vector is output according to the last updated vector set; the corresponding prediction terrain data obtained by inversion solution according to the optimal prediction vector is used as the terrain solution result of the prediction area. The embodiment of the present invention realizes the dual improvement of prediction accuracy and efficiency through the innovative combination of multi-dimensional starting search and genetic algorithm optimization mechanism, and has significant technical advancement and broad application prospects. The present invention includes the following beneficial effects: Prediction performance is significantly improved: The fusion of multi-dimensional starting search and genetic algorithm optimization mechanism enables this solution to show excellent performance in complex and changeable prediction tasks, and the prediction accuracy and stability are greatly improved. Computational efficiency optimization: The dynamic mutation probability adjustment mechanism effectively reduces unnecessary computational overhead and accelerates the convergence process, so that this solution has high computational efficiency while maintaining high-precision prediction. The embodiment of the present invention can efficiently and accurately perform simulated annealing inversion solutions for terrain. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The accompanying drawings are used to provide a further understanding of the technical solution of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the technical solution of the present invention and do not constitute a limitation on the technical solution of the present invention.

[0074] Figure 1 A schematic diagram of an implementation environment for performing simulated annealing inversion solution of terrain provided in an embodiment of the present invention;

[0075] Figure 2 A schematic flow chart of a simulated annealing inversion solution method for terrain provided in an embodiment of the present invention;

[0076] Figure 3 A schematic diagram of the expansion process of the initialization vector set provided in an embodiment of the present invention;

[0077] Figure 4 A schematic diagram of the expansion process of the inversion solution and fitness evaluation provided in an embodiment of the present invention;

[0078] Figure 5 A schematic diagram of an expansion process for outputting an optimal prediction vector provided by an embodiment of the present invention;

[0079] Figure 6 A schematic diagram of the expansion flow of the selection operation provided by an embodiment of the present invention;

[0080] Figure 7 A schematic diagram of the expansion process of the mutation operation provided by an embodiment of the present invention;

[0081] Figure 8 A schematic diagram of the overall process principle of the simulated annealing inversion solution method for terrain provided by an embodiment of the present invention;

[0082] Fig. 9 A schematic diagram of the structure of a simulated annealing inversion solution device for terrain provided by an embodiment of the present invention;

[0083] Fig.10 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0084] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0085] It should be noted that, although the functional modules are divided in the system schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the system or the order in the flowchart. The terms "first / S100", "second / S200", etc. in the specification, claims and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0086] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0087] It is understandable that the simulated annealing inversion solution method for terrain provided in the embodiment of the present invention can be applied to any computer device with data processing and computing capabilities, and this computer device can be various terminals or servers. When the computer device in the embodiment is a server, the server is an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. In some embodiments, the terminal is a smart phone, a tablet computer, a laptop computer, and a desktop computer, but is not limited to this.

[0088] To facilitate the understanding of the technical solution of the present invention, the technical features that may appear in the embodiments of the present invention and the specific terms of the variable parameters that may be applied are first explained:

[0089] The technical features are described as follows:

[0090] Seabed topography: the solid earth surface under water. Due to the cover of seawater, the ups and downs of the seabed topography are difficult to observe directly. The early plumb-bob sounding method was time-consuming and of low precision. Since the 1920s, ships have used echo sounders during navigation, which can quickly measure the depth of the seabed and reveal the truth of the seabed topography in combination with precise positioning. In the present invention, the seabed topography refers to the depth of seawater at various latitudes and longitudes of the ocean.

[0091] Heuristic search: Use the heuristic information of the problem to guide the search, so as to reduce the search scope and reduce the complexity of the problem. For example, simulated annealing and genetic algorithm are classic heuristic search algorithms.

[0092] Simulated annealing: It is derived from solid annealing and is a probability-based algorithm. The solid is heated to a sufficiently high temperature and then slowly cooled. When heated, the particles inside the solid become disordered as the temperature rises, and the internal energy increases. When slowly cooled, the particles gradually become ordered and reach equilibrium at each temperature. Simulated annealing is a general optimization algorithm. In theory, the algorithm has probabilistic global optimization performance. It has been widely used in engineering, such as VLSI, production scheduling, control engineering and other fields. Simulated annealing is an optimization algorithm that effectively avoids falling into local minima and eventually tends to the global optimal serial structure by giving the search process a time-varying and eventually zero probability jump.

[0093] Genetic Algorithm (GA) is a computational model that simulates the biological evolution process, using mechanisms such as natural selection and gene recombination to search and optimize complex problems. Its basic idea is to create a set of artificial prediction vectors for optimization by simulating the mechanisms of natural inheritance and evolution. In a genetic algorithm, each prediction vector represents a potential solution, and each prediction vector has its own fitness evaluation criteria. These prediction vectors are selected, crossed, and mutated to find the optimal solution. Genetic algorithms have the characteristics of a wide range of applications, strong global search, and easy parallel computing. They are suitable for optimization solutions to various types of problems, including combinatorial optimization, continuous optimization, and multi-objective optimization.

[0094] Genetic operators: Simply put, genetic operators refer to mutation, crossover, and selection operations, which are mainly used to evaluate and generate new prediction vectors or solution vectors.

[0095] The Parker-Oldenburg inversion method is a frequency domain density interface iterative inversion method proposed by Oldenburg based on the Parker formula. Since this method has the advantage of fast calculation, its application has been rapidly developed. However, there is a downward extension factor in the formula, which causes the formula to have an oscillation phenomenon of high-frequency signals during the inversion iteration process, and as the inversion depth increases, this phenomenon becomes more obvious, which seriously affects the convergence of the inversion result. To solve this problem, a low-pass filter is usually applied in the inversion process to eliminate the high-frequency oscillation phenomenon. The Parker-Oldenburg inversion method is one of the important methods for calculating the seabed topography, and is also the starting point of the core of the present invention. At present, the input parameters of the method (filter-related parameters, density difference, depth), etc. are designed according to constants based on experience, but in fact, there is still room for tuning and mining these parameters.

[0096] RMS: Abbreviation for Root Mean Square, which is the result of taking the square root after dividing the sum of the squares of N terms by N, that is, the result of the root mean square.

[0097] The description of the variable parameters is as follows:

[0098] The parameters mainly include two parts: the simulated annealing method driven by genetic operators and the parameters of the Parker - Oldenburg inversion method, etc.

[0099] There are a total of 2 variables searched by the simulated annealing method driven by genetic operators, which are contrast (physical meaning: density difference, unit g / cm 3 ), and depth (physical meaning: depth, unit km). Then, the parameters of the simulated annealing method driven by genetic operators are respectively:

[0100] The maximum density difference maxContrast: the maximum value of contrast, which is determined by the average value of the target search area, generally obtained based on experience, and satisfies minContrast < maxContrast. In most cases, it is sufficient to satisfy maxContrast - minContrast >= 0.4, and it can be set according to the actual situation.

[0101] The minimum density difference minContrast: the minimum value of contrast, which is determined by the average value of the target search area, generally obtained based on experience, and satisfies minContrast < maxContrast. In most cases, it is sufficient to satisfy maxContrast - minContrast >= 0.4, and it can be set according to the actual situation.

[0102] The maximum depth maxDepth: the maximum value of depth, which is determined by the average value of the target search area, generally obtained based on experience, and satisfies maxDepth < minDepth. In most cases, it is sufficient to satisfy maxDepth - minDepth >= 10, and it can be set according to the actual situation.

[0103] The minimum depth minDepth: the minimum value of depth, which is determined by the average value of the target search area, generally obtained based on experience, and satisfies maxDepth < minDepth. In most cases, it is sufficient to satisfy maxDepth - minDepth >= 10, and it can be set according to the actual situation.

