Electromagnetic radiation source target threat level evaluation method and device, storage medium and equipment

Abstract

The invention discloses an electromagnetic radiation source target threat level assessment method and device, a storage medium and equipment, and belongs to the technical field of radiation source target threat assessment. Processing the target threat element data based on a pre-constructed threat level evaluation model to obtain a threat level reasoning result; according to the threat level reasoning method, Bayesian network modeling advantages under small sample data and efficient search capability under a node sequence space are fully utilized, and meanwhile, a simulated annealing method is creatively improved and then is combined with a Bayesian network structure learning algorithm OBS method under a classic node sequence space; while certain learning efficiency is ensured, the precision of structure learning is improved to a great extent, so that the established radiation source target threat level evaluation model can better adapt to a radiation source target threat level evaluation task in a complex electromagnetic environment.

Description

Technical Field

[0001] The present application relates to the technical field of radiation source target threat assessment, and in particular to an electromagnetic radiation source target threat level assessment method, device, storage medium and equipment. Background Art

[0002] Under the current complex electromagnetic environment, there are radar, communication and other signals of various types, quantities and threat levels from the ground or air in the region, and there are also electronic countermeasures loads distributed on various manned and unmanned heterogeneous platforms. The distributed electromagnetic application mode has become an indispensable choice with its unique advantages and high efficiency. It achieves accurate reconnaissance and interference of electronic equipment by distributing electromagnetic resources on multiple nodes. An efficient and reliable electromagnetic resource allocation strategy is an important part of distributed electromagnetic applications, which can promote more accurate and agile electromagnetic resource utilization. The environmental situation often changes rapidly. Radar and communication targets will adjust their work tasks and working modes according to real-time conditions. If the assigned method is continued, it will be impossible to adjust the resource allocation according to the real-time situation, resulting in poor electromagnetic suppression effect; in addition, when the number of radiation sources is large, the assigned method is inefficient, inaccurate and unreliable, and it is impossible to assign the best interference resource allocation method. In engineering, it is necessary to use an autonomous method to determine the allocation of collaborative interference resources and adjust them in time according to changes in the situation, so that the electromagnetic countermeasures can be maximized at all times. Effective threat level assessment of radiation source targets can help decision makers understand the power comparison between the two sides, identify potential dangers in a complex and changing environment, and make quick and wise resource allocation decisions. How to do a good job in target threat level assessment and do the preliminary work for making correct resource allocation decisions is an urgent problem to be solved in the command and control of electronic countermeasure systems.

[0003] Scholars at home and abroad have already conducted extensive research on the issue of threat level assessment. At present, there are mainly two types of radiation source target threat level assessment methods:

[0004] (1) Construct theoretical models based on expert knowledge. By establishing a mathematical model of the threat level assessment index of the radiation source target, the threat level of the target can be quantified. Commonly used theoretical models include hierarchical analysis method and fuzzy set theory. These methods rely on expert experience and simple mathematical models, so the evaluation results are highly subjective, have weak interpretability, and have limited ability to process uncertain information. These methods have obvious limitations when facing complex electromagnetic environments;

[0005] (2) Build an artificial intelligence model based on data, and regard threat level assessment as a nonlinear mapping problem from threat attributes to threat level assessment results. Typical intelligent algorithms include neural networks, support vector machines, etc. The radiation source target threat level assessment method based on neural networks is currently a more popular method. It is based on threat attributes and their corresponding threat level label data, and uses algorithms such as back propagation to learn neural network parameters and then build a network model. However, it cannot effectively combine existing expert knowledge in the process of establishing the radiation source target threat level assessment model. In order to ensure sufficient model accuracy, it requires a large amount of sample data.

[0006] Bayesian Network (BN) was first proposed by Pearl in the 1980s. Its theoretical basis is graph theory and probability theory. It can describe the causal relationship between variables, thereby helping people to effectively mine potential information in data. It has received widespread attention in recent years. As a powerful tool for uncertainty reasoning and data analysis, Bayesian Network has the following characteristics:

[0007] (1) Bayesian networks organically combine graph theory and probability theory, intuitively presenting the causal relationship between variables in a graphical way and quantifying these relationships in a probabilistic form.

[0008] (2) Bayesian networks have a solid mathematical foundation, and their inference results have profound theoretical support.

[0009] (3) Expert experience and prior information can be used to construct the Bayesian network, so that existing information can be effectively combined with sample data to improve modeling accuracy.

[0010] (4) Bayesian networks are good at handling small sample data modeling problems and can learn the probabilistic relationship between variables. Even when the data is incomplete, they can still achieve good modeling results.

[0011] Because of the unique advantages of Bayesian networks, they can be fully applied in threat level assessment. The research on Bayesian networks covers structural learning, parameter learning and reasoning. Structural learning determines the network structure that best fits the sample data set based on sample data and the prior knowledge of experts; parameter learning calculates the conditional probability of nodes based on the structure and data; reasoning solves the marginal or maximum probability problems of variables based on the structure and parameters. Among them, structural learning is crucial. Inaccurate structural learning will lead to incorrect causal relationships, resulting in incorrect threat level assessment results. These errors cannot be effectively compensated by parameter learning, and may even increase its burden. Summary of the invention

[0012] The main purpose of this application is to provide a method, device, storage medium and equipment for assessing the threat level of an electromagnetic radiation source target, aiming to construct an accurate Bayesian network structure using a small amount of sample data.

[0013] To achieve the above-mentioned purpose, the present application provides a method for assessing the target threat level of an electromagnetic radiation source, comprising: obtaining target threat element data; processing the target threat element data based on a pre-constructed threat level assessment model to obtain a threat level inference result, wherein a Bayesian network structure is learned from sample data based on a pre-constructed simulated annealing structure learning algorithm, and network parameters are learned from the Bayesian network structure based on a maximum likelihood estimation algorithm, the threat level assessment model is constructed based on the Bayesian network structure and the network parameters, and the simulated annealing structure learning algorithm is jointly constructed based on an improved simulated annealing algorithm and an improved OBS algorithm; and the threat level inference result is output.

