Method, device, computer equipment and storage medium for evaluating effectiveness of adversarial system

Abstract

The present application relates to a method, device, computer equipment and storage medium for evaluating the effectiveness of an adversarial system, which constructs a hierarchical indicator Bayesian stereo network according to system-level indicators, system-level indicators and function-level indicators based on an initial performance indicator sample data set, and uses a Bayesian reasoning mechanism to predict the value of an unknown performance indicator in the same-layer indicator Bayesian network in the network according to known conditions, and uses a Gaussian process regression model to horizontally optimize the predicted unknown performance indicators in each layer of the indicator Bayesian network, and vertically predicts the unknown performance indicators in the upper layer indicator Bayesian network according to the known conditions in the lower layer indicator Bayesian network, until the performance indicator of the top layer indicator Bayesian network is obtained, and the effectiveness of the adversarial system is evaluated according to the performance indicator, and the hierarchical indicator Bayesian stereo network is optimized according to the new indicator sample data and the objective function. The present method can be used for comprehensive performance evaluation.

Description

Technical Field

[0001] The present application relates to the technical field of energy efficiency evaluation of adversarial systems, and in particular to a method, device, computer equipment and storage medium for evaluating the effectiveness of adversarial systems. Background Art

[0002] In the field of modern adversarial system effectiveness assessment, traditional evaluation methods often rely on empirical formulas and qualitative analysis, which perform well when dealing with simple or known conflict scenarios.

[0003] However, with the evolution of confrontation forms and the increasing complexity of the environment, the accuracy and reliability of the evaluation results of these traditional methods have been seriously challenged when faced with changing strategic conditions, multi-dimensional confrontation elements and highly uncertain confrontation environments.

[0004] In addition, although existing machine learning methods, such as support vector machines (SVM) and neural networks, have made breakthroughs in some aspects, they often require a lot of computing resources when processing small sample data, nonlinear relationships, and high-dimensional feature spaces, and the generalization ability of the model is limited. Summary of the invention

[0005] Based on this, it is necessary to provide a method, device, computer equipment and storage medium for evaluating the effectiveness of an adversarial system that can perform comprehensive effectiveness evaluation in response to the above technical problems.

[0006] A method for evaluating the effectiveness of a confrontation system, the method comprising:

[0007] Acquire an initial performance indicator sample data set of the adversarial system, and construct a hierarchical indicator Bayesian stereo network according to the initial performance indicator sample data set according to system-level indicators, system-level indicators and function-level indicators, wherein the hierarchical indicator Bayesian stereo network includes a plurality of layers of indicator Bayesian networks stacked from bottom to top;

[0008] Using the Bayesian inference mechanism, the same-layer indicator Bayesian network in the hierarchical indicator Bayesian stereo network predicts the value of the unknown performance indicator according to known conditions;

[0009] Using the Gaussian process regression model, in the hierarchical indicator Bayesian stereoscopic network, the unknown performance indicators predicted in the Bayesian network of each layer of indicators are horizontally optimized, and the unknown performance indicators in the upper layer of indicator Bayesian network are vertically predicted according to the known conditions in the lower layer of indicator Bayesian network, until the performance indicator of the top layer of indicator Bayesian network is obtained, and the performance of the adversarial system is evaluated according to the performance indicator;

[0010] A new indicator sample data set is obtained, and the hierarchical indicator Bayesian stereo network is optimized according to the new indicator sample data set and the objective function.

[0011] In one embodiment, when constructing a hierarchical index Bayesian stereo network:

[0012] The indicator samples in the initial performance indicator sample data set are defined as top-level indicators, middle-level indicators and bottom-level indicators in sequence according to system-level indicators, system-level indicators and function-level indicators;

[0013] The corresponding top-level indicator Bayesian network, middle-level indicator Bayesian network and bottom-level indicator Bayesian network are constructed according to the top-level indicators, middle-level indicators and bottom-level indicators respectively.

[0014] In one embodiment, when constructing a certain layer of indicator Bayesian network, all indicators in the current layer are used as nodes, and the edges between the corresponding nodes are defined according to the relationship between the indicators;

[0015] The product-moment correlation coefficient is used to measure the influence of the lower-level Bayesian network on the upper-level Bayesian network and define the edges between cross-layer nodes.

[0016] In one embodiment, when constructing a certain layer of indicator Bayesian network, after normalizing the values ​​of all indicators in the current layer, conditional probability distribution is used as the edge between two related nodes.

[0017] In one embodiment, when the Gaussian process regression model is used to horizontally optimize the unknown performance indicators predicted in the Bayesian network of each layer of indicators:

[0018] The conditional probability distribution obtained from the Bayesian network and the probability distribution of the Gaussian process regression are fused to calculate the expectation of the unknown performance index and the following formula is used:

[0019] ;

[0020] In the above formula, X represents the expectation after optimization of an unknown performance indicator, represents the probability distribution predicted by the Bayesian network, and represents the probability distribution obtained through the Gaussian process regression model, where .

[0021] In one embodiment, when optimizing the hierarchical indicator Bayesian stereo network according to the new indicator sample data set and the objective function, it includes:

[0022] The objective function is defined as an abstract description of the performance indicator in the top-level indicator Bayesian network, and the objective function is modeled using Gaussian process regression;

[0023] Based on the objective function through the acquisition function, multiple performance indicators that maximize the objective function value are selected in the multi-layer indicator Bayesian network as reference performance indicators;

[0024] Under the guidance of each of the reference performance indicators, optimizing the value of the unknown performance indicator in the hierarchical indicator Bayesian stereo network to achieve optimization of the hierarchical indicator Bayesian stereo network;

[0025] When optimizing Gaussian process regression based on the new indicator sample data set, a recursive formula is used to reduce the amount of calculation, wherein the recursive formula is expressed as:

[0026] ;

[0027] In the above formula, represents the covariance matrix, represents the Gaussian kernel function, and Respectively represent the indicator sample data set and the new indicator sample data set, Represents the next prediction indicator sample data set. Since there is only one sample point for each prediction, , .