[0104] The initial temperature initTemp, which is taken according to requirements, and the default value is 100.

[0105] The final temperature finalTemp is set according to the requirements and the default value is 1.

[0106] The temperature decay rate tempReduceRate has a value range of [0, 1] and is determined based on demand. The default value is 0.99.

[0107] MaxPredCnt, the number of prediction vectors involved in the genetic operator, takes a value based on demand and is 100 by default.

[0108] Crossover operator probability crossoverRate, a parameter used by the crossover operator, has a value range of [0,1] and is determined based on demand. The default value is 0.9.

[0109] Crossover operator ratio crossoverAlpha, a parameter used by the crossover operator, with a value range of [0,1]. The value is determined according to requirements and the default value is 0.5.

[0110] Iterate the optimal threshold minIterDelt. If a prediction vector less than or equal to minIterDelt is found, it means that the optimal solution has been found and the entire search process ends directly.

[0111] The main parameters of the Parker-Oldenburg inversion method are as follows:

[0112] Density difference contrast, unit: g / cm 3 , which is one of the variables searched by the simulated annealing method driven by an improved genetic operator, and is one of the two search directions mentioned in the present invention.

[0113] Depth depth: unit is km, which is one of the variables searched by the simulated annealing method driven by the improved genetic operator, and is one of the two search directions mentioned in the present invention.

[0114] Gravity outlier matrix GravityMat: It is a matrix composed of (longitude, latitude, gravity outlier) triplets, generally data observed by seismic measurement, satellite measurement or other means, and is a constant in the present invention.

[0115] Seismic depth matrix SemisMat: It is a matrix composed of (longitude, latitude, depth) triplets, generally data observed by seismic measurement, satellite measurement or other means, and is a constant in the present invention.

[0116] The iteration threshold criterio, i.e., the iteration rms threshold of the Parker-Oldenburg inversion method, is generally a constant, and the default value is 0.1, which is a constant in the present invention.

[0117] Frequency threshold WH: low-pass filter threshold, the initial value is obtained based on experience, the default value is 0.1, and it is a constant in the present invention.

[0118] Frequency threshold SH: high-pass filter threshold, the initial value is obtained based on experience, the default value is 0.12, and it is a constant in the present invention.

[0119] Window function threshold truncation: It is a signal sampling weighting function in signal processing. The initial value is obtained based on experience, and the default value is 0.1. It is a constant in the present invention.

[0120] like Figure 1 FIG. 1 is a schematic diagram of an implementation environment provided by an embodiment of the invention. Figure 1 The implementation environment includes at least one terminal 102 and a server 101. The terminal 102 and the server 101 can be connected to a network wirelessly or wired to complete data transmission and exchange.

[0121] Server 101 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), as well as big data and artificial intelligence platforms.

[0122] In addition, the server 101 can also be a node server in the blockchain network. Blockchain is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithm.

[0123] The terminal 102 may be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited thereto. The terminal 102 and the server 101 may be directly or indirectly connected via wired or wireless communication, which is not limited in the embodiment of the present invention.

[0124] Based on the example Figure 1 In the implementation environment shown, an embodiment of the present invention provides a simulated annealing inversion solution method for terrain. The following is an example of applying the simulated annealing inversion solution method for terrain in the server 101. It can be understood that the simulated annealing inversion solution method for terrain can also be applied to the terminal 102.

[0125] Reference Figure 2 , Figure 2The flowchart of the simulated annealing inversion solution method for terrain applied to a server provided in an embodiment of the present invention, the execution subject of the simulated annealing inversion solution method for terrain can be any of the aforementioned computer devices (including a server or a terminal). Figure 2 , the method comprises the following steps:

[0126] S100, initializing a vector set, a first fitness value and an initial temperature; using the initial temperature as the annealing temperature;

[0127] The vector set includes a preset number of prediction vectors; the value of the first fitness value initially set is large enough to ensure that the first round of loop iteration can be updated by the second fitness value;

[0128] It should be noted that in some embodiments, a vector set is first initialized, each vector consisting of a density difference contrast and an average seafloor depth depth. The core of the initialization vector set is to generate MaxPredCnt vectors. There are three strategies, such as Figure 3 As shown, the initialization vector set may include at least one of the following steps:

[0129] S101, obtaining a preset standard vector value, and randomly generating a preset number of prediction vectors near the standard vector value;

[0130] The standard vector values ​​include standard values ​​of density difference and depth;

[0131] For example, in some specific implementations, the generation is based on the center point (i.e., the standard vector value). If the center point is given, the generation method is as follows:

[0132] contrast=initContrast+(maxContrast-minContrast)*normal(0,1);

[0133] depth=initDepth+(maxDepth-minDepth)*normal(0,1);

[0134] S102, randomly generating a preset number of prediction vectors within a target interval; the target interval represents a value range of the prediction vector;

[0135] The target interval includes a density difference interval (determined based on a preset maximum density difference value and a preset minimum density difference value) and a seafloor depth interval (determined based on a preset maximum depth value and a preset minimum depth value);

[0136] For example, in some specific implementations, the random generation is performed. If no center point is given, the random generation is performed within the search interval, and the generation method is as follows:

[0137] contrast=minContrast+(maxContrast-minContrast)*random();

[0138] depth=minDepth+(maxDepth-minDepth)*random();

[0139] S103 . Directly obtain a preset number of prediction vectors in response to a manual setting instruction.

[0140] For example, in some specific implementations, the genes are manually specified. In this case, the initialization genes of all prediction vectors need to be manually specified.

[0141] Among them, initContrast and initDepth represent the center point; random() represents the random value generated in [0,1]; normal(0,1) represents a normal distribution with a mean of 0 and a standard deviation of 1. The obtained random numbers satisfy the normal distribution, thereby ensuring the fairness of gene generation.

[0142] S200, traversing all prediction vectors in the vector set to perform inversion solution and fitness evaluation on the prediction area, and obtaining the fitness function value of each prediction vector for inversion solution of the prediction area;

[0143] The result of the inversion solution is the predicted terrain data of the predicted area; the value of the fitness function is negatively correlated with the adaptability of the prediction vector for the inversion solution of the predicted area;

[0144] It should be noted that in some embodiments, Figure 4 As shown, step S200 may include the following steps:

[0145] S201, traversing the prediction vectors in the vector set in turn, and solving by the Parker-Oldenburg inversion method to obtain the prediction terrain data of the prediction area obtained by the inversion solution corresponding to each prediction vector;

[0146] S202, based on the predicted terrain data, combined with the earthquake depth matrix of the predicted area, a root mean square calculation is performed to obtain a root mean square value as a fitness function value of the corresponding prediction vector for the adaptability of the inversion solution in the predicted area.

[0147] In some embodiments, the root mean square calculation specifically includes the following steps: obtaining the seismic depth matrix of the predicted area; traversing the first terrain values ​​of all data points of the predicted terrain data corresponding to the predicted vector, and collecting the second terrain values ​​of the corresponding data points from the seismic depth matrix; obtaining the third terrain value of the corresponding data point based on the square of the difference between the first terrain value and the second terrain value; performing square root processing based on the average value of the third terrain values ​​of all data points to obtain the root mean square value corresponding to the corresponding prediction vector.

[0148] For example, in some specific implementations, the corresponding fitness function value can be calculated based on the prediction vector set, which is specifically implemented as follows: for the current vector set, the contrast and depth of each vector are traversed one by one, and then the corresponding seabed topography data is calculated based on the Parker-Oldenburg inversion method. Then, combined with the seismic depth matrix SemisMat, the corresponding RMS is calculated, and each prediction vector will calculate its own RMS, and finally an array of size MaxPredCnt is obtained.