[0014] Optionally, before processing the target threat factor data based on the pre-built threat level assessment model, the method further includes obtaining an initial node sequence based on the discretized target threat factor data.

[0015] Optionally, the pre-built simulated annealing-based structural learning algorithm learns the Bayesian network structure from sample data, including: determining the initial conditions of the threat level assessment model, the initial conditions including the initial temperature value, the termination temperature value, the annealing rate, the minimum tolerated number of searches without improvement, the initial value of the number of invalid searches, the initial node sequence and the current optimal node sequence; using the improved OBS algorithm to iteratively search from the initial node sequence to obtain a local optimal node sequence, and obtaining a new node sequence as the initial node sequence for the next round of improved OBS method search; using a scoring function to process the local optimal node sequence and the current optimal node sequence respectively, and correspondingly obtaining a first score and a second score; if the first score is greater than the second score, using the local optimal node sequence to update the current optimal node sequence, and resetting the number of invalid searches to zero; processing the optimal node sequence through the K2 algorithm to obtain the Bayesian network structure.

[0016] Optionally, after using the scoring function to process the local optimal node sequence and the current optimal node sequence respectively and obtaining the first score and the second score respectively, the method further includes: if the first score is less than the second score, the current optimal node sequence is not updated, and the number of invalid searches is increased by one to obtain a new number of invalid searches; based on the new node sequence as the start of the next search, and based on the first expression to update the current temperature value, continue to iteratively search the local optimal node sequence and the new node sequence based on the improved OBS algorithm, and perform subsequent operations until the iteration is terminated; calculate the iteration termination parameter value, terminate the iteration according to the new number of invalid searches, the iteration termination parameter value, the current temperature value and the termination temperature value, and output the current optimal node sequence.

[0017] Optionally, the improved OBS algorithm is used to iteratively search from the initial node sequence to obtain a local optimal node sequence, and a new node sequence is obtained as the initial node sequence for the next round of improved OBS method search, including: during the search process, if the score of the node sequence obtained by an operation in a round of search is greater than the score of the current node sequence, then this node sequence is accepted to continue the search; if the score of the node sequence obtained by an operation in a round of search is less than the score of the current node sequence, then based on the Metropolis criterion, the node sequence is selected to be accepted or abandoned; if the node sequence is selected to be accepted, the scores of the current node sequence and the current optimal node sequence are compared, and if the score of the current node sequence is greater than the score of the current optimal node sequence, the current node sequence is stored in a pre-built memory module; if the node sequence is selected to be abandoned, the search is continued based on the current node sequence.

[0018] Optionally, the calculating of the iteration termination parameter value, terminating the iteration according to the new number of invalid searches, the iteration termination parameter value, the current temperature value and the termination temperature value, and outputting the current optimal node sequence includes: calculating the iteration termination parameter value based on a second expression, wherein the second expression is:

[0019]

[0020] Among them, α min represents the minimum number of tolerated searches without improvement, T i Indicates that the current temperature value is obtained after each iteration, T f Represents the termination temperature value; if the current value of the invalid search times is greater than the iteration termination parameter value, or the current temperature value is greater than the termination temperature value, the iteration is stopped and the current optimal node sequence is output.

[0021] Optionally, the updating of the current temperature value based on the first expression, continuing to iteratively search the optimal node sequence and the new node sequence based on the improved OBS algorithm, and performing subsequent operations until the termination iteration is reached, includes: if the current value of the invalid search times is less than the iteration termination parameter value, and the current temperature value is less than the termination temperature value, updating the current temperature based on the first expression and outputting the new node sequence, wherein the first expression is:

[0022] T c =λT c

[0023] Where λ represents the annealing rate, T c Represents the current temperature output after each iteration; if the current value of the invalid search times is greater than the iteration termination parameter value, or the current temperature value is greater than the termination temperature value, the iteration is stopped and the optimal node sequence is output; if the current value of the invalid search times is less than the iteration termination parameter value, and the current temperature value is less than the termination temperature value, the current temperature value of the next iteration is obtained based on the product of the annealing rate and the current temperature value, and the new node sequence is used as the initial node sequence of the next iteration, and the optimal node sequence is continued to be iteratively searched based on the improved OBS algorithm and the new node sequence is output until the preset judgment condition is met, and the preset judgment condition is that the current value of the invalid search times is greater than the iteration termination parameter value, or the current temperature value is greater than the termination temperature value.

[0024] To achieve the above-mentioned purpose, the present application also provides an electromagnetic radiation source target threat level assessment device, including: an acquisition module, used to obtain target threat element data; a processing module, used to process the target threat element data based on a pre-built threat level assessment model to obtain a threat level reasoning result, wherein a Bayesian network structure is learned from sample data based on a pre-built simulated annealing structure learning algorithm, and network parameters are learned from the Bayesian network structure based on a maximum likelihood estimation algorithm, and the threat level assessment model is constructed based on the Bayesian network structure and the network parameters, and the simulated annealing structure learning algorithm is jointly constructed based on an improved simulated annealing algorithm and an improved OBS algorithm; an output module, used to output the threat level reasoning result.

[0025] To achieve the above objectives, the present application also provides a computer-readable storage medium, which includes instructions, which, when executed on a computer, enables the computer to execute the electromagnetic radiation source target threat level assessment method described in any one of the aforementioned embodiments.

[0026] To achieve the above object, the present application also provides an electronic device, the electronic device comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus.

[0027] a memory for storing a computer program,

[0028] A processor, wherein the processor is used to execute a computer program stored in a memory to implement the electromagnetic radiation source target threat level assessment method in any one of the aforementioned embodiments of the claims.