[0028] The present application also provides a device for evaluating the effectiveness of a confrontation system, the device comprising:

[0029] A hierarchical indicator Bayesian stereo network construction module is used to obtain an initial performance indicator sample data set of the adversarial system, and to construct a hierarchical indicator Bayesian stereo network according to system-level indicators, system-level indicators and function-level indicators based on the initial performance indicator sample data set. The hierarchical indicator Bayesian stereo network includes multiple layers of indicator Bayesian networks stacked from bottom to top.

[0030] The unknown performance indicator prediction module in the same-layer indicator Bayesian network is used to predict the value of the unknown performance indicator according to known conditions using the same-layer indicator Bayesian network in the hierarchical indicator Bayesian stereo network by using the Bayesian reasoning mechanism;

[0031] A system performance evaluation module based on Gaussian process regression is used to use the Gaussian process regression model to horizontally optimize the unknown performance indicators predicted in the Bayesian network of each layer of indicators in the hierarchical indicator Bayesian stereo network, and to vertically predict the unknown performance indicators in the upper layer indicator Bayesian network according to the known conditions in the lower layer indicator Bayesian network, until the performance indicator of the top layer indicator Bayesian network is obtained, and the performance of the adversarial system is evaluated according to the performance indicator;

[0032] The hierarchical index Bayesian stereo network optimization module is used to obtain a new index sample data set and optimize the hierarchical index Bayesian stereo network according to the new index sample data set and the objective function.

[0033] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0034] Acquire an initial performance indicator sample data set of the adversarial system, and construct a hierarchical indicator Bayesian stereo network according to the initial performance indicator sample data set according to system-level indicators, system-level indicators and function-level indicators, wherein the hierarchical indicator Bayesian stereo network includes a plurality of layers of indicator Bayesian networks stacked from bottom to top;

[0035] Using the Bayesian inference mechanism, the same-layer indicator Bayesian network in the hierarchical indicator Bayesian stereo network predicts the value of the unknown performance indicator according to known conditions;

[0036] Using the Gaussian process regression model, in the hierarchical indicator Bayesian stereoscopic network, the unknown performance indicators predicted in the Bayesian network of each layer of indicators are horizontally optimized, and the unknown performance indicators in the upper layer of indicator Bayesian network are vertically predicted according to the known conditions in the lower layer of indicator Bayesian network, until the performance indicator of the top layer of indicator Bayesian network is obtained, and the performance of the adversarial system is evaluated according to the performance indicator;

[0037] A new indicator sample data set is obtained, and the hierarchical indicator Bayesian stereo network is optimized according to the new indicator sample data set and the objective function.

[0038] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0039] Acquire an initial performance indicator sample data set of the adversarial system, and construct a hierarchical indicator Bayesian stereo network according to the initial performance indicator sample data set according to system-level indicators, system-level indicators and function-level indicators, wherein the hierarchical indicator Bayesian stereo network includes a plurality of layers of indicator Bayesian networks stacked from bottom to top;

[0040] Using the Bayesian inference mechanism, the same-layer indicator Bayesian network in the hierarchical indicator Bayesian stereo network predicts the value of the unknown performance indicator according to known conditions;

[0041] Using the Gaussian process regression model, in the hierarchical indicator Bayesian stereoscopic network, the unknown performance indicators predicted in the Bayesian network of each layer of indicators are horizontally optimized, and the unknown performance indicators in the upper layer of indicator Bayesian network are vertically predicted according to the known conditions in the lower layer of indicator Bayesian network, until the performance indicator of the top layer of indicator Bayesian network is obtained, and the performance of the adversarial system is evaluated according to the performance indicator;

[0042] A new indicator sample data set is obtained, and the hierarchical indicator Bayesian stereo network is optimized according to the new indicator sample data set and the objective function.

[0043] The above-mentioned confrontation system effectiveness evaluation method, device, computer equipment and storage medium, by constructing a hierarchical indicator Bayesian stereo network composed of multiple layers of indicator Bayesian networks stacked from bottom to top according to the system-level indicators, system-level indicators and function-level indicators according to the initial effectiveness indicator sample data set, using the Bayesian reasoning mechanism, the same-layer indicator Bayesian network in the hierarchical indicator Bayesian stereo network predicts the value of the unknown effectiveness indicator according to the known conditions, using the Gaussian process regression model, in the hierarchical indicator Bayesian stereo network, the predicted unknown effectiveness indicators in the Bayesian network of each layer of indicators are horizontally optimized, and the unknown effectiveness indicators in the upper layer indicator Bayesian network are vertically predicted according to the known conditions in the lower layer indicator Bayesian network, until the effectiveness indicator of the top layer indicator Bayesian network is obtained, and the effectiveness of the confrontation system is evaluated according to the effectiveness indicator, and the hierarchical indicator Bayesian stereo network is optimized according to the new indicator sample data set and the objective function. The method can be used to perform comprehensive effectiveness evaluation, while ensuring the accuracy and flexibility of confrontation effectiveness evaluation, and can dynamically adapt to complex confrontation environments and multi-dimensional uncertainty factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic diagram of a process for evaluating the effectiveness of an adversarial system in one embodiment;

[0045] Figure 2 A schematic diagram of the overall indicator system architecture in a confrontation system in an embodiment;

[0046] Figure 3 A schematic diagram of the structure of a Bayesian network of peer indicators in one embodiment;

[0047] Figure 4 A schematic diagram of the structure of a hierarchical index Bayesian stereo network in one embodiment;

[0048] Figure 5 A schematic diagram of a Gaussian process regression flow chart in one embodiment;

[0049] Figure 6 A schematic diagram of a Bayesian optimization flow chart in one embodiment;

[0050] Figure 7 A flowchart of performance evaluation based on Bayesian network optimization and Gaussian process regression in one embodiment;

[0051] Figure 8 For Figure 7 A flowchart further describing the process;

[0052] Fig. 9 is a structural block diagram of a device for evaluating the effectiveness of an adversarial system in one embodiment;

[0053] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application 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 application and are not used to limit the present application.