[0149] S300, when there is a prediction vector in the vector set whose corresponding fitness function value is less than the iterative optimal threshold, the corresponding prediction vector is output as the optimal prediction vector, and the process goes to step S900; otherwise, the process goes to the next step;

[0150] It should be noted that in some embodiments, Figure 5 As shown, outputting the corresponding prediction vector as the optimal prediction vector may include the following steps: S301, when there is a prediction vector in the vector set whose corresponding fitness function value is less than the iterative optimal threshold, directly outputting the prediction vector as the optimal prediction vector; S302, when there are multiple prediction vectors in the vector set whose corresponding fitness function values ​​are less than the iterative optimal threshold, outputting the prediction vector corresponding to the smallest fitness function value as the optimal prediction vector.

[0151] Exemplarily, in some specific implementations, all population prediction vectors are traversed to determine whether there is an RMS less than the iterative optimal threshold minIterDelt. If so, go to step S900; otherwise, execute the next step.

[0152] S400, determining the excellence of each prediction vector according to the fitness function value, performing a random probability selection operation on the prediction vectors in the vector set based on the excellence, and updating the vector set according to the selection result of the selection operation; and then taking the minimum fitness function value corresponding to the vector set as the second fitness value;

[0153] Among them, the degree of excellence is negatively correlated with the numerical value of the fitness function;

[0154] It should be noted that in some embodiments, Figure 6 As shown, based on the degree of excellence, the prediction vector in the vector set is randomly selected with probability, which may include the following steps:

[0155] S401, normalizing the excellence levels corresponding to all prediction vectors in the vector set to obtain a normalized excellence level corresponding to each prediction vector;

[0156] The expression of excellence is:

[0157]

[0158] In the formula, e represents the natural constant; RMS represents the fitness function value; minIterDelt represents the iterative optimal threshold; nowTemp represents the annealing temperature;

[0159] S402, sorting out the normalized excellence levels corresponding to all prediction vectors in the vector set to obtain an excellence level array;

[0160] The value of the nth element in the excellence array is the cumulative sum of all normalized excellences corresponding to the first prediction vector to the nth prediction vector in the vector set;

[0161] S403, generating a random number in the first interval as a selection probability;

[0162] S404, sequentially traverse the excellence array to find the first element whose probability is greater than the selection probability as the target element;

[0163] S405, selecting a corresponding prediction vector from the vector set as an update vector according to the element number of the target element, and returning to execute the step of generating a random number as a selection probability within the first interval until a preset number of update vectors are selected from the vector set;

[0164] S406: Use a preset number of update vectors as a selection result of the selection operation.

[0165] For example, in some specific implementations, applying a selection operator to a prediction vector based on probability may be implemented as follows:

[0166] The selection operator is mainly based on probability to select a better prediction vector. The smaller the RMS, the better the prediction vector, and the larger the RMS, the worse the prediction vector. At the same time, the influence of temperature is introduced. The larger the temperature, the better the prediction vector, and the smaller the temperature, the worse the prediction vector. The present invention uses To indicate the excellence of the prediction vector, The larger the value, the better the prediction vector. Considering that solid particles tend to become disordered at high temperatures, that is, the search variable is more likely to change dramatically at high temperatures, the current temperature is introduced into the evaluation system. At the same time, this is also used in the mutation operator. The higher the temperature, the greater the mutation probability.

[0167] Formula (1) is used to solve the sum of the priorities of all prediction vectors; Formula (2) is used to calculate the normalized prediction vector priority; referring to Formula (3), the sum of the priorities of all prediction vectors is 1; Formula (4) is used to calculate the array of cumulative priority sums. How to select an excellent prediction vector? Calculate a random number R0 in the range [0,1], traverse the accRMS array to find the first prediction vector greater than R0, and the prediction vector will be selected.

[0168] Through this selection operator, the larger the averRMS, the greater the probability of being selected, and at the same time, the prediction vector with a smaller averRMS still has a chance to be selected. The reason for this design is that the current excellent prediction vector means that the gene expression is excellent and worthy of being selected, but considering that the reason for the low averRMS of the prediction vector may be that the gene expression is relatively slow, it still needs to be selected with a certain probability.

[0169]

[0170] Generally, the number of prediction vectors is greater than MaxPredCnt. Here, the selection operator is run MaxPredCnt times to finally obtain MaxPredCnt excellent prediction vectors. Note that the same prediction vector may be selected multiple times.

[0171] S500, when the second fitness value is greater than or equal to the first fitness value, based on the annealing temperature, performing a random probability mutation operation on the prediction vector in the vector set, and updating the vector set according to the mutation result of the mutation operation; otherwise, updating the first fitness value by the second fitness value;

[0172] It should be noted that in some embodiments, Figure 7 As shown, the prediction vector includes two types of inversion parameters: density difference and depth. Based on the annealing temperature, the prediction vector in the vector set is subjected to a random probability mutation operation, which may include the following steps:

[0173] S501, based on the annealing temperature, the mutation probability is calculated in combination with the first fitness value and the second fitness value;

[0174] Among them, the expression of mutation probability is:

[0175]

[0176] In the formula, mutationRate represents the mutation probability; e represents the natural constant; RMS1 represents the first fitness value; RMS2 represents the second fitness value; nowTemp represents the annealing temperature;

[0177] S502, generating a random number in the second interval as a mutation determination probability; adding 1 to the number of mutation cycles;

[0178] S503, when the mutation determination probability is less than the preset mutation probability, randomly taking a prediction vector from the vector set as a mutation prediction vector;

[0179] S504, performing numerical variation on any inversion parameter in the variation prediction vector within the target interval, and placing the prediction vector after the numerical variation back into the vector set; the target interval represents the value range of the prediction vector;

[0180] S505, returning to the step of generating a random number in the second interval as a mutation determination probability, until the number of mutation cycles is equal to a preset number, and obtaining a mutation result of the mutation operation;

[0181] Among them, the number of mutation cycles is initialized to 0.

[0182] Exemplarily, in some specific implementations, it is assumed that RMS1 represents the best evaluation value currently searched globally, and RMS2 represents the smallest evaluation value among all current prediction vectors. If RMS2 is less than RMS1, it means that there is a prediction vector that is better than the current one, and go to step S600, otherwise the mutation operation is performed and the first fitness value is updated. It is emphasized here that step S600 is a crossover operator process, and step S500 involves a mutation operator process. The crossover operator is used in each iteration, but the mutation operator is not necessarily used. This design is different from the genetic algorithm. Because the core point of the mutation operator is to jump out of the local optimum, only when a better solution is not found, it means that it is trapped in the local optimum, and only then is a mutation operator needed.

[0183] Specifically, the specific process of the mutation operation can be implemented as follows:

[0184] First, the mutation probability is calculated based on the current temperature. The probability calculation of the mutation operator refers to formula (5).

[0185]

[0186] Taking into account that solid particles tend to become disordered at high temperatures, that is, the search variable is more likely to undergo drastic changes at high temperatures, the current temperature is introduced into the mutation operator probability calculation formula. The higher the temperature, the greater the probability of mutation, and the lower the temperature, the smaller the probability of mutation. At the same time, the closer RMS2 is to RMS1, the greater the probability of mutation, and the farther RMS2 is from RMS1, the smaller the probability of mutation. In the critical state, that is, when RMS2=RMS1, the probability of mutation is 1, that is, mutation is bound to occur. In this case, it means that the search may have fallen into a local optimum, and the probability of mutation needs to be increased to jump out of the local optimum. Based on this, the present invention implements dynamic adjustment of the probability of mutation.