[0029] The embodiments of the present application propose a method, device, storage medium and equipment for assessing the threat level of an electromagnetic radiation source target, which obtains target threat factor data; processes the target threat factor data based on a pre-constructed threat level assessment model to obtain a threat level inference result, wherein a Bayesian network structure is learned from sample data based on a pre-constructed simulated annealing structure learning algorithm, and network parameters are learned from the Bayesian network structure based on a maximum likelihood estimation algorithm, and the threat level assessment model is constructed based on the Bayesian network structure and the network parameters, and the simulated annealing structure learning algorithm is jointly constructed based on an improved simulated annealing algorithm and an improved OBS algorithm; and the threat level inference result is output. The present invention makes full use of the advantages of Bayesian network modeling under small sample data and the efficient search capability under node order space, improves the efficiency of Bayesian network structure learning, and enables the established radiation source target threat level assessment model to better adapt to the radiation source target threat level assessment task under complex electromagnetic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a flow chart of an embodiment of a method for assessing the target threat level of an electromagnetic radiation source of the present application;

[0031] Figure 2 A threat level assessment flow chart of an embodiment of a method for assessing the threat level of an electromagnetic radiation source target of the present application;

[0032] Figure 3 This is a flow chart of a cluster tree propagation algorithm of an embodiment of a method for assessing the threat level of an electromagnetic radiation source target of the present application;

[0033] Figure 4 A Bayesian network structure diagram based on the SAOBS algorithm learned in an embodiment of the electromagnetic radiation source target threat level assessment method of the present application;

[0034] Figure 5 A structural block diagram of an embodiment of the electromagnetic radiation source target threat level assessment of the present application.

[0035] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0036] 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.

[0037] The present invention also creatively improves the simulated annealing method and combines it with the Bayesian network structure learning algorithm OBS algorithm in the classic node order space. While ensuring a certain learning efficiency, it greatly improves the accuracy of Bayesian network structure learning, thereby making the results of radiation source target threat level assessment more convincing.

[0038] Faced with a complex electromagnetic environment with a large electromagnetic frequency band span, complex multipath propagation, multiple types of frequency-using equipment, overlapping signals in time and space, and cross-actions, establishing a radiation source target threat level assessment model often faces the situation of small sample size and high data dimension, which makes it difficult to ensure the accuracy and timeliness of the radiation source target threat level assessment results, and cannot meet the needs of precise and flexible electromagnetic confrontation under complex electromagnetic backgrounds. In order to address this problem, the present invention proposes a radiation source target threat level assessment method based on Bayesian network learning in node order space, and establishes a radiation source target threat level assessment model by combining the modeling advantages of Bayesian network under a small amount of sample data. In this process, a simulated annealing structure learning algorithm is innovatively proposed and applied.

[0039] The simulated annealing algorithm and the OBS algorithm are introduced below:

[0040] Simulated Annealing (SA) is a heuristic optimization algorithm inspired by the physical phenomenon that solids reach a stable state during cooling. i Start searching, T i Gradually decreases over time. If the current state is S i , whose internal energy is E i , at the next temperature point T j , if the state S j Internal energy E j <E i , then S j will be considered as the new current state; if E j >E i , then it will probabilistically decide whether to accept S j This new state, the acceptance probability function is:

[0041]

[0042] This acceptance criterion is called the Metropolis criterion. In general, the simulated annealing algorithm usually follows the following steps:

[0043] Set initial temperature T i , final temperature T f , annealing rate λ and the number of iterations L at each temperature, randomly generate an initial state S and set it as the current best solution, that is, S * =S;

[0044] Repeat (3) and (4) for k = 1, ..., L;

[0045] Generate a new state S' and calculate its internal energy ΔE = E(S') - E(S);

[0046] If ΔE≤0, set S' to the current state and check whether the best solution needs to be updated, that is, E(S * )-E(S') is greater than 0. If so, update it. When ΔE>0, decide whether to accept the new state S' according to the Metropolis criterion. If accepted, update the current state and detect the optimal solution; if not accepted, keep the current state unchanged; determine whether the current temperature reaches the termination temperature T f , if reached, the algorithm ends and outputs the optimal solution S * If the termination temperature is not reached, continue to cool down until T i+1 =λT i , return to (2).

[0047] The OBS algorithm is a simple and efficient method for learning the Bayesian network structure in the node order space. It explores the node order space by using the exchange operator and determines the Bayesian network structure accordingly. The exchange operator is defined as:

[0048]

[0049] The OBS exchange operator explores the neighborhood of the current node order by exchanging the positions of adjacent variables. After each exchange, the algorithm compares the scores of the new node order with the original node order (the score here refers to the quantitative value of the degree of matching between the network structure and the data, which is the BIC score in this invention), and selects the node order with the higher score to continue searching until the local optimum is found. In the new node order, except for the changed node X i and X i+1 The parent node sets of the remaining nodes do not change, so only X needs to be recalculated. i and X i+1 The fraction after the change of the parent set.

[0050] Reference Figure 1 , Figure 1This is a flow chart of the electromagnetic radiation source target threat level assessment method of the present application. The electromagnetic radiation source target threat level assessment method may include:

[0051] S10, obtaining target threat factor data;

[0052] In an embodiment of the present application, before step S10, the electromagnetic radiation source target threat level assessment method may further include the following execution process:

[0053] The initial node sequence is obtained based on the discretized target threat factor data.

[0054] Specifically, Bayesian networks usually need to discretize data during processing. Qualitative indicators can be discretized based on experience, while quantitative indicators need to select appropriate discretization methods to avoid reasoning errors. In this invention, we use the K-means algorithm to discretize data. The K-means algorithm is an effective and simple clustering method that clusters data into K clusters through iteration, with the aim of making data points in the same cluster close to each other, while data points in different clusters are far away from each other. The specific steps of the algorithm are as follows:

[0055] (1) Randomly select K data points as the starting cluster centers;

[0056] (2) Each data point is assigned to the cluster where the nearest cluster center is located. The commonly used distance measurement method here is the Euclidean distance, which is as follows:

[0057]

[0058] (3) Recalculate the cluster center of each cluster, that is, the mean of all data points in the cluster;

[0059] (4) Repeat (2) and (3) until the cluster center no longer changes significantly or the preset number of iterations is reached, and the discretized target threat factor data is obtained, and the initial node sequence is obtained based on the target threat factor data.