[0055] Regarding the existing adversarial system effectiveness evaluation technology, great progress has been made in the evaluation of single adversarial systems, but there are still many deficiencies in the evaluation of complex overall adversarial systems, dynamic changes in actual adversarial environments, and synergy effects. Especially in the scenario of deepening modern informatization, the adversarial system is not only composed of a single equipment platform, but also relies more on the coordinated operations of multiple equipment platforms, command and control, communication networks, and logistics support. These factors influence each other and change dynamically, and existing evaluation technologies are difficult to effectively deal with these complex correlations and uncertainties. In this application, if Figure 1 As shown, a method for evaluating the effectiveness of an adversarial system is provided, which specifically includes the following steps:

[0056] Step S100, obtaining an initial performance indicator sample data set of the adversarial system, and constructing a hierarchical indicator Bayesian stereo network according to the initial performance indicator sample data set according to system-level indicators, system-level indicators and function-level indicators. The hierarchical indicator Bayesian stereo network includes multiple layers of indicator Bayesian networks stacked from bottom to top.

[0057] Step S110, using the Bayesian inference mechanism, the same-layer indicator Bayesian network in the hierarchical indicator Bayesian stereo network predicts the value of the unknown performance indicator according to known conditions.

[0058] Step S120, using the Gaussian process regression model, horizontally optimizes the unknown performance indicators predicted in each layer of the Bayesian network in the hierarchical indicator Bayesian stereo network, and vertically predicts the unknown performance indicators in the upper layer of the Bayesian network according to the known conditions in the lower layer of the Bayesian network, until the performance indicator of the top layer of the Bayesian network is obtained, and the effectiveness of the adversarial system is evaluated based on the performance indicator.

[0059] Step S130, obtaining a new indicator sample data set, and optimizing the hierarchical indicator Bayesian stereo network according to the new indicator sample data set and the objective function.

[0060] In this method, a method for evaluating the effectiveness of system (system) confrontation based on Bayesian network optimization (combined with Gaussian process regression) is proposed. The uncertainty and correlation in the system are modeled using Bayesian networks, and a hierarchical indicator Bayesian stereo network is constructed by first layering and then arranging the Bayesian network by layer. Gaussian process regression (GPR) is used to estimate and optimize unknown factors, including the horizontal estimation of the same-layer indicators and the vertical prediction of bottom-up information transmission. Moreover, when new indicator data is added, the objective function (top-level effectiveness indicator, which represents the entire confrontation effectiveness) is estimated through Bayesian networks and Gaussian process regression to complete the evaluation of the combat effectiveness at that time. In addition, the parameter structure of the pre-constructed Bayesian network and Gaussian process regression is optimized, and the acquisition function is used to determine the acquisition location of the next sample point, which is conducive to the optimization of the overall network. Based on the global correlation analysis of the data, the Bayesian network combines the prediction ability of the Gaussian process to enable the system to adaptively adjust the evaluation strategy in a dynamically changing environment, thereby improving the accuracy and robustness of the combat effectiveness evaluation.

[0061] Furthermore, this method improves the accuracy and adaptability of effectiveness evaluation of complex combat systems through data-driven and statistical inference techniques. Its core lies in the system modeling of Bayesian networks and the dynamic prediction capabilities of Gaussian process regression (GPR). Among them, structured dependency modeling is performed through Bayesian networks, and Gaussian process regression is used to deal with unknown and complex environmental factors. On this basis, combined with Bayesian optimization, the iterative update of the effectiveness evaluation model and the search for the global optimal solution are realized.

[0062] In step S100, the performance indicators in the initial performance indicator sample data set are a series of key performance indicators determined by analyzing the characteristics of the adversarial system. However, not all performance indicators in the initial performance indicator sample data set are known, and there are also uncertain and positional performance indicator values, or only a range of possible values ​​is given. For example, the basic flight capability of a drone can be determined by cruising altitude, cruising speed, flight speed, practical ceiling, endurance time, range, and attitude angle. The target detection capability is determined by the target successful screening probability, target capture time, and target capture probability.

[0063] In this embodiment, the Bayesian network is used to model the performance index of the entire adversarial system.

[0064] Specifically, the Bayesian network can clearly represent the causal relationship between adversarial elements through a directed acyclic graph (DAG), and dynamically update it using prior information and data to adapt to changes in the situation in the adversarial system. This method can effectively handle the complex dependencies between multiple variables and dynamically update the network structure based on known prior knowledge, thereby improving the accuracy and real-time performance of the evaluation.

[0065] In general, when using the Bayesian network to model the effectiveness indicators of the entire adversarial system, some key effectiveness indicators such as attack power, mobility, communication capability, environmental adaptability, etc. are set as nodes of the Bayesian network. That is, each effectiveness indicator corresponds to a node in the Bayesian network. At the same time, these key indicators, that is, the relationship between nodes, are also determined based on experience, historical data or automatic learning. For example, the effectiveness of some adversarial systems may depend on the interaction between attack power and communication capability.

[0066] In this embodiment, when constructing a hierarchical indicator Bayesian stereo network according to the system-level indicators, system-level indicators and function-level indicators, each indicator sample in the initial performance indicator sample data set is defined as the top-level indicator, the middle-level indicator and the bottom-level indicator in turn according to the system-level indicators, the system-level indicators and the function-level indicators, and the corresponding top-level indicator Bayesian network, middle-level indicator Bayesian network and bottom-level indicator Bayesian network are constructed according to the top-level indicators, the middle-level indicators and the bottom-level indicators respectively.