[0187] Then, the mutation operator is applied to the prediction vector based on probability. The mutation operator randomly selects a single prediction vector and then selects the gene for mutation. The maximum number of mutation operations is MaxPredCnt, and then a random number R1 in [0,1] is calculated. If R1 is less than mutationRate, a prediction vector is randomly selected to be mutated; then a random one is selected from Case1 and Case2 for mutation operator operation.

[0188] Case 1: Density difference mutation, only the density difference mutation operation is performed. The mutation formula is as follows:

[0189] A1[contrast]=A1[contrast]+(maxContrast-minContrast)*normal(0,1);

[0190] Case 2: Deep mutation, only deep mutation operation is performed. The mutation formula is as follows:

[0191] A1[depth]=A1[depth]+(maxDepth-minDepth)*normal(0,1);

[0192] Among them, normal(0,1) represents a normal distribution with a mean of 0 and a standard deviation of 1.

[0193] In general, when facing the selected MaxPredCnt prediction vector, the maximum number of mutations is MaxPredCnt, and then whether mutation is needed is determined based on the mutationRate probability; if mutation is clear, individuals are randomly selected, and then which gene is selected for mutation. The new prediction vector is obtained after completing the mutation operator.

[0194] S600, based on the second fitness value, performing a crossover operation with a random probability on the prediction vectors in the vector set, and updating the vector set according to the crossover result of the crossover operation;

[0195] It should be noted that, in some embodiments, the prediction vector includes two types of inversion parameters, namely, density difference and depth; and based on the second fitness value, performing a crossover operation with a random probability on the prediction vectors in the vector set may include the following steps:

[0196] Based on the second fitness value, the number of crossovers is calculated by combining the iterative optimal threshold and the preset number;

[0197] The expression of the number of crossovers is:

[0198]

[0199] In the formula, crossCnt represents the number of crossovers; random(x) represents a random value generated in [0,x]; RMS2 represents the second fitness value; minIterDelt represents the iterative optimal threshold; MaxPredCnt represents the preset number;

[0200] Generate a random number in the third interval as the crossover determination probability; add 1 to the number of crossover cycles;

[0201] When the crossover determination probability is less than the preset crossover operator probability, two prediction vectors are randomly obtained from the vector set as crossover vectors; the crossover vectors include the first vector and the second vector;

[0202] Performing numerical weighted crossover on at least one type of inversion parameters in the first vector and the second vector, and placing the predicted vector after the numerical weighted crossover back into the vector set;

[0203] Return to the step of generating a random number in the third interval as the crossover determination probability until the number of crossover cycles is equal to the number of crossovers, and obtain a crossover result of the crossover operation;

[0204] Among them, the number of crossover cycles is initialized to 0.

[0205] For example, in some specific implementations, applying a crossover operator to a prediction vector based on probability may be implemented as follows:

[0206] The crossover operator selects two prediction vectors and then crosses them. Referring to formula (6), RMS2 represents the minimum evaluation value among all current prediction vectors, round() represents rounding up, and crossCnt represents the number of crossovers. The advantage of this design is that when RMS2 is larger, it means that the distance from the global optimal solution is farther, and the crossover operator operates more; when RMS2 is smaller, it means that the distance from the global optimal solution is closer, and the crossover operator can guarantee at least MaxPredCnt times.

[0207]

[0208] For a single crossover operation, based on the excellent prediction vector obtained by the selection operator, two prediction vectors A1 and A2 are randomly selected; then the random number R2 of [0,1] is calculated. If R2 is less than crossoverRate, it is determined to crossover the two prediction vectors A1 and A2; then one is randomly selected from Case1, Case2 and Case3 for the crossover operator operation; then a single crossover operator operation is completed; this is repeated, and crossCnt crossover operations are performed, and finally the crossover operation of the entire population is completed.

[0209] Case 1: Density difference crossover, only the density difference crossover operation is performed. The crossover formula is as follows:

[0210] t1 = A1[contrast];

[0211] t2 = A2[contrast];

[0212] A1[contrast]=t1*crossoverAlpha+t2*(1-crossoverAlpha);

[0213] A2[contrast]=t2*crossoverAlpha+t1*(1-crossoverAlpha);

[0214] Case 2: Depth crossover, only depth crossover operation is performed. The crossover formula is as follows:

[0215] t1 = A1[depth];

[0216] t2 = A2[depth];

[0217] A1[depth]=t1*crossoverAlpha+t2*(1-crossoverAlpha);

[0218] A2[depth]=t2*crossoverAlpha+t1*(1-crossoverAlpha);

[0219] Case 3: Density difference crossover and depth crossover, all genes are crossovered, the operation formula is as above.

[0220] In general, when facing the selected MaxPredCnt prediction vector, the total number of crossovers is calculated by RMS1, and then a random prediction vector is selected to confirm whether to participate in the crossover according to probability; the prediction vector participating in the crossover selects which genes to participate in the crossover according to probability. The new prediction vector is obtained after completing the crossover operator.

[0221] S700, performing temperature attenuation processing on the annealing temperature based on a preset temperature attenuation rate, and updating the annealing temperature according to the attenuation result of the temperature attenuation processing;

[0222] For example, in some specific implementations, after the operation from step S400 to step S600, a round of temperature iteration is completed, and a new vector is obtained. After the random iteration of the current temperature is completed, the temperature needs to be attenuated, where the attenuation formula is: tNowTemp = tNowTemp*tempReduceRate, where the initial value of tNowTemp (i.e., annealing temperature) is the initial temperature initTemp.

[0223] S800, when the annealing temperature is greater than or equal to the preset termination temperature, return to execute the steps of traversing all prediction vectors in the vector set to perform inversion solution and fitness evaluation of the prediction area, until the annealing temperature is less than the termination temperature, and output the optimal prediction vector according to the last updated vector set;

[0224] It should be noted that, in some embodiments, outputting the optimal prediction vector based on the last updated vector set may include the following steps: obtaining the minimum value of the fitness function values ​​corresponding to all prediction vectors in the last updated vector set, and outputting the prediction vector corresponding to the minimum value as the optimal prediction vector.

[0225] Exemplarily, in some specific implementations, steps S200 to S700 are executed in a loop iteratively. Each time the execution is performed, the temperature decays once. It is determined whether tNowTemp is less than the final temperature finalTemp. If so, it indicates that the iterative process ends and the process goes to step S900. Otherwise, the process goes to step S200.

[0226] S900 , using the inversion method to obtain the corresponding predicted terrain data according to the optimal predicted vector as the terrain solution result of the predicted area.

[0227] For example, in some specific implementations, this step obtains the optimal solution of the entire search process. There are only two solutions obtained here: the optimal vector that satisfies the iterative optimal threshold minIterDelt, or the optimal vector obtained by searching the entire space. At this point, the entire search process ends, and not only the optimal solution is obtained, but also the corresponding seabed topography data.

[0228] In order to explain the principle of the technical solution of the present invention in detail, the overall process of the present invention is described below in combination with some specific embodiments. It is easy to understand that the following is an explanation of the technical principle of the present invention and cannot be regarded as a limitation of the present invention.

