A method and device for adaptive interference target allocation based on machine learning

By employing a machine learning-based adaptive jamming target allocation method, which utilizes receiver-acquired target feature parameters and KNN algorithm for signal clustering, the problem of jamming target allocation error and time loss for radar electronic warfare equipment in complex environments is solved, achieving fast and accurate jamming target allocation and resource optimization.

CN115062545BActive Publication Date: 2024-11-19CHINA SHIPBUILDING IND CORP NO 723 RESEARCH INSTITUTE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210701216.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-06-21
Publication Date
2024-11-19
Estimated Expiration
2042-06-21

AI Technical Summary

Technical Problem

Existing radar electronic warfare equipment struggles to quickly and accurately allocate jamming targets in complex electromagnetic environments, resulting in errors and time losses, which fails to meet the demands of modern warfare.

Method used

An adaptive interference target allocation method based on machine learning is adopted. The target feature parameters are obtained by the receiver, a target allocation model with multiple constraints is constructed, and the optimal interference mode is adaptively allocated for multiple target signals by combining the KNN algorithm and signal clustering, thereby reducing human operation error and time loss.

Benefits of technology

It enables rapid and accurate target allocation for jamming in complex electromagnetic environments, improves jamming efficiency and effectiveness, provides faster and more accurate jamming solutions, and optimizes battlefield resource utilization.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115062545B_ABST
    Figure CN115062545B_ABST
Patent Text Reader

Abstract

The present invention discloses an adaptive interference target allocation method based on machine learning, which is as follows: first, according to the characteristic parameter information of the target signal, the threat degree and threat weight are calculated, and a target allocation model with multiple constraints is constructed; then, a clustering algorithm based on the signal type is used to obtain the similarity matrix of the target signal, and then the optimal solution is obtained by judging and verifying the signal category to which it belongs according to the KNN classification algorithm, and the optimal interference mode is allocated to different targets. The present invention can provide decision makers with a faster and more accurate solution on the premise of ensuring that the interference mode meets all signal types, eliminates the errors and time loss existing in manual operation, realizes the adaptive allocation of interference target signals, and improves the interference efficiency and interference effect.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to electronic countermeasure technology, and in particular to a method and device for adaptive jamming target allocation based on machine learning. Background Art

[0002] In radar electronic warfare, jamming and anti-jamming technologies are constantly evolving amidst the interplay between friendly and enemy forces. As the electromagnetic environment becomes increasingly complex, accurate and rapid differentiation between friendly and enemy radars, as well as the ability to assess the threat level of enemy radars and implement rapid and effective jamming control, are essential capabilities for current electronic warfare equipment. Traditional, manually operated equipment cannot meet the demands of today's complex warfare environments. Therefore, improving electronic warfare equipment's adaptive target allocation and jamming control capabilities not only impacts decision-makers' target judgment but also the outcome of the entire war.

[0003] In a complex combat environment characterized by a constantly changing electromagnetic environment and multiple target signals, the types of threatening signals and the countermeasures employed are diverse. Even for the same target, different countermeasures may be employed or certain countermeasure parameters may be modified to achieve optimal countermeasure effectiveness under different combat factors. Therefore, when an enemy target attacks, timely and accurate decisions on the specific jamming time and method for multiple target signals are crucial to optimizing combat effectiveness and maximizing the use of friendly resources for rapid and effective strikes against targets.

[0004] Whether the target allocation plan is timely and effective is a key factor in determining victory or defeat on the battlefield. The traditional approach relies on operators analyzing targets and then implementing a jamming method based on experience or on-site evaluation. The effectiveness of the jamming is not evaluated afterwards, making it impossible to determine whether the jamming achieved the intended effect. This approach fails to form a closed loop and is no longer able to meet the rapidly changing battlefield conditions of the modern environment.

[0005] Most of the current jammers are operated manually. After receiving the target signal, the operator manually selects the corresponding jamming style based on the signal type and threat level. Multiple signals require multiple operations, which will lead to errors and time loss. In addition, the operator has difficulty in analyzing and judging certain unknown signals. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and device for adaptive interference target allocation based on machine learning, which provides decision makers with a faster and more accurate solution while ensuring that the interference mode meets all signal types, eliminates the errors and time loss existing in manual operation, realizes adaptive allocation of interference target signals, and improves interference efficiency and interference effect.