[0067] In one embodiment, an overall indicator system architecture in a confrontation system is provided, such as Figure 2 As shown in the figure, the system-level indicators are system performance, that is, the top-level indicators, and the corresponding top-level indicator Bayesian network is constructed based on this layer of indicators. The system-level indicators include execution performance, guarantee performance, UAV platform performance, command decision-making performance, and experiment completion performance, which are middle-level indicators. Based on this layer of indicators, the middle-level indicator Bayesian network can be generated. Below each system-level indicator are function-level indicators, which will not be discussed here. Figure 2 The functional level indicators in the above are described in detail, and the corresponding underlying indicator Bayesian network can be generated according to each functional level indicator. Figure 2 In the figure, we can also see the relationship between the indicators at each level. The upper-level performance indicators are all affected by the associated lower-level performance indicators.

[0068] In this embodiment, when constructing a certain layer of indicator Bayesian network, all indicators in the current layer are used as nodes, and the edges between corresponding nodes are defined according to the relationship between the indicators.

[0069] Specifically, for indicators of the same level, a correlation matrix is ​​constructed through the Pearson correlation coefficient. Due to the same level, the diagonal elements of the correlation matrix represent the relationship between the indicator and itself, which has no practical meaning. In addition, the correlation matrix is ​​a symmetric matrix. After connecting the indicators of the same level through the correlation matrix, a Bayesian network is constructed based on historical data and expert experience. Bayesian network (BN) is a graphical model based on probabilistic reasoning, which is used to represent the relationship between variables. It consists of a directed acyclic graph and a conditional probability table, where the connection between different indicators has been established, and the direction is determined by expert experience to indicate the causal relationship.

[0070] Furthermore, taking a simple flight tracking task as an example, the same-level indicators include basic flight capability, target detection capability, target tracking capability, target recognition capability, and target loss recovery capability. The performance indicators are defined as random variables The hierarchical Bayesian network initially constructed is as follows Figure 3 shown.

[0071] In this embodiment, the product-moment correlation coefficient is also used to measure the influence of the lower-layer indicator Bayesian network on the upper-layer indicator Bayesian network indicator to define the edges between cross-layer nodes.

[0072] Specifically, the Pearson correlation coefficient (also known as the product-moment correlation coefficient) is used to measure the influence of the lower-level indicators on the upper-level indicators to form a correlation matrix:

[0073] (1)

[0074] In formula (1), Indicates the number of parent indicators, Indicates the number of subordinate indicators. Indicates The first The correlation coefficient between the subordinate indicators is calculated as follows:

[0075] Two indicator series with the same time labels and time scale and It can be viewed as two vectors. Without loss of generality, assuming that the two vectors have been centered, then the angle between them is The cosine of is:

[0076] (2)

[0077] In formula (2), represents the inner product of vectors, Represents the length of a vector.

[0078] In general, the correlation matrix is ​​divided by threshold Transformed into:

[0079] (3)

[0080] In formula (3), , Indicates the index system The first There is no connection between the subordinate indicators. Indicates the index system The first There is a connection relationship between the subordinate indicators. The general direction is that the subordinate indicators point to the superior indicators.

[0081] Before performing subsequent operations, various indicator data are preprocessed. Whether it is a discrete indicator or a continuous indicator, it is limited to the interval [0,1] through the normalization formula:

[0082] (4)

[0083] In formula (4), and are the maximum and minimum values ​​of the indicator respectively.

[0084] In step S110, a Bayesian inference mechanism is used to predict the value of the unknown performance indicator based on known conditions in the same-layer indicator Bayesian network in the hierarchical indicator Bayesian stereo network.

[0085] The Bayesian network structurally represents the dependencies of various factors in the system and can effectively deal with information uncertainty and missing issues. Through Bayesian reasoning, the value of unknown nodes can be predicted based on known conditions. The Bayesian reasoning formula is as follows:

[0086] (5)

[0087] In formula (5), represents an unknown variable in performance evaluation, represents its prior probability, It represents the likelihood function of the variable under the condition of evidence, and the Bayesian network realizes reasoning by joint conditional probability.

[0088] In practical applications, in order to reduce the amount of calculation, the original continuous variable is expressed in discrete form by setting intervals. Since the normalization operation has been performed, the interval of the variable is limited to [0,1]. It is further divided into common n interval, denoted as The original discrete or continuous value variables are represented by the discrete values ​​of the interval to which they belong, thus achieving the unification of discrete values.

[0089] Based on the initial performance index sample data set, the conditional probability distribution table can be calculated by the library function of the statistical software or programming software, such as Figure 3 It should be noted that in the "condition" (for example Figure 3 middle )When confirmed:

[0090] (6)

[0091] This means that in the new round of evaluation, if there are unknown or difficult-to-determine indicators in the same level indicators, the probability distribution of an unknown indicator can be obtained from the conditional probability distribution table in the Bayesian network.

[0092] After step S100 and step S110, on the basis of dividing the indicators into three levels: system level, system level and function level, the Bayesian network of each layer (mainly system level and function level) is constructed based on the correlation matrix and expert experience for the same-level indicators of each layer. For the same-level indicators, black arrows are used to represent the dependency relationship between them, including conditional probability information. For the indicators between the upper and lower levels, gray arrows are used to represent the dependency relationship, without probability information, and the following can be obtained: Figure 4 A relatively complete hierarchical indicator Bayesian stereo network structure is shown.

[0093] Specifically, the expert experience method is a qualitative evaluation method that mainly evaluates research problems by obtaining expert knowledge, and analyzes and predicts based on the wisdom and experience of many experts. For example, when evaluating the credibility of the UAV confrontation system, according to expert experience, it should be evaluated from the perspective of the UAV platform credibility, radar equipment credibility, and data link credibility. For some credibility, only experts can score and obtain weights. In addition, expert experience can determine the relationship between different indicators. For example, experts believe that the ability to recover a lost target may be affected by both the target tracking and recognition capabilities.