[0229] First of all, it should be noted that the core idea of ​​the current technical solution is to simulate annealing to search for the optimal value, then confirm the optimal value by the Parker-Oldenburg inversion method, and finally obtain the final result by adaptively updating the filter matrix parameters. A large number of experiments have proved that although the existing technical solution can better solve the optimal density difference contrast and depth depth, it has the problems of low solution efficiency, unstable solution accuracy, and easy to fall into local optimality. Disadvantages of existing technology:

[0230] 1) Low solution efficiency. Existing solutions generate density contrast and depth through purely random ideas, which means that a large number of random tests are required to obtain the optimal solution, resulting in low solution efficiency. For example, sometimes it takes 10 hours to get the optimal solution;

[0231] 2) The average solution accuracy is unstable. Existing solutions generate density contrast and depth through purely random ideas. A large number of tests have found that for all optimal solutions, 40% of the results have good accuracy, while the rest have poor accuracy, resulting in unstable average solution accuracy;

[0232] 3) It is easy to fall into the local optimum and fail to obtain the optimal solution. When the iteration parameter configuration is low, such as when the temperature is set to 50 degrees and the temperature decay rate is 0.6, it is easy to fail to find the optimal solution. Through a large number of tests, the optimal solution can be found in 50% of the cases, and the optimal solution cannot be obtained in the rest of the cases.

[0233] In view of this, the present invention proposes a simulated annealing method driven by a genetic operator and a method for solving the seafloor topography by Parker-Oldenburg inversion, which finds the optimal parameters by using the simulated annealing method driven by a genetic operator and obtains a more accurate seafloor topography by using the Parker-Oldenburg inversion method. Figure 8 As shown, the implementation process of the present invention can be as follows:

[0234] Step 1: Parameter initialization. The parameters mainly include two parts: the simulated annealing method driven by the genetic operator and the Parker-Oldenburg inversion method parameters (for details of the specific parameters, please refer to the aforementioned variable parameter description, which will not be repeated here)).

[0235] Step 2: Initialize the vector set. Each vector consists of density difference contrast and average seafloor depth depth. The core of the initialization vector set is to generate MaxPredCnt vectors. There are three strategies:

[0236] Case 1: Generate based on the center point. If the center point is given, the generation method is as follows:

[0237] contrast=initContrast+(maxContrast-minContrast)*normal(0,1);

[0238] depth=initDepth+(maxDepth-minDepth)*normal(0,1);

[0239] Case 2: Random generation. If no center point is given, it is randomly generated within the search interval. The generation method is as follows:

[0240] contrast=minContrast+(maxContrast-minContrast)*random();

[0241] depth=minDepth+(maxDepth-minDepth)*random();

[0242] Case 3: Manually specified. In this case, the initialization genes of all prediction vectors need to be manually specified.

[0243] Among them, initContrast and initDepth represent the center point; random() represents the random value generated in [0,1]; normal(0,1) represents the normal distribution with a mean of 0 and a standard deviation of 1. The obtained random numbers satisfy the normal distribution, thus ensuring the fairness of gene generation. Go to step 3.

[0244] Step 3: Calculate the corresponding fitness function value based on the predicted vector set. For the current vector set, traverse the contrast and depth of each vector one by one, and then calculate the corresponding seafloor topography data based on the Parker-Oldenburg inversion method. Then combine the seismic depth matrix SemisMat to calculate the corresponding RMS. Each predicted vector will calculate its own RMS, and finally get an array of size MaxPredCnt. Go to step 4.

[0245] Step 4: Is there a vector smaller than the threshold? Traverse all population prediction vectors to determine whether there is a vector with RMS smaller than the iterative optimal threshold minIterDelt. If so, go to step 13, indicating that the optimal prediction vector (optimal solution) has been found. Otherwise, go to step 5.

[0246] Step 5: Apply the selection operator to the prediction vector based on probability. The selection operator mainly selects a better prediction vector based on probability. The smaller the RMS, the better the prediction vector, and the larger the RMS, the worse the prediction vector. At the same time, the influence of temperature is introduced. The larger the temperature, the better the prediction vector, and the smaller the temperature, the worse the prediction vector. The present invention uses To indicate the excellence of the prediction vector, The larger the value, the better the prediction vector. Considering that solid particles tend to become disordered at high temperatures, that is, the search variable is more likely to change dramatically at high temperatures, the current temperature is introduced into the evaluation system. At the same time, this is also used in the mutation operator. The higher the temperature, the greater the mutation probability.

[0247] Formula (1) is used to solve the sum of the priorities of all prediction vectors; Formula (2) is used to calculate the normalized prediction vector priority; referring to Formula (3), the sum of the priorities of all prediction vectors is 1; Formula (4) is used to calculate the array of cumulative priority sums. How to select an excellent prediction vector? Calculate a random number R0 in the range [0,1], traverse the accRMS array to find the first prediction vector greater than R0, and the prediction vector will be selected.

[0248] Through this selection operator, the larger the averRMS, the greater the probability of being selected, and at the same time, the prediction vector with a smaller averRMS still has a chance to be selected. The reason for this design is that the current excellent prediction vector means that the gene expression is excellent and worthy of being selected, but considering that the reason for the low averRMS of the prediction vector may be that the gene expression is relatively slow, it still needs to be selected with a certain probability.

[0249]

[0250]

[0251] For example, if minIterDelt is 1.5 and nowTemp is 1.0, the array of the original RMS is [4.5, 8.5, 2.5, 10.5], that is, there are four predicted vectors with RMS of 4.5, 8.5, 2.5, and 10.5 respectively.

[0252] By formula (1) we can get

[0253] The averRMS array is obtained by formula (2): The calculated value is averRMS = [0.1189, 0.0022, 0.8786, 0.0003]. Formula (3) can be used to verify that the sum of the averRMS array is 1.

[0254] Calculate the accRMS array through formula (4).

[0255] accRMS[0] = 0.1189;

[0256] accRMS[1] = 0.1189 + 0.0022 = 0.1211;

[0257] accRMS[2] = 0.1189 + 0.0022 + 0.8786 = 0.9997;

[0258] accRMS[3] = 0.1189 + 0.0022 + 0.8786 + 0.0003 = 1.0000;

[0259] Then calculate the random value R1. If R1 < accRMS[0], then accRMS[0] is selected; if R1 >= accRMS[0] and R1 < accRMS[1], then accRMS[1] is selected; if R1 >= accRMS[1] and R1 < accRMS[2], then accRMS[2] is selected; if R1 >= accRMS[2] and R1 < accRMS[3], then accRMS[3] is selected; if R1 >= accRMS[3], then accRMS[3] is selected, and this is the critical case.

[0260] Generally, the number of prediction vectors is greater than MaxPredCnt. Here, the selection operator is run MaxPredCnt times, and finally MaxPredCnt excellent prediction vectors are obtained. Note that the same prediction vector may be selected twice (this is the difference from simulated annealing). Go to step 6.

[0261] Step 6: Whether there is a vector better than the current one. Assume that RMS1 represents the currently globally searched optimal evaluation value, and RMS2 represents the minimum evaluation value among all current prediction vectors. If RMS2 is less than RMS1, it means there is a prediction vector better than the current one, go to step 9, otherwise go to step 7.

[0262] Here, it should be emphasized that step 9 is the crossover operator process, and steps 7 and 8 are the mutation operator processes. The crossover operator is used in each iteration, but the mutation operator may not be used. This design is different from the genetic algorithm. Because the core point of the mutation operator is to jump out of the local optimum. Only when a better solution has not been found does it mean that the local optimum has been reached, and at this time, the mutation operator is needed.

[0263] Step 7: Calculate the mutation probability based on the current temperature. The probability calculation of the mutation operator refers to formula (5).