[0060] S20. Processing target threat factor data based on a pre-constructed threat level assessment model to obtain a threat level inference result, wherein a structure learning algorithm based on a pre-constructed simulated annealing learns a Bayesian network structure from sample data, and learns network parameters from the Bayesian network structure based on a maximum likelihood estimation algorithm, and converts the network parameters into a connection tree according to a clique propagation algorithm to obtain a threat level assessment model, and the structure learning algorithm of simulated annealing is jointly constructed based on an improved simulated annealing algorithm and an improved OBS algorithm;

[0061] Among them, after discretizing the sample data, the Bayesian network model is established. The establishment of the Bayesian network model is mainly divided into two stages: structure learning and parameter learning. First, the Bayesian network structure is learned from the sample data using the SAOBS algorithm. Then, the network parameters are learned based on the network structure using the maximum likelihood estimation method (MLE). The MLE algorithm judges the degree of fit between the parameters and the model through the likelihood of the samples and parameters. The likelihood of the parameter refers to the conditional probability P(D|θ) of the data D given the parameter θ. The conditional probability is the likelihood of the parameter θ, which is recorded as L(θ|D)=P(D|θ). The larger the conditional probability, that is, the larger the likelihood, the higher the degree of fit between the parameter θ and the data D. At this point, a complete Bayesian network is established, which is the threat level assessment model used to infer the threat level of the target.

[0062] Specifically, step S20 may include the following execution process:

[0063] S201, determining the initial conditions of the threat level assessment model, the initial conditions including an initial temperature value, a termination temperature value, an annealing rate, a minimum tolerable number of non-boosting searches, an initial value of the number of invalid searches, an initial node sequence, and a current optimal node sequence;

[0064] S202, using the improved OBS algorithm to iteratively search from the initial node sequence to obtain a local optimal node sequence, and obtain a new node sequence as the initial node sequence for the next round of improved OBS method search;

[0065] Among them, the improved OBS algorithm is used to iteratively search from the initial node sequence to obtain the local optimal node sequence, and the new node sequence is obtained as the initial node sequence for the next round of improved OBS method search, including:

[0066] During the search process, if the score of the node sequence obtained by an operation in a round of search is greater than the score of the current node sequence, then this node sequence is accepted and the search continues;

[0067] If the score of the node sequence obtained by an operation in a round of search is less than the score of the current node sequence, the node sequence is selected to be accepted or abandoned based on the Metropolis criterion;

[0068] If the node sequence is chosen to be accepted, the scores of the current node sequence and the current optimal node sequence are compared. If the score of the current node sequence is greater than the score of the current optimal node sequence, the current node sequence is stored in a pre-built memory module.

[0069] If you choose to abandon the node sequence, continue searching based on the current node sequence.

[0070] S203, using a scoring function to process the local optimal node sequence and the current optimal node sequence respectively, and obtain a first score and a second score respectively;

[0071] S204, if the first score is greater than the second score, the current optimal node sequence is updated using the local optimal node sequence, and the number of invalid searches is reset to zero;

[0072] In an embodiment of the present invention, after step S203, the electromagnetic radiation source target threat level assessment method may further include the following execution process:

[0073] If the first score is less than the second score, the current optimal node order is not updated, and the invalid search count is increased by one to obtain a new invalid search count;

[0074] The new node sequence is used as the start of the next search, and the current temperature value is updated based on the first expression. The local optimal node sequence and the new node sequence are iteratively searched based on the improved OBS algorithm, and subsequent operations are performed until the termination iteration is reached.

[0075] Wherein, updating the current temperature value based on the first expression, continuing to iteratively search for the optimal node sequence and the new node sequence based on the improved OBS algorithm, and performing subsequent operations until the termination iteration is reached include:

[0076] If the current value of the invalid search times is less than the iteration termination parameter value, and the current temperature value is less than the termination temperature value, the current temperature is updated based on the first expression and a new node sequence is output, where the first expression is:

[0077] T c =λT c

[0078] Where λ represents the annealing rate, T c Represents the current temperature output after each iteration;

[0079] If the current value of the invalid search times is greater than the iteration termination parameter value, or the current temperature value is greater than the termination temperature value, the iteration is stopped and the optimal node sequence is output;

[0080] If the current value of the invalid search times is less than the iteration termination parameter value, and the current temperature value is less than the termination temperature value, the current temperature value of the next iteration is obtained based on the product of the annealing rate and the current temperature value, and the new node sequence is used as the initial node sequence of the next iteration, and the optimal node sequence is continued to be iteratively searched based on the improved OBS algorithm and the new node sequence is output until the preset judgment condition is met. The preset judgment condition is that the current value of the invalid search times is greater than the iteration termination parameter value, or the current temperature value is greater than the termination temperature

[0081] Calculate the iteration termination parameter value, terminate the iteration according to the new invalid search times, iteration termination parameter value, current temperature value and termination temperature value, and output the current optimal node sequence;

[0082] Among them, the iteration termination parameter value is calculated, the iteration is terminated according to the new invalid search times, the iteration termination parameter value, the current temperature value and the termination temperature value, and the current optimal node sequence is output, including:

[0083] The iteration termination parameter value is calculated based on the second expression, which is:

[0084]

[0085] Among them, α min represents the minimum number of tolerated searches without improvement, T i Indicates the current temperature value after each iteration, T f Indicates the termination temperature value;

[0086] If the current value of the invalid search times is greater than the iteration termination parameter value, or the current temperature value is greater than the termination temperature value, the iteration is stopped and the current optimal node sequence is output.