[0094] Since the Bayesian reasoning mechanism is limited to predicting unknown performance indicators based on known conditions in the Bayesian network of the same level, and cannot predict unknown performance indicators in the previous layer across levels, but in fact, the energy efficiency indicators in the previous layer will be affected by the related indicators in the next layer, so when the Bayesian reasoning mechanism is used to reason about unknown performance indicators, it is not accurate. Therefore, in step S120, for unknown factors that cannot be directly observed or difficult to model in the Bayesian network, the present invention introduces Gaussian Process Regression (GPR) to perform non-parametric estimation and prediction. GPR has extremely high flexibility and accuracy when processing small sample data or noisy data.

[0095] Gaussian process regression is a powerful non-parametric statistical method that can predict the unknown behavior of a system by learning from existing data when data is scarce or uncertainty is high. GPR does not require assumptions about the precise physical model of the system, but instead builds an effectiveness prediction model in a data-driven way, which is particularly important for dealing with dynamic changes in adversarial systems.

[0096] Furthermore, the core of GPR lies in the covariance function , the function defines two input points and In this embodiment, a kernel function is used to characterize the correlation between different indicators in the system performance evaluation. Commonly used kernel functions include square exponential kernel (RBF kernel) expressed as:

[0097] (7)

[0098] In formula (7), is the length scale, is the signal variance.

[0099] Specifically, given an initial performance indicator sample data set ,in The sample data points are taken as independent variables. is the normalized weight corresponding to different indicators, For (Or ) has a causal relationship with the dependent variable. We hope to use the independent variables and dependent variables in the sample data set to predict the independent variables of the new sample. The corresponding dependent variable . Let the training set sample point set and the prediction sample point set be and , the dependent variable vector in the dataset . Here, since only one sample point needs to be predicted each time, Among them, the dependent variable is the unknown performance indicator, and the independent variable is a known performance indicator, and Gaussian process regression is used to The prediction is:

[0100] (8)

[0101] (9)

[0102] In formula (8) and formula (9), is the noise variance, is the covariance matrix between the new input and the training data, expressed as:

[0103] (10)

[0104] Through the posterior distribution of GPR, the performance evaluation value can be obtained The mean and uncertainty (variance) , thereby providing dynamic prediction and optimization capabilities for performance evaluation.

[0105] Furthermore, noting that only one sample point needs to be predicted, formula (10) can actually be expressed as , at the next prediction sample point When As new training set data, it is necessary to reduce the amount of calculation. Based on the calculation in the previous step, the following formula is given:

[0106] (11)

[0107] (12)

[0108] In formula (11) and formula (12), , , , the most computationally intensive It is given by the following recursive formula:

[0109] (13)

[0110] This will greatly reduce the amount of calculations required when updating the training set data at each step. The update process of Gaussian process regression and the application of the recursive formula are as follows: Figure 5 shown.

[0111] In this embodiment, Gaussian process regression can be applied to the hierarchical Bayesian stereo network obtained in step 110 in the horizontal direction and the vertical direction respectively.

[0112] Specifically, horizontal prediction means that among the indicators of the same level, if there is an unknown factor that cannot be directly observed or difficult to model, if an indicator with a causal relationship with it can be found in the previously constructed Bayesian network ( Figure 4 The black arrows in the figure represent the edges of the Bayesian network. Based on a certain training set of data (which can be obtained from expert experience or rough estimates), for the indicator data that cannot be measured in the new round of performance evaluation, the Gaussian process regression can be used to predict the indicator through the data of other related indicators.

[0113] Specifically, vertical prediction is used to predict the performance of upper-level indicators when the lower-level indicator data is known. Based on the data training set, when a new set of lower-level indicator data (including indicators that are difficult to model through horizontal prediction) is obtained, Figure 4 The middle grey arrow uses Gaussian process regression to transmit the new information from the lower layer to the upper layer for prediction and obtain the corresponding performance evaluation value.

[0114] In this embodiment, when the Gaussian process regression model is used to horizontally optimize the unknown performance indicators predicted in the Bayesian network of each layer of indicators:

[0115] Considering that the previously established Bayesian network and Gaussian process regression can both estimate unknown indicators in the same level indicators, we choose to fuse the data of the two. Figure 3 In As an example, a conditional probability distribution table is obtained through the Bayesian network. , , In the case of The probability distribution of ,in j represents n intervals. After Gaussian process regression, , , Based on the historical data and the values ​​determined in this evaluation, we get Gaussian distribution Since the historical data and evaluation indicators are normalized, the Gaussian distribution should be in the interval [0,1]. According to the intervals divided in the Bayesian network, the probabilities in the intervals are calculated in turn. :

[0116] (14)

[0117] Data fusion is performed by multiplying the two probabilities, normalizing the fused probability and calculating the expected performance of the unknown indicator:

[0118] (15)

[0119] In order to consider both the accuracy of the results and the efficiency of the calculation, when the index changes greatly, choose n =6 is sufficient for a rough estimate. When the calculation accuracy is high, you can choose n =16 or even higher. In general, choose Quite reasonable.

[0120] This completes the estimation of unknown indicators among indicators of the same level.

[0121] Considering that the environment in which the adversarial system is located is dynamically changing, it is necessary to evaluate the performance of the given sample data and also to find the corresponding indicator sample point set that maximizes the performance evaluation. Therefore, every time new indicator sample data is received or selected through the acquisition function, in step S130, the hierarchical indicator Bayesian stereo network is further optimized to keep the objective function value, that is, the final performance evaluation value, at the highest.