[0264]

[0265] Taking into account that solid particles tend to become disordered at high temperatures, that is, the search variable is more likely to undergo drastic changes at high temperatures, the current temperature is introduced into the mutation operator probability calculation formula. The higher the temperature, the greater the probability of mutation, and the lower the temperature, the smaller the probability of mutation. At the same time, the closer RMS2 is to RMS1, the greater the probability of mutation, and the farther RMS2 is from RMS1, the smaller the probability of mutation. In the critical state, that is, when RMS2=RMS1, the probability of mutation is 1, that is, mutation is bound to occur. In this case, it means that the search may have fallen into a local optimum, and the probability of mutation needs to be increased to jump out of the local optimum. Based on this, the present invention implements dynamic adjustment of the probability of mutation. Go to step 8.

[0266] Step 8: Apply the mutation operator to the prediction vector based on probability. The mutation operator randomly selects a single prediction vector and then selects the gene for mutation. The maximum number of mutation operations is MaxPredCnt, and then calculates the random number R1 in [0,1]. If R1 is less than mutationRate, a prediction vector is randomly selected for mutation; then a random one is selected from Case1 and Case2 for mutation operator operation.

[0267] Case 1: Density difference mutation, only the density difference mutation operation is performed. The mutation formula is as follows:

[0268] A1[contrast]=A1[contrast]+(maxContrast-minContrast)*normal(0,1);

[0269] Case 2: Deep mutation, only deep mutation operation is performed. The mutation formula is as follows:

[0270] A1[depth]=A1[depth]+(maxDepth-minDepth)*normal(0,1);

[0271] Among them, normal(0,1) represents a normal distribution with a mean of 0 and a standard deviation of 1.

[0272] In general, when facing the selected MaxPredCnt prediction vector, the maximum number of mutations is MaxPredCnt, and then confirm whether mutation is needed based on the mutationRate probability; if mutation is clear, randomly select individuals and then choose which gene to use for mutation. After completing the mutation operator, the new prediction vector is obtained. Go to step 10.

[0273] Step 9: Apply the crossover operator to the prediction vector based on probability. The crossover operator selects two prediction vectors and then crosses the genes. Referring to formula (6), RMS2 represents the minimum evaluation value of all current prediction vectors, round() represents rounding up, and crossCnt represents the number of crossovers. The advantage of this design is that when RMS2 is larger, it means that the distance to the global optimal solution is farther, and the crossover operator operates more; when RMS2 is smaller, it means that the distance to the global optimal solution is closer, and the crossover operator can guarantee at least MaxPredCnt times.

[0274]

[0275] For a single crossover operation, based on the excellent prediction vector obtained by the selection operator, two prediction vectors A1 and A2 are randomly selected; then the random number R2 of [0,1] is calculated. If R2 is less than crossoverRate, it is determined to crossover the two prediction vectors A1 and A2; then one is randomly selected from Case1, Case2 and Case3 for the crossover operator operation; then a single crossover operator operation is completed; this is repeated, and crossCnt crossover operations are performed, and finally the crossover operation of the entire population is completed.

[0276] Case 1: Density difference crossover, only the density difference crossover operation is performed. The crossover formula is as follows:

[0277] t1 = A1[contrast];

[0278] t2 = A2[contrast];

[0279] A1[contrast]=t1*crossoverAlpha+t2*(1-crossoverAlpha);

[0280] A2[contrast]=t2*crossoverAlpha+t1*(1-crossoverAlpha);

[0281] Case 2: Depth crossover, only depth crossover operation is performed. The crossover formula is as follows:

[0282] t1 = A1[depth];

[0283] t2 = A2[depth];

[0284] A1[depth]=t1*crossoverAlpha+t2*(1-crossoverAlpha);

[0285] A2[depth]=t2*crossoverAlpha+t1*(1-crossoverAlpha);

[0286] Case 3: Density difference crossover and depth crossover, all genes are crossovered, the operation formula is as above.

[0287] In general, when facing the selected MaxPredCnt prediction vector, the total number of crossovers is calculated by RMS1, and then a random prediction vector is selected to confirm whether to participate in the crossover according to probability; the prediction vector participating in the crossover selects which genes to participate in the crossover according to probability. The new prediction vector is obtained after completing the crossover operator.

[0288] Here is a specific example. If minRMS is 1.2, minIterDelt is 1, sizePop is 4, and crossoverAlpha is 0.8, then corssCnt = round(1.2 / 1)*4 = 8, that is, at most 8 pairs of individuals participate in the crossover operation. Here, we take the crossover of a pair of individuals as an example. Assume that the density difference of individual A1 is A1[contrast] = 0.5, and the depth of individual A1 is A1 [depth] = 20; the density difference of individual A2 is A2[contrast] = 0.3, and the depth of individual A2 is A2 [depth] = 10. If the random number R2 is less than crossoverRate, A1 and A2 are definitely crossed.

[0289] If we only do the density intersection:

[0290] A1[contrast]=0.5*0.8+(1-0.8)*0.3=0.46,

[0291] A2[contrast]=0.3*0.8+(1-0.8)*0.5=0.34;

[0292] If only the depth intersection is done, then:

[0293] A1[depth]=20*0.8+(1-0.8)*10=18,

[0294] A2[depth]=10*0.8+(1-0.8)*20=12;

[0295] If you do the intersection of density difference and depth at the same time:

[0296] A1[contrast]=0.5*0.8+(1-0.8)*0.3=0.46,

[0297] A2[contrast]=0.3*0.8+(1-0.8)*0.5=0.34;

[0298] A1[depth]=20*0.8+(1-0.8)*10=18,

[0299] A2[depth]=10*0.8+(1-0.8)*20=12;

[0300] Go to step 10.

[0301] Step 10: Get a new round of vector sets based on the genetic operator. After the operations from step 5 to step 9, a round of temperature iteration is completed, and a new vector is obtained. With the help of the survival of the fittest and natural selection of the genetic operator, determine whether the global optimal solution is found based on the new population. Go to step 11.

[0302] Step 11: Reduce the temperature. After the random iteration of the current temperature is completed, the temperature needs to be attenuated. The attenuation formula is: tNowTemp = tNowTemp*tempReduceRate, where the initial value of tNowTemp is the initial temperature initTemp. Go to step 12.

[0303] Step 12: Whether the minimum temperature threshold is reached. Steps 3 to 11 are executed in a loop. Each time the temperature is executed, it is determined whether tNowTemp is less than the final temperature finalTemp. If so, it indicates that the iteration process is over and the process goes to step 13. Otherwise, the process goes to step 3.

[0304] Step 13: Find the optimal solution. This step yields the optimal solution for the entire search process. There are only two solutions: the optimal vector that satisfies the iterative optimal threshold minIterDelt, or the optimal vector obtained by searching the entire space. At this point, the entire search process ends, and not only the optimal solution is obtained, but also the corresponding seafloor topography data.

[0305] In summary, the present invention proposes a method for solving seabed topography by using simulated annealing driven by a genetic operator and Parker-Oldenburg inversion. The optimal parameters are found by using the simulated annealing method driven by a genetic operator, and a more accurate seabed topography is obtained by using the Parker-Oldenburg inversion method. The technical features of the present invention include but are not limited to:

[0306] 1) A genetic operator-driven simulated annealing method is proposed to search for the optimal density difference contrast and depth depth. Real number coding has advantages in continuous parameter optimization and continuous gradient problems, which ultimately improves the accuracy of Parker-Oldenburg inversion of seafloor topography.

[0307] 2) It is proposed to apply the optimized genetic operator to the vector generation of simulated annealing. The mutation probability of the improved genetic operator is dynamically adjusted with the temperature and search progress, which improves the ability to jump out of the local optimum, ensures that the improved simulated annealing search process is more efficient, and improves the accuracy and efficiency of data solution;

[0308] 3) The genetic operator-driven simulated annealing method and Parker-Oldenburg inversion are combined to ensure the accuracy of the seabed topography and improve computational efficiency.