[0087] S205. Process the optimal node sequence through the K2 algorithm to obtain a Bayesian network structure.

[0088] The embodiment of the present application proposes a method for assessing the threat level of an electromagnetic radiation source target, by acquiring target threat factor data; processing the target threat factor data based on a pre-constructed threat level assessment model to obtain a threat level reasoning result, wherein a structure learning algorithm based on a pre-constructed simulated annealing learns a Bayesian network structure from sample data, and learns network parameters from the Bayesian network structure based on a maximum likelihood estimation algorithm, and a threat level assessment model is constructed based on the Bayesian network structure and the network parameters, and the structure learning algorithm of simulated annealing is jointly constructed based on an improved simulated annealing algorithm and an improved OBS algorithm; and the threat level reasoning result is output. The present invention makes full use of the advantages of Bayesian network modeling under small sample data and the efficient search capability under node order space, improves the efficiency of Bayesian network structure learning, and enables the established radiation source target threat level assessment model to better adapt to the radiation source target threat level assessment task under complex electromagnetic environments.

[0089] For example, simulated annealing is good at breaking through local optimality, but it is time-consuming; while the OBS algorithm provides a simple and effective search operator in local search, but it may fall into local optimality. These two methods are complementary, and combining them can produce an optimization algorithm that balances efficiency and accuracy. In the present invention, the idea of ​​probabilistic acceptance is applied to the OBS algorithm, and the Metropolis criterion is used to decide whether to accept a new node order. At this time, the Metropolis criterion can be defined as:

[0090]

[0091] in, is the current node sequence, yes The node order obtained after applying the exchange operator once, T c is the current temperature value.

[0092] For the current node sequence (Old node sequence), its score is:

[0093]

[0094]

[0095] Swap X k and X k+1 Afterwards, the new node sequence The rating is:

[0096]

[0097] In the formula, X k and X k+1 The family score is:

[0098]

[0099] However, the introduction of the Metropolis criterion may lead to the selection of a poor node order in the local search process, thus losing the optimal solution. Therefore, a memory module is set to save the previous optimal solution. In OBS, this memory module is named It represents the optimal solution within a stage, in which the node order score obtained by each search is higher than the current node order score, and is only updated at the end of this stage (a worse solution is accepted).

[0100] Consider the score gap between two node sequences:

[0101]

[0102] if It is established, indicating that a better node order has been found. Remain unchanged from Continue searching. If is a poor solution, It will be chosen or abandoned probabilistically according to the Metropolis criterion, rather than being abandoned directly:

[0103] If a bad node sequence is chosen probabilistically Then through Evaluate To decide whether to update, if renew Then from Continue local search.

[0104] When a poor node sequence When abandoned, Remain unchanged, search based on the current node order Keep going.

[0105] By tolerating suboptimal solutions, the OBS algorithm can largely get rid of the influence of local optimal solutions. At the same time, under the guidance of the Metropolis criterion, the quality of the initial node order has a weaker impact on the entire search. The present invention refers to this improved OBS algorithm as the MOBS algorithm.

[0106] In order to more efficiently combine simulated annealing and OBS algorithm for Bayesian network structure learning, SA in the present invention is also improved accordingly. First, the iteration termination condition is designed based on the number of consecutive invalid local searches. Here, invalid local search refers to the local search that fails to find a higher score than the best node sequence in the current stage. Let the number of consecutive invalid searches be count invalid , set the iteration termination parameter α, which is a dynamic adaptive parameter, as shown in the following formula (second expression):

[0107]

[0108] Where α min represents the minimum number of tolerated searches without improvement, T i represents the temperature at the i-th iteration, T f Indicates the end temperature. invalid >α, the local search ends; once a solution with a higher score than the current optimal node order is found, count invalid Reset to 0.

[0109] The present invention refers to this new algorithm as the SAOBS algorithm, and its basic steps for learning the Bayesian network structure are as follows:

[0110] Initialization parameters: initial temperature T0, end temperature T f , annealing rate λ and minimum number of tolerable searches without boosting α min , count of consecutive invalid searches invalid Initialize to 0; randomly generate an initial sequence And set it as the optimal node order, that is Current temperature T c Initialize to T0;

[0111] Will As a new node sequence and other parameters as input to the MOBS algorithm to run the search to obtain a new node sequence and local optimal order like The optimal node order Reset count invalid ;otherwise No change, at the same time, update count invalid Value: count invalid =count invalid +1.

[0112] According to formula (8), calculate α. If count invalid >α or current temperature T c <T f , then the algorithm search stops and the output Otherwise, update the parameter T c =λT c And return to step (2), and As the initial sequence for the next MOBS algorithm.

[0113] The optimal node sequence is obtained by K2 algorithm. converted into a network structure.

[0114] The SAOBS algorithm incorporates the idea of ​​simulated annealing to probabilistically accept poor solutions, so that the algorithm no longer stagnates in the local optimum during the search process, but continues to seek breakthroughs, effectively expanding the search range and improving learning accuracy. At the same time, it also retains the simple and efficient search method of the OBS algorithm, ensuring a certain learning efficiency. In addition, a memory function is added to the algorithm to record the current optimal solution when accepting a poor solution, avoiding the loss of possible global optimal solutions. In view of the large number of redundant search problems in simulated annealing itself, it is proposed to set the iteration termination condition based on the number of consecutive decreases of the scoring function during the search process to improve the accuracy and efficiency of Bayesian network modeling, thereby improving the performance of the overall radiation source target threat level assessment algorithm.