[0122] In step S130, Bayesian optimization is used to further optimize the system's adversarial effectiveness. Bayesian optimization is an algorithm that iteratively updates the model to find the global optimal solution. In adversarial effectiveness evaluation, Bayesian optimization continuously optimizes the objective function of effectiveness evaluation through the prediction results of Gaussian process regression to find the optimal combination of evaluation parameters. When optimizing the hierarchical indicator Bayesian stereo network according to the new indicator sample data set and the objective function, it includes: defining the objective function as an abstract description of the effectiveness indicator in the top-level indicator Bayesian network, and modeling the objective function using Gaussian process regression, based on the objective function through the acquisition function, selecting multiple effectiveness indicators that maximize the objective function value in the multi-layer indicator Bayesian network as reference effectiveness indicators, and under the guidance of each reference effectiveness indicator, optimizing the value of the unknown effectiveness indicator in the hierarchical indicator Bayesian stereo network to achieve the optimization of the hierarchical indicator Bayesian stereo network.

[0123] Specifically, the system performance evaluation problem is transformed into an optimization problem, and the objective function Represents the performance evaluation value. Generally speaking, the objective function is an abstract description of the top-level performance indicator.

[0124] Next, use GPR to analyze the objective function Modeling is performed to give the predicted mean and uncertainty of the performance evaluation value. Here we start with the bottom-level indicators and use GPR to predict and evaluate the performance of the upper-level indicators layer by layer, and finally obtain the performance of the top layer, that is, the objective function. Suppose that the predicted performance by Gaussian process regression is Normal distribution .

[0125] In this embodiment, a common acquisition function Expected Improvement (EI) is used, which is expressed as:

[0126] (16)

[0127] In formula (16), It is the current optimal solution.

[0128] Furthermore, formula (16) can be written as:

[0129] (17)

[0130] In formula (17), is the cumulative distribution function (CDF) of the standard normal distribution, and is the probability density function (PDF) of the standard normal distribution.

[0131] Further, take the next assessment sample collection point That is, the value of the reference performance index, expressed as:

[0132] (18)

[0133] However, in actual experiments, it is impossible to collect samples at any point. For example, the target discovery capability may be certain, so we choose to collect samples at the point that satisfies the constraints. The closest to the ideal evaluation sample collection point Collecting at:

[0134] (19)

[0135] It should be noted that there are generally two approaches to acquisition functions: exploration and exploitation. "Exploration" means trying to select points far away from known points as reference points for the next iteration, that is, trying to explore unknown areas, and the distribution of points will be as even as possible. "Exploitation" means trying to select points close to known points as reference points for the next iteration, that is, trying to explore points around known points, and the distribution of points will appear in a dense area, which is easy to enter the local maximum. In practical applications, the EI acquisition function can help us sample in areas with greater uncertainty and greater prediction improvements, thereby effectively balancing the relationship between "exploration" and "exploitation".

[0136] Furthermore, by optimizing the acquisition function, new input points are selected , and continuously iterate and optimize until the optimal performance evaluation parameters are found, that is, the optimal hierarchical indicator Bayesian stereo network is obtained.

[0137] like Figure 6 As shown, it is a schematic diagram of the process of optimizing the hierarchical indicator Bayesian stereo network.

[0138] In this embodiment, the combination of Bayesian network and Gaussian process regression makes the system confrontation effectiveness evaluation have strong adaptive ability. Figure 7-8 As shown. First, based on historical data and the current environment, the Bayesian network is used to infer the initial distribution of the system's effectiveness in the war of resistance, and the dependency relationship of each indicator is preliminarily set to construct a hierarchical indicator Bayesian stereo network (steps S100 and S110). Then, Gaussian process regression is used to predict and estimate unknown factors that are difficult to model. The lower-level indicator information is passed upward by alternating horizontal and vertical predictions, and finally the effectiveness evaluation value is generated (step S120). Furthermore, the effectiveness evaluation model is continuously adjusted through Bayesian optimization to find the optimal effectiveness evaluation parameters. When new data arrives, the Bayesian network and GPR model are updated in real time, so that the system evaluation can adapt to changes and maintain high accuracy (step S130). Among them, Figure 8 Hierarchical Indicator Bayesian Stereo Network Reference Figure 4 , Bayesian network and conditional probability distribution table reference Figure 3 This evaluation process ensures the accuracy and flexibility of the confrontation effectiveness evaluation and can dynamically adapt to complex tactical environments and multi-dimensional uncertain factors.

[0139] In the above-mentioned effectiveness evaluation method of the adversarial system, the Bayesian network is used to model the uncertainty and correlation in the system, and the hierarchical indicator Bayesian stereo network is constructed by first stratifying and then arranging the Bayesian network by layer. Gaussian process regression (GPR) is used to realize the estimation and optimization of unknown factors, including the horizontal estimation of the same-layer indicators and the vertical prediction of bottom-up information transmission. When new indicator data is added, the objective function (top-level effectiveness indicator, which represents the entire combat effectiveness) is estimated through Bayesian network and Gaussian process regression to complete the evaluation of combat effectiveness at that time. In addition, the parameter structure of the pre-constructed Bayesian network and Gaussian process regression is optimized, and the acquisition function is used to determine the acquisition location of the next sample point, which is conducive to the optimization of the overall network. Based on the global correlation analysis of the data, the Bayesian network combines the predictive ability of the Gaussian process to enable the system to adaptively adjust the evaluation strategy in a dynamically changing environment, thereby improving the accuracy and robustness of the combat effectiveness evaluation.

[0140] Compared with the existing technology, this method has higher adaptability: Bayesian network can dynamically process multi-dimensional uncertainty factors, enhancing the flexibility of system evaluation; higher evaluation accuracy: Gaussian process regression can still guarantee high prediction accuracy under small sample and noisy data conditions; less manual intervention: compared with traditional methods that rely on expert experience, the present invention reduces subjective bias through a data-driven approach.