[0309] It is important to point out that, in the face of the common Parker-Oldenburg inversion method improvement problem, unlike existing work, the present invention proposes a genetic operator-driven simulated annealing search strategy, which improves the efficiency of solution and the stability of solution accuracy. A small number of steps are consistent with existing work, while the core steps are unique to the present invention. Through a large number of tests, the experimental results of the present invention meet expectations. In summary, compared with existing simulated annealing related work, the core points of the present invention are:

[0310] 1) Based on the initial search of multiple prediction vectors, existing work is a single initial search variable;

[0311] 2) Introduce selection, mutation and crossover operators to predict new vectors respectively. The predicted vectors obtained by mutation and crossover operators are added to the existing vectors, and the redundant vectors are eliminated by selection operator;

[0312] 3) Dynamic changes in mutation probability based on temperature and search progress.

[0313] Compared with the existing genetic algorithm related work, the core points of the present invention are:

[0314] 1) The mutation probability changes dynamically based on temperature and search progress. The mutation probability of existing work is a fixed value set manually;

[0315] 2) The calculation formulas for introducing selection, mutation and crossover operators have different perspectives.

[0316] 3) The processes of mutation and crossover operators are different. In genetic algorithms, mutation operators and crossover operators may act on individuals at the same time, but the present invention will not. For example, for individuals A1 and A2, based on the present invention, A1 and A2 cross to obtain A3 and A4, A1 mutates to obtain A5, and finally obtains A1, A2, A3, A4, A5. Mutation and crossover will not act on the newly obtained prediction vector at the same time. By selecting operators to eliminate redundant vectors, what may be obtained in the end is A3 and A5. In the existing genetic algorithm work, A1 and A2 cross to obtain A3 and A4, A4 mutates to obtain A5, and finally obtains A3 and A5, among which A3 has undergone the crossover operator, and A5 has undergone the crossover and mutation operators.

[0317] On the other hand, Fig. 9 As shown, the embodiment of the present invention provides a simulated annealing inversion solution device 900 for terrain, which may include:

[0318] The first module 901 is used to initialize a vector set, a first fitness value and an initial temperature; the initial temperature is used as an annealing temperature; wherein the vector set includes a preset number of prediction vectors;

[0319] The second module 902 is used to traverse all prediction vectors in the vector set to perform inversion solution and fitness evaluation on the prediction area, and obtain the fitness function value of each prediction vector for inversion solution of the prediction area;

[0320] The result of the inversion solution is the predicted terrain data of the predicted area; the value of the fitness function is negatively correlated with the adaptability of the prediction vector for the inversion solution of the predicted area;

[0321] The third module 903 is used to output the corresponding prediction vector as the optimal prediction vector when the fitness function value corresponding to the prediction vector in the vector set is less than the iterative optimal threshold; otherwise,

[0322] The fourth module 904 is used to determine the excellence of each prediction vector according to the fitness function value, perform a random probability selection operation on the prediction vectors in the vector set based on the excellence, and update the vector set according to the selection result of the selection operation; and then use the minimum fitness function value corresponding to the vector set as the second fitness value; the excellence is negatively correlated with the numerical value of the fitness function value;

[0323] The fifth module 905 is used for performing a random probability mutation operation on the prediction vector in the vector set based on the annealing temperature when the second fitness value is greater than or equal to the first fitness value, and updating the vector set according to the mutation result of the mutation operation; otherwise, updating the first fitness value by the second fitness value;

[0324] The sixth module 906 is used to perform a crossover operation with a random probability on the prediction vectors in the vector set based on the second fitness value, and update the vector set according to the crossover result of the crossover operation;

[0325] The seventh module 907 is used to perform temperature decay processing on the annealing temperature based on a preset temperature decay rate, and update the annealing temperature according to the decay result of the temperature decay processing;

[0326] The eighth module 908 is used for returning to the step of performing inversion solution and fitness evaluation of the prediction area by traversing all prediction vectors in the vector set when the annealing temperature is greater than or equal to the preset termination temperature, until the annealing temperature is less than the termination temperature, and outputting the optimal prediction vector according to the last updated vector set;

[0327] The ninth module 909 is used to obtain the corresponding predicted terrain data through inversion according to the optimal prediction vector as the terrain solution result of the prediction area.

[0328] The contents of the method embodiments of the present invention are all applicable to the device embodiments. The functions specifically implemented by the device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0329] On the other hand, an embodiment of the present invention further provides an electronic device, the electronic device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the above-mentioned hydrate stability region bottom boundary prediction method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0330] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0331] like Fig.10 As shown, Fig.10 The hardware structure of an electronic device 1000 of another embodiment is illustrated. The electronic device 1000 includes:

[0332] The processor 1001 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present invention.

[0333] The memory 1002 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1002 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 1002, and the processor 1001 calls and executes the network node population optimization method of the embodiment of the present invention;

[0334] Input / output interface 1003, used to implement information input and output;

[0335] The communication interface 1004 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.);

[0336] A bus 1005 , which transmits information between various components of the device (e.g., the processor 1001 , the memory 1002 , the input / output interface 1003 , and the communication interface 1004 );

[0337] The processor 1001 , the memory 1002 , the input / output interface 1003 and the communication interface 1004 are connected to each other in communication within the device via the bus 1005 .

[0338] The electronic device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0339] The contents of the method embodiments of the present invention are all applicable to the electronic device embodiments. The functions specifically implemented by the electronic device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0340] Another aspect of an embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a program, and the program is executed by a processor to implement the above method.

[0341] The contents of the method embodiments of the present invention are all applicable to the computer-readable storage medium embodiments. The functions specifically implemented by the computer-readable storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0342] The embodiment of the present invention also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the above method.

[0343] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0344] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present invention.

Claims

1. A simulated annealing inversion method for terrain, characterized in that: The following steps are involved: Initializing a vector set, a first fitness value and an initial temperature; using the initial temperature as an annealing temperature; wherein the vector set includes a preset number of prediction vectors; Traversing all the prediction vectors in the vector set to perform inversion solution and fitness evaluation on the prediction area, and obtaining the fitness function value of each prediction vector for performing the inversion solution on the prediction area; The result of the inversion solution is the predicted terrain data of the predicted area; the value of the fitness function value is negatively correlated with the adaptability of the prediction vector for the inversion solution in the predicted area; When the fitness function value corresponding to the prediction vector in the vector set is less than the iterative optimal threshold, the corresponding prediction vector is output as the optimal prediction vector; otherwise, Determine the excellence of each of the prediction vectors according to the fitness function value, perform a random probability selection operation on the prediction vectors in the vector set based on the excellence, and update the vector set according to the selection result of the selection operation; and then use the minimum fitness function value corresponding to the vector set as the second fitness value; the excellence is negatively correlated with the numerical value of the fitness function value; When the second fitness value is greater than or equal to the first fitness value, based on the annealing temperature, performing a random probability mutation operation on the prediction vector in the vector set, and updating the vector set according to the mutation result of the mutation operation; otherwise, updating the first fitness value by the second fitness value; Based on the second fitness value, performing a crossover operation with a random probability on the prediction vectors in the vector set, and updating the vector set according to a crossover result of the crossover operation; Performing temperature decay processing on the annealing temperature based on a preset temperature decay rate, and updating the annealing temperature according to a decay result of the temperature decay processing; When the annealing temperature is greater than or equal to the preset termination temperature, returning to the step of traversing all the prediction vectors in the vector set to perform inversion solution and fitness evaluation of the prediction area, until the annealing temperature is less than the termination temperature, outputting the optimal prediction vector according to the last updated vector set; The predicted terrain data corresponding to the optimal prediction vector obtained through the inversion solution is used as the terrain solution result of the prediction area.