[0115] Through the above-mentioned structure learning and parameter learning, a complete Bayesian network has been obtained. Based on this, the target threat value can be calculated by Bayesian network reasoning. In the present invention, the cluster propagation algorithm is used to infer the following posterior probability. The process of the cluster tree propagation algorithm is as follows: Figure 3As shown in the figure, it first converts the Bayesian network into a connection tree. This process involves assigning the conditional probability table of the Bayesian network (i.e., the network parameters obtained by parameter learning) to the corresponding cluster nodes of the connection tree, that is, completing the initialization of the connection tree. Once the connection tree is initialized, it will start the message passing process, and make the connection tree reach global consistency through message exchange between cluster nodes. When the connection tree reaches global consistency, the probability distribution of any node can be calculated. When a new evidence variable is introduced, the message passing between cluster nodes must be started again to ensure that the connection tree reaches global consistency again.

[0116] P(Y=y i |X1=x1,X2=x2,...,X n =x n )

[0117] Where Y represents the threat level of a target, X1, X2, X3, ..., X n are the parameters of the target, such as target type, target speed, target intention, etc. The posterior probability of the above formula means that a target has various parameters X1=x1,X2=x2,...,X n =x n , at this time, its threat level is y i After the probability values ​​of each threat level of each target in its own state are inferred, the probability values ​​of the highest threat level are compared. If they are the same, the next level is compared. In this way, the threat level of each target is ranked.

[0118] S30: Output the threat level reasoning result.

[0119] The beneficial effects of the present invention are:

[0120] (1) The present invention applies the Bayesian network learning theory in node order space to the field of radiation source target threat level assessment. While utilizing the data modeling advantage of the Bayesian network itself under a small number of samples, it also effectively improves the structural learning efficiency through the search advantage in node order space, thereby enabling the established radiation source target threat level assessment model to better adapt to the threat level assessment task in a complex electromagnetic environment.

[0121] (2) The present invention creatively improves the simulated annealing method and combines it with the Bayesian network structure learning algorithm OBS method in the classical node order space. While ensuring a certain learning efficiency, it greatly improves the accuracy of structure learning, thereby making the results of radiation source target threat level assessment more convincing.

[0122] In summary, the present invention proposes a radiation source target threat level assessment method based on Bayesian network learning in node order space, which can accurately and effectively analyze and infer the size of the radiation source target threat level.

[0123] The present invention is further described below in conjunction with the accompanying drawings and through the threat level assessment process in the example. In this example, the electronic countermeasure system is composed of a command and control system and three electronic countermeasure loads. The command and control system is the core of the electronic countermeasure system, which controls the other three loads to implement reconnaissance and interference tasks. The working process of the electronic countermeasure system is: the three electronic countermeasure loads receive reconnaissance instructions to conduct reconnaissance on the target area. When the target enters the coverage range of the equipment, the electromagnetic radiation released by the reconnaissance load is captured by the reconnaissance load. After the load obtains the target information, it is transmitted back to the command and control system. The command and control system further processes the target information (information fusion, target matching and identification, etc.), generates interference strategies based on the processed information, and the commander issues a control command. The electronic countermeasure load receives the interference command and interferes with the target.

[0124] In the above example, threat level assessment is an important part of information fusion in the command and control system, and plays a vital role in the generation of subsequent strategies. After the electronic countermeasure payload captures the target electromagnetic radiation, it analyzes and processes the electromagnetic information to obtain the frequency, pulse repetition period, bandwidth, pulse width, amplitude, etc. of the target radiation, and further extracts the signal characteristics through sorting and identification, and calculates the target speed, direction, and distance at the same time, and then transmits this target information back to the command and control system, which further processes the initially acquired target information and extracts other target threat elements:

[0125] Target type: The radiation released by different types of radiation source targets has different characteristics in terms of signal frequency, pulse repetition period, pulse width, modulation type, etc. The target type can be identified and matched based on the target radiation information combined with the radiation source target threat library. In this example, the radiation source targets targeted by the electronic countermeasure system are digital and analog radio stations, UAV measurement and control, reconnaissance radars, and airborne / unmanned SAR.

[0126] Target working mode: Different radiation source targets have different characteristics in different working modes. The target working mode and status can be preliminarily judged by further analyzing the radiation source information.

[0127] Target intent: Preliminary inference of target intent based on target type, altitude, heading angle, and speed. In this example, target intent is strongly correlated with target type.

[0128] Target mission capability: After the target type is determined based on the target radiation information, namely frequency, pulse repetition period, etc., combined with the target threat library, different target types have different performance, functions and quantities, and their mission capabilities are also different.

[0129] In this example, based on target type, target intention, target mission capability, target speed, target heading angle, target distance, target altitude and target working mode, a total of 8 threat index elements are involved, represented by x1, x2, ..., x8, plus a threat level index x9. The dimension of the data to be analyzed is 9 dimensions, and the sample size is set to 2000. The specific threat level index is determined by investigating the military threat level assessment criteria and expert experience. Next, based on this data, the present invention is applied to assess the threat level of the target. The entire assessment process is as follows: Figure 2 shown.

[0130] Step 1: Data Preprocessing

[0131] First, the data is discretized. Qualitative indicators such as target type and target intention are directly discretized, and quantitative targets such as target speed and target height are discretized using the K-means algorithm. The discretization results are as follows:

[0132] Target Type

[0133] x1={0,1,2,3}={digital and analog radio, UAV measurement and control, reconnaissance radar, airborne / unmanned SAR}

[0134] Target intention x2 = {0, 1, 2} = {communication, imaging, surveillance, fire control guidance};

[0135] Target task capability x3 = {0, 1, 2} = {weak, medium, strong};

[0136] Target speed x4 = {0, 1, 2} = {slow, medium, fast};

[0137] Target heading angle x5 = {0,1} = {small, large};

[0138] Target distance x6 = {0, 1, 2} = {near, medium, far};

[0139] Target height x7 = {0, 1, 2} = {low, medium, high};

[0140] Target working mode x8 = {0, 1, 2, 3} = {communication, search, tracking, locking};

[0141] Target threat level x9 = {0, 1, 2} = {low, medium, high}.