[0141] Furthermore, this method represents the objective function in the form of a random process, and can continuously optimize the prediction of the objective function as the number of evaluations increases. This is a global consideration of the variable and is not prone to falling into a local optimal solution. At the same time, it is not limited by the generalization ability of neural networks, and can fully consider "exploration" or "utilization" when using the acquisition function. Even for unknown sample data, it still has a certain performance evaluation capability.

[0142] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0143] In one embodiment, Fig. 9 As shown, a device for evaluating the effectiveness of an adversarial system is provided, comprising: a hierarchical index Bayesian stereo network construction module 200, an unknown effectiveness index prediction module 210 in the same-layer index Bayesian network, a system effectiveness evaluation module 220 based on Gaussian process regression, and a hierarchical index Bayesian stereo network optimization module 230, wherein:

[0144] A hierarchical indicator Bayesian stereo network construction module 200 is used to obtain an initial performance indicator sample data set of the adversarial system, and to construct a hierarchical indicator Bayesian stereo network according to the system-level indicators, system-level indicators and function-level indicators based on the initial performance indicator sample data set. The hierarchical indicator Bayesian stereo network includes multiple layers of indicator Bayesian networks stacked from bottom to top.

[0145] The unknown performance indicator prediction module 210 in the same-layer indicator Bayesian network is used to predict the value of the unknown performance indicator according to known conditions using the same-layer indicator Bayesian network in the hierarchical indicator Bayesian stereo network by using the Bayesian reasoning mechanism;

[0146] The system performance evaluation module 220 based on Gaussian process regression is used to use the Gaussian process regression model to horizontally optimize the unknown performance indicators predicted in the Bayesian network of each layer of indicators in the hierarchical indicator Bayesian stereo network, and to vertically predict the unknown performance indicators in the upper layer indicator Bayesian network according to the known conditions in the lower layer indicator Bayesian network, until the performance indicator of the top layer indicator Bayesian network is obtained, and the performance of the adversarial system is evaluated according to the performance indicator;

[0147] The hierarchical index Bayesian stereo network optimization module 230 is used to obtain a new index sample data set, and optimize the hierarchical index Bayesian stereo network according to the new index sample data set and the objective function.

[0148] For the specific definition of the adversarial system effectiveness evaluation device, please refer to the definition of the adversarial system effectiveness evaluation method above, which will not be repeated here. Each module in the above-mentioned adversarial system effectiveness evaluation device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0149] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.10 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for evaluating the effectiveness of an adversarial system is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0150] Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0151] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0152] Acquire an initial performance indicator sample data set of the adversarial system, and construct a hierarchical indicator Bayesian stereo network according to the initial performance indicator sample data set according to system-level indicators, system-level indicators and function-level indicators, wherein the hierarchical indicator Bayesian stereo network includes a plurality of layers of indicator Bayesian networks stacked from bottom to top;

[0153] Using the Bayesian inference mechanism, the same-layer indicator Bayesian network in the hierarchical indicator Bayesian stereo network predicts the value of the unknown performance indicator according to known conditions;

[0154] Using the Gaussian process regression model, in the hierarchical indicator Bayesian stereoscopic network, the unknown performance indicators predicted in the Bayesian network of each layer of indicators are horizontally optimized, and the unknown performance indicators in the upper layer of indicator Bayesian network are vertically predicted according to the known conditions in the lower layer of indicator Bayesian network, until the performance indicator of the top layer of indicator Bayesian network is obtained, and the performance of the adversarial system is evaluated according to the performance indicator;

[0155] A new indicator sample data set is obtained, and the hierarchical indicator Bayesian stereo network is optimized according to the new indicator sample data set and the objective function.

[0156] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0157] Acquire an initial performance indicator sample data set of the adversarial system, and construct a hierarchical indicator Bayesian stereo network according to the initial performance indicator sample data set according to system-level indicators, system-level indicators and function-level indicators, wherein the hierarchical indicator Bayesian stereo network includes a plurality of layers of indicator Bayesian networks stacked from bottom to top;

[0158] Using the Bayesian inference mechanism, the same-layer indicator Bayesian network in the hierarchical indicator Bayesian stereo network predicts the value of the unknown performance indicator according to known conditions;

[0159] Using the Gaussian process regression model, in the hierarchical indicator Bayesian stereoscopic network, the unknown performance indicators predicted in the Bayesian network of each layer of indicators are horizontally optimized, and the unknown performance indicators in the upper layer of indicator Bayesian network are vertically predicted according to the known conditions in the lower layer of indicator Bayesian network, until the performance indicator of the top layer of indicator Bayesian network is obtained, and the performance of the adversarial system is evaluated according to the performance indicator;

[0160] A new indicator sample data set is obtained, and the hierarchical indicator Bayesian stereo network is optimized according to the new indicator sample data set and the objective function.