2. The simulated annealing inversion solution method for terrain according to claim 1, characterized in that: The initialization vector set comprises at least one of the following steps: Obtaining a preset standard vector value, and randomly generating a preset number of prediction vectors near the standard vector value; Randomly generating the preset number of prediction vectors within a target interval; the target interval represents a value range of the prediction vector; The preset number of prediction vectors are directly obtained in response to a manual setting instruction.

3. The simulated annealing inversion solution method for terrain according to claim 1, characterized in that: The traversing all the prediction vectors in the vector set to perform inversion solution and fitness evaluation in the prediction area, and obtaining the fitness function value of each prediction vector for performing the inversion solution in the prediction area, comprises the following steps: The prediction vectors in the vector set are traversed in sequence, and the prediction terrain data of the prediction area obtained by inverting each prediction vector using the Parker-Oldenburg inversion method; Based on the predicted terrain data, a root mean square calculation is performed in combination with the seismic depth matrix of the predicted area, and the root mean square value is obtained as the fitness function value of the adaptability of the corresponding prediction vector for the inversion solution in the predicted area.

4. The simulated annealing inversion solution method for terrain according to claim 1, characterized in that: Outputting the corresponding prediction vector as the optimal prediction vector comprises the following steps: When there is a prediction vector in the vector set whose corresponding fitness function value is smaller than the iterative optimal threshold, the prediction vector is directly output as the optimal prediction vector; When there are multiple prediction vectors in the vector set whose corresponding fitness function values ​​are smaller than the iterative optimal threshold, the prediction vector corresponding to the smallest fitness function value is output as the optimal prediction vector.

5. The simulated annealing inversion solution method for terrain according to claim 1, characterized in that: The step of performing a random probability selection operation on the prediction vector in the vector set based on the excellence level comprises the following steps: Normalizing the excellence levels corresponding to all the prediction vectors in the vector set to obtain a normalized excellence level corresponding to each prediction vector; Among them, the expression of the excellence degree is: In the formula, e represents the natural constant; RMS represents the fitness function value; minIterDelt represents the iterative optimal threshold; nowTemp represents the annealing temperature; Arranging the normalized excellence levels corresponding to all the prediction vectors in the vector set to obtain an excellence level array; The value of the nth element in the excellence degree array is the cumulative sum of all the normalized excellence degrees corresponding to the first prediction vector to the nth prediction vector in the vector set; Generate a random number in the first interval as the selection probability; Sequentially traverse the excellence array to find the first element whose value is greater than the selection probability as the target element; Selecting the corresponding prediction vector from the vector set as the update vector according to the element number of the target element, and returning to execute the step of generating a random number in the first interval as the selection probability until the preset number of update vectors are selected from the vector set; The preset number of update vectors is used as the selection result of the selection operation.

6. The simulated annealing inversion solution method for terrain according to claim 1, characterized in that: The prediction vector includes two types of inversion parameters, namely, density difference and depth; and performing a random probability mutation operation on the prediction vector in the vector set based on the annealing temperature includes the following steps: Based on the annealing temperature, the mutation probability is calculated in combination with the first fitness value and the second fitness value; Among them, the expression of the mutation probability is: In the formula, mutationRate represents the mutation probability; e represents the natural constant; RMS1 represents the first fitness value; RMS2 represents the second fitness value; nowTemp represents the annealing temperature; Generate a random number in the second interval as the mutation determination probability; add 1 to the number of mutation cycles; When the mutation determination probability is less than the preset mutation probability, randomly taking one of the prediction vectors from the vector set as a mutation prediction vector; Performing numerical variation on any of the inversion parameters in the variation prediction vector within a target interval, and placing the prediction vector after the numerical variation back into the vector set; the target interval represents the value range of the prediction vector; Returning to the step of generating a random number in the second interval as a mutation determination probability until the number of mutation cycles is equal to the preset number, obtaining the mutation result of the mutation operation; Wherein, the number of mutation cycles is initialized to 0.

7. The simulated annealing inversion solution method for terrain according to claim 1, characterized in that: The prediction vector includes two types of inversion parameters, namely, density difference and depth; and based on the second fitness value, performing a crossover operation of random probability on the prediction vector in the vector set includes the following steps: Based on the second fitness value, the number of crossovers is calculated in combination with the iterative optimal threshold and the preset number; The expression of the crossover number is: In the formula, crossCnt represents the number of crossovers; random(x) represents rounding up x; RMS2 represents the second fitness value; minIterDelt represents the iterative optimal threshold; MaxPredCnt represents the preset number; Generate a random number in the third interval as the crossover determination probability; add 1 to the number of crossover cycles; When the crossover determination probability is less than a preset crossover operator probability, randomly obtaining two of the prediction vectors from the vector set as crossover vectors; the crossover vectors include a first vector and a second vector; Performing numerical weighted crossover on at least one type of the inversion parameters in the first vector and the second vector, and placing the predicted vector after the numerical weighted crossover back into the vector set; Returning to the step of generating a random number as the crossover determination probability within the third interval until the number of crossover cycles is equal to the number of crossovers, and obtaining the crossover result of the crossover operation; The number of crossover cycles is initialized to 0.

8. A simulated annealing inversion solution device for terrain, characterized in that: include: The first module is used to initialize the vector set, the first fitness value and the initial temperature; The initial temperature is used as the annealing temperature; wherein the vector set includes a preset number of prediction vectors; The second module is used to traverse all the prediction vectors in the vector set to perform inversion solution and fitness evaluation on the prediction area, and obtain the fitness function value of each prediction vector for performing the inversion solution on the prediction area; The result of the inversion solution is the predicted terrain data of the predicted area; the value of the fitness function value is negatively correlated with the adaptability of the prediction vector for the inversion solution in the predicted area; The third module is used to output the corresponding prediction vector as the optimal prediction vector when the fitness function value corresponding to the prediction vector in the vector set is less than the iterative optimal threshold; otherwise, A fourth module is used to determine the excellence of each of the prediction vectors according to the fitness function value, perform a random probability selection operation on the prediction vectors in the vector set based on the excellence, and update the vector set according to the selection result of the selection operation; and then use the minimum fitness function value corresponding to the vector set as the second fitness value; the excellence is negatively correlated with the numerical value of the fitness function value; A fifth module is configured to, when the second fitness value is greater than or equal to the first fitness value, perform a random probability mutation operation on the prediction vector in the vector set based on the annealing temperature, and update the vector set according to the mutation result of the mutation operation; otherwise, update the first fitness value by the second fitness value; A sixth module is used to perform a crossover operation with a random probability on the prediction vectors in the vector set based on the second fitness value, and update the vector set according to a crossover result of the crossover operation; A seventh module is used to perform temperature attenuation processing on the annealing temperature based on a preset temperature attenuation rate, and update the annealing temperature according to the attenuation result of the temperature attenuation processing; The eighth module is used for returning to the step of traversing all the prediction vectors in the vector set to perform inversion solution and fitness evaluation of the prediction area when the annealing temperature is greater than or equal to the preset termination temperature, until the annealing temperature is less than the termination temperature, and outputting the optimal prediction vector according to the vector set updated last time; The ninth module is used to obtain the predicted terrain data corresponding to the inversion solution according to the optimal prediction vector as the terrain solution result of the prediction area.

9. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the method according to any one of claims 1 to 7.

10. A computer storage medium storing a program executable by a processor, characterized in that: The program executable by the processor is used to implement the method according to any one of claims 1 to 7 when executed by the processor.

Images