[0142] Step 2: Build a threat level assessment model

[0143] The above SAOBS method is used to learn the Bayesian network structure from the sample data. The parameter settings of the SAOBS method are shown in Table 1:

[0144] Table 1 SAOBS method parameter settings

[0145]

[0146] Learning results such as Figure 4 Then, the MLE algorithm is used to learn the network parameters based on the network structure and sample data. At this point, the complete Bayesian network is established, which is the threat level assessment model.

[0147] Step 3: Reasoning about the threat level assessment results

[0148] Based on the threat level assessment model established in step 2, the target threat value is calculated by the clique tree propagation algorithm. Now three sets of target situation information are input into the network for inferring the threat level. The two sets of situation information are:

[0149] Target 1: x1=0, x2=1, x3=0, x4=1, x5=0, x6=1, x7=0, x8=0;

[0150] Target 2: x1=1, x2=2, x3=1, x4=0, x5=0, x6=1, x7=1, x8=1;

[0151] Target 3: x1=2, x2=2, x3=2, x4=0, x5=0, x6=2, x7=1, x8=2;

[0152] Inference is performed on the following posterior probabilities:

[0153] P(x9=0|x1=0,x2=1,x3=0,x4=1,x5=0,x6=1,x7=0,x8=0)

[0154] P(x9=1|x1=0,x2=1,x3=0,x4=1,x5=0,x6=1,x7=0,x8=0)

[0155] P(x9=2|x1=0,x2=1,x3=0,x4=1,x5=0,x6=1,x7=0,x8=0)

[0156] P(x9=0|x1=1,x2=2,x3=1,x4=0,x5=0,x6=1,x7=1,x8=1)

[0157] P(x9=1|x1=1,x2=2,x3=1,x4=0,x5=0,x6=1,x7=1,x8=1)

[0158] P(x9=2|x1=1,x2=2,x3=1,x4=0,x5=0,x6=1,x7=1,x8=1)

[0159] P(x9=0|x1=2,x2=2,x3=2,x4=0,x5=0,x6=2,x7=1,x8=2)

[0160] P(x9=1|x1=2,x2=2,x3=2,x4=0,x5=0,x6=2,x7=1,x8=2)

[0161] P(x9=2|x1=2,x2=2,x3=2,x4=0,x5=0,x6=2,x7=1,x8=2)

[0162] The inference values ​​of the above posterior probabilities are shown in Table 2:

[0163] Table 2 Reasoning results

[0164]

[0165] By comparison, the threat levels of the three targets are ranked as follows:

[0166] Goal 3 > Goal 2 > Goal 1.

[0167] refer to Figure 5 On the basis of the above embodiments, the present application further provides an electromagnetic radiation source target threat level assessment device, the assessment device 100 includes an acquisition module 1001, a processing module 1002 and an output module 1003: the acquisition module 1001 is used to acquire target threat factor data; the processing module 1002 is used to process the target threat factor data based on a pre-built threat level assessment model to obtain a threat level reasoning result, wherein a structure learning algorithm based on a pre-built simulated annealing learns a Bayesian network structure from sample data, and learns network parameters from the Bayesian network structure based on a maximum likelihood estimation algorithm, and constructs a threat level assessment model based on the Bayesian network structure and the network parameters, and the structure learning algorithm of simulated annealing is jointly constructed based on an improved simulated annealing algorithm and an improved OBS algorithm; the output module 1003 is used to output the threat level reasoning result.

[0168] It is not difficult to find that this embodiment is a system embodiment corresponding to the above method embodiment, and this embodiment can be implemented in conjunction with the above method embodiment. The relevant technical details and technical effects mentioned in the above embodiments are still valid in this embodiment, and in order to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied in the above embodiments.

[0169] It is worth mentioning that all modules involved in this embodiment are logic modules. In practical applications, a logic unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, in order to highlight the innovative part of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed by this application, but this does not mean that there are no other units in this embodiment.

[0170] Based on the above embodiments, the present application also proposes an electronic device, which includes a processor, a communication interface, a memory and a communication bus. The processor, the communication interface and the memory communicate with each other through the communication bus. The characteristics are: the memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to implement the electromagnetic radiation source target threat level assessment method in any of the above embodiments.

[0171] Among them, the memory and the processor are connected in a bus manner, and the bus may include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and memories together. The bus can also connect various other circuits such as peripheral devices, voltage regulators and power management circuits together, which are all well known in the art, so this article will not describe them further. The bus interface provides an interface between the bus and the transceiver. The transceiver can be one element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. The data processed by the processor is transmitted on the wireless medium through the antenna, and further, the antenna also receives the data and transmits the data to the processor.

[0172] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory can be used to store data used by the processor when performing operations.

[0173] Based on the above embodiments, the present application also proposes a computer-readable storage medium, characterized in that it includes instructions, which, when executed on a computer, enable the computer to execute the electromagnetic radiation source target threat level assessment method in any of the above embodiments.

[0174] Those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including a number of instructions to enable a device (which can be a single-chip microcomputer, chip, etc.) or a processor (processor) to execute all or part of the steps of each embodiment method of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, ROM (Read-Only Memory), RAM (Random Access Memory), disk or optical disk and other media that can store program codes.

[0175] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.

Claims

1. A method for assessing the threat level of an electromagnetic radiation source target, characterized in that: include: Obtain target threat factor data; The target threat factor data is processed based on a pre-constructed threat level assessment model to obtain a threat level inference result, wherein a Bayesian network structure is learned from sample data based on a pre-constructed simulated annealing structure learning algorithm, and network parameters are learned from the Bayesian network structure based on a maximum likelihood estimation algorithm, and the network parameters are converted into a connection tree according to a clique propagation algorithm to obtain the threat level assessment model, and the simulated annealing structure learning algorithm is jointly constructed based on an improved simulated annealing algorithm and an improved OBS algorithm; The threat level inference result is outputted.