[0161] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0162] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0163] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for evaluating the effectiveness of an adversarial system, characterized in that: The method comprises: Acquire an initial performance indicator sample data set of the adversarial system, and construct a hierarchical indicator Bayesian stereo network according to the initial performance indicator sample data set according to system-level indicators, system-level indicators and function-level indicators, wherein the hierarchical indicator Bayesian stereo network includes a plurality of layers of indicator Bayesian networks stacked from bottom to top; Using the Bayesian inference mechanism, the same-layer indicator Bayesian network in the hierarchical indicator Bayesian stereo network predicts the value of the unknown performance indicator according to known conditions; Using the Gaussian process regression model, the unknown performance indicators predicted in the Bayesian network of each layer of indicators in the hierarchical indicator Bayesian stereoscopic network are horizontally optimized, and the unknown performance indicators in the upper layer of indicator Bayesian network are predicted according to the known conditions in the lower layer of indicator Bayesian network, until the performance indicator of the top layer of indicator Bayesian network is obtained, and the performance of the adversarial system is evaluated according to the performance indicator, wherein the horizontal optimization refers to optimizing the unknown performance indicators predicted by the current layer according to the known performance indicators of the current layer. When the Gaussian process regression model is used to horizontally optimize the unknown performance indicators predicted in the Bayesian network of each layer of indicators: The conditional probability distribution obtained from the Bayesian network and the probability distribution of the Gaussian process regression are fused to calculate the expectation of the unknown performance index and the following formula is used: ； In the above formula, X represents the expectation after optimization of an unknown performance indicator, represents the probability distribution predicted by the Bayesian network, j represents n intervals, and represents the probability distribution obtained through the Gaussian process regression model, where n =8,12; A new indicator sample data set is obtained, and the hierarchical indicator Bayesian stereo network is optimized according to the new indicator sample data set and the objective function. When the hierarchical indicator Bayesian stereo network is optimized according to the new indicator sample data set and the objective function, it includes: defining the objective function with an abstract description of the performance indicator in the top-level indicator Bayesian network, and modeling the objective function using Gaussian process regression; based on the objective function, a plurality of performance indicators that maximize the value of the objective function are selected in the multi-layer indicator Bayesian network as reference performance indicators through an acquisition function; under the guidance of each of the reference performance indicators, the values ​​of unknown performance indicators in the hierarchical indicator Bayesian stereo network are optimized to achieve the optimization of the hierarchical indicator Bayesian stereo network.

2. The method for evaluating the effectiveness of an adversarial system according to claim 1, characterized in that: When constructing a hierarchical indicator Bayesian stereo network: The indicator samples in the initial performance indicator sample data set are defined as top-level indicators, middle-level indicators and bottom-level indicators in sequence according to system-level indicators, system-level indicators and function-level indicators; The corresponding top-level indicator Bayesian network, middle-level indicator Bayesian network and bottom-level indicator Bayesian network are constructed according to the top-level indicators, middle-level indicators and bottom-level indicators respectively.

3. The method for evaluating the effectiveness of an adversarial system according to claim 2, characterized in that: When constructing a Bayesian network of indicators in a certain layer, all indicators in the current layer are used as nodes, and the edges between the corresponding nodes are defined according to the relationship between the indicators; The product-moment correlation coefficient is used to measure the influence of the lower-level Bayesian network on the upper-level Bayesian network and define the edges between cross-layer nodes.

4. The method for evaluating the effectiveness of an adversarial system according to claim 3, characterized in that: When constructing a Bayesian network of indicators at a certain layer, after normalizing the values ​​of all indicators in the current layer, conditional probability distribution is used as the edge between two related nodes.

5. The method for evaluating the effectiveness of an adversarial system according to any one of claims 1 to 4, characterized in that: When optimizing the hierarchical index Bayesian stereo network according to the new index sample data set and the objective function, when optimizing the Gaussian process regression based on the new index sample data set, a recursive formula is used to reduce the amount of calculation, wherein the recursive formula is expressed as: ； In the above formula, represents the covariance matrix, represents the Gaussian kernel function, and Respectively represent the indicator sample data set and the new indicator sample data set, Represents the next prediction indicator sample data set. Since there is only one sample point for each prediction, , .

6. A device for evaluating the effectiveness of an adversarial system, characterized in that: The device comprises: A hierarchical indicator Bayesian stereo network construction module is used to obtain an initial performance indicator sample data set of the adversarial system, and to construct a hierarchical indicator Bayesian stereo network according to system-level indicators, system-level indicators and function-level indicators based on the initial performance indicator sample data set. The hierarchical indicator Bayesian stereo network includes multiple layers of indicator Bayesian networks stacked from bottom to top. The unknown performance indicator prediction module in the same-layer indicator Bayesian network is used to predict the value of the unknown performance indicator according to known conditions using the same-layer indicator Bayesian network in the hierarchical indicator Bayesian stereo network by using the Bayesian reasoning mechanism; The system performance evaluation module based on Gaussian process regression is used to use the Gaussian process regression model to horizontally optimize the unknown performance indicators predicted in the Bayesian network of each layer of indicators in the hierarchical indicator Bayesian stereoscopic network, and predict the unknown performance indicators in the upper layer indicator Bayesian network according to the known conditions in the lower layer indicator Bayesian network until the performance indicator of the top layer indicator Bayesian network is obtained, and the performance of the adversarial system is evaluated according to the performance indicator, wherein horizontal optimization refers to optimizing the unknown performance indicators predicted by the current layer according to the known performance indicators of the current layer. When the Gaussian process regression model is used to horizontally optimize the unknown performance indicators predicted in the Bayesian network of each layer of indicators: The conditional probability distribution obtained from the Bayesian network and the probability distribution of the Gaussian process regression are fused to calculate the expectation of the unknown performance index and the following formula is used: ； In the above formula, X represents the expectation after optimization of an unknown performance indicator, represents the probability distribution predicted by the Bayesian network, j represents n intervals, and represents the probability distribution obtained through the Gaussian process regression model, where n =8,12; A hierarchical indicator Bayesian stereo network optimization module is used to obtain a new indicator sample data set, and optimize the hierarchical indicator Bayesian stereo network according to the new indicator sample data set and the objective function. When optimizing the hierarchical indicator Bayesian stereo network according to the new indicator sample data set and the objective function, it includes: defining the objective function with an abstract description of the performance indicator in the top-level indicator Bayesian network, and modeling the objective function using Gaussian process regression; based on the objective function, through the acquisition function, selecting multiple performance indicators that maximize the value of the objective function in the multi-layer indicator Bayesian network as reference performance indicators; under the guidance of each of the reference performance indicators, optimizing the values ​​of unknown performance indicators in the hierarchical indicator Bayesian stereo network to achieve the optimization of the hierarchical indicator Bayesian stereo network.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images