2. The electromagnetic radiation source target threat level assessment method according to claim 1, characterized in that: Before processing the target threat factor data based on the pre-built threat level assessment model, the method includes: An initial node sequence is obtained based on the discretized target threat factor data.

3. The electromagnetic radiation source target threat level assessment method according to claim 2, characterized in that: The pre-built simulated annealing-based structure learning algorithm learns the Bayesian network structure from sample data, including: Determining initial conditions of the threat level assessment model, the initial conditions including an initial temperature value, a termination temperature value, an annealing rate, a minimum tolerable number of non-boosting searches, an initial value of the number of invalid searches, the initial node sequence, and a current optimal node sequence; Iteratively searching the initial node sequence using the improved OBS algorithm to obtain a local optimal node sequence, and obtaining a new node sequence as the initial node sequence for the next round of improved OBS search; Using a scoring function to process the local optimal node sequence and the current optimal node sequence respectively, and obtaining a first score and a second score respectively; If the first score is greater than the second score, the current optimal node sequence is updated using the local optimal node sequence, and the number of invalid searches is reset to zero; The optimal node sequence is processed by the K2 algorithm to obtain the Bayesian network structure.

4. The electromagnetic radiation source target threat level assessment method according to claim 3, characterized in that: After the local optimal node sequence and the current optimal node sequence are processed respectively by using a scoring function to obtain a first score and a second score respectively, the method further includes: If the first score is less than the second score, the current optimal node sequence is not updated, and the invalid search count is increased by one to obtain a new invalid search count; Based on the new node sequence as the start of the next search, and based on the first expression to update the current temperature value, continue to iteratively search the local optimal node sequence and the new node sequence based on the improved OBS algorithm, and perform subsequent operations until the termination iteration is reached; The iteration termination parameter value is calculated, the iteration is terminated according to the new invalid search times, the iteration termination parameter value, the current temperature value and the termination temperature value, and the current optimal node sequence is output.

5. The electromagnetic radiation source target threat level assessment method according to claim 3, characterized in that: The method of iteratively searching the initial node sequence using the improved OBS algorithm to obtain a local optimal node sequence, and obtaining a new node sequence as an initial node sequence for the next round of improved OBS method search, includes: During the search process, if the score of the node sequence obtained by an operation in a round of search is greater than the score of the current node sequence, then this node sequence is accepted and the search continues; If the score of the node sequence obtained by an operation in a round of search is less than the score of the current node sequence, then the node sequence is selected to be accepted or abandoned based on the Metropolis criterion; If the node sequence is chosen to be accepted, the scores of the current node sequence and the current optimal node sequence are compared, and if the score of the current node sequence is greater than the score of the current optimal node sequence, the current node sequence is stored in a pre-built memory module; If you choose to abandon the node sequence, continue searching based on the current node sequence.

6. The electromagnetic radiation source target threat level assessment method according to claim 4, characterized in that: The calculating of the iteration termination parameter value, terminating the iteration according to the new invalid search times, the iteration termination parameter value, the current temperature value and the termination temperature value, and outputting the current optimal node sequence includes: The iteration termination parameter value is calculated based on the second expression, wherein the second expression is: Among them, α min represents the minimum number of tolerated searches without improvement, T i Indicates that the current temperature value is obtained after each iteration, T f Indicates the termination temperature value; If the current value of the invalid search times is greater than the iteration termination parameter value, or the current temperature value is greater than the termination temperature value, the iteration is stopped and the current optimal node sequence is output.

7. The electromagnetic radiation source target threat level assessment method according to claim 4, characterized in that: The updating of the current temperature value based on the first expression, continuing to iteratively search for the optimal node sequence and the new node sequence based on the improved OBS algorithm, and performing subsequent operations until the iteration is terminated, includes: If the current value of the invalid search times is less than the iteration termination parameter value, and the current temperature value is less than the termination temperature value, the current temperature is updated based on the first expression and the new node sequence is output, wherein the first expression is: T c =λT c Where λ represents the annealing rate, T c Represents the current temperature output after each iteration; If the current value of the invalid search times is greater than the iteration termination parameter value, or the current temperature value is greater than the termination temperature value, the iteration is stopped and the optimal node sequence is output; If the current value of the number of invalid searches is less than the iteration termination parameter value, and the current temperature value is less than the termination temperature value, the current temperature value of the next iteration is obtained based on the product of the annealing rate and the current temperature value, and the new node sequence is used as the initial node sequence of the next iteration, and the optimal node sequence is continued to be iteratively searched based on the improved OBS algorithm and the new node sequence is output until the preset judgment condition is met, and the preset judgment condition is that the current value of the number of invalid searches is greater than the iteration termination parameter value, or the current temperature value is greater than the termination temperature value.

8. An electromagnetic radiation source target threat level assessment device, characterized in that: include: An acquisition module, used to acquire target threat factor data; A processing module, used for processing the target threat factor data based on a pre-constructed threat level assessment model to obtain a threat level inference result, wherein a Bayesian network structure is learned from sample data based on a pre-constructed simulated annealing structure learning algorithm, and network parameters are learned from the Bayesian network structure based on a maximum likelihood estimation algorithm, and the threat level assessment model is constructed based on the Bayesian network structure and the network parameters, and the simulated annealing structure learning algorithm is jointly constructed based on an improved simulated annealing algorithm and an improved OBS algorithm; An output module is used to output the threat level reasoning result.

9. A computer-readable storage medium, characterized in that: The method comprises instructions, which, when executed on a computer, enable the computer to execute the electromagnetic radiation source target threat level assessment method as claimed in any one of claims 1 to 7.

10. An electronic device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus, characterized in that: a memory for storing a computer program, A processor, wherein the processor is used to execute a computer program stored in a memory to implement the electromagnetic radiation source target threat level assessment method according to any one of claims 1 to 7.

Images