[0007] The technical solution to achieve the purpose of the present invention is: a method for adaptive interference target allocation based on machine learning, comprising the following steps:

[0008] Step 1: Obtain target information through the receiver and extract the characteristic parameters of each target, including target distance D, target direction W, target speed V, target type T and target state C;

[0009] Step 2: Analyze the extracted target features to obtain the threat level of the target, and normalize the threat level to obtain the threat weight of different targets;

[0010] Step 3: Construct a target allocation model with multiple constraints based on the extracted target feature information and the target threat weight;

[0011] Step 4: When a target signal is received, the target signal is compared and analyzed with the known radar database. If it matches the database, the interference pattern of the target is assigned directly using the target allocation model in step 3 based on the correspondence between the radar database and the interference source database. If it does not match the database, it is determined to be a new target signal and the process goes to step 5.

[0012] Step 5: Cluster the new target signal with the signals in the radar database to obtain a similarity matrix between the new target signal and other signals in the radar database;

[0013] Step 6: Based on the similarity matrix after clustering, use the KNN algorithm to determine the signal category to which the new target signal belongs, use the interference pattern corresponding to the category to perform interference, and record the interference results;

[0014] Step 7: If the interference is effective, the interference pattern in step 6 is the optimal solution, and the radar database and interference source database are updated; if the interference is ineffective, proceed to step 8;

[0015] Step 8: Based on the similarity matrix in step 5, select the next interference pattern corresponding to the signal most similar to the signal to interfere, record the interference results, and proceed to step 7 to determine whether the interference is effective until the optimal solution is obtained. Based on the optimal solution, use the target allocation model in step 3 to allocate the interference pattern to the target.

[0016] A machine learning-based adaptive interference target allocation device, comprising:

[0017] The characteristic parameter extraction module is used to obtain target information through the receiver and extract the characteristic parameters of each target, including target distance D, target direction W, target speed V, target type T and target state C;

[0018] The target threat level acquisition module is used to analyze the extracted target features to obtain the target threat level, and normalize the threat level to obtain the threat weight of different targets;

[0019] A target allocation model construction module with multiple constraints is used to construct a target allocation model with multiple constraints based on the extracted target feature information and the target's threat weight;

[0020] The target signal and radar database matching module is used to compare and analyze the target signal with the known radar database. If it matches the database, the target interference pattern is directly assigned using the target allocation model based on the correspondence between the radar database and the interference source database. If it does not match the database, it is determined to be a new target signal and transferred to the target signal clustering module.

[0021] The target signal clustering module is used to cluster the new target signal with the signals in the radar database to obtain the similarity matrix between the new target signal and other signals in the radar database;

[0022] The target signal category determination module is used to determine the signal category to which the new target signal belongs based on the clustered similarity matrix using the KNN algorithm, perform interference using the interference pattern corresponding to the category, and record the interference results;

[0023] The interference effect determination module is used to determine whether the interference is effective. If the interference is effective, the interference pattern in the target signal category determination module is the optimal solution, and the radar database and the interference source database are updated. If the interference is ineffective, the process proceeds to step 8.

[0024] The optimal solution determination module is used to select the next interference style corresponding to the signal most similar to the signal according to the similarity matrix, record the interference results, and transfer it to the interference effect judgment module to determine whether the interference is effective until the optimal solution is obtained. The target allocation model is used to allocate the interference style of the target according to the optimal solution.

[0025] A mobile terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, an adaptive interference target allocation method based on machine learning is implemented.

[0026] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the adaptive interference target allocation method based on machine learning.

[0027] Compared with the prior art, the present invention has the following significant advantages: (1) the interference target allocation problem is converted into a multi-objective optimization problem. On the premise of ensuring that the interference mode satisfies all signal types, a clustering algorithm based on signal type and a KNN classification algorithm are used to obtain the optimal solution, and the most effective interference mode is adaptively allocated to the detected multi-target signals, thereby improving the interference efficiency; (2) the time of manual operation is greatly reduced. The adaptive interference processing has the ability to detect and receive, and can make the optimal interference decision by analyzing the interference effect of its own transmission. In target allocation, it can accurately and quickly make the correct interference style allocation for each target, providing operators with a faster and more accurate interference solution for the analysis and judgment of certain unknown signals, thereby improving the interference effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 The figure is a flow chart of a method for adaptive interference target allocation based on machine learning according to the present invention.

[0029] Figure 2 Schematic diagram of the process of adaptive interference processing of the present invention.

[0030] Figure 3 Schematic diagram of selecting the optimal k value for clustering in an embodiment of the present invention. DETAILED DESCRIPTION

[0031] The present invention provides an adaptive interference target allocation method based on machine learning, characterized by comprising the following steps:

[0032] Step 1: Obtain target information through the receiver and extract the characteristic parameters of each target, including target distance D, target direction W, target speed V, target type T and target state C;

[0033] Step 2: Analyze the extracted target features to obtain the threat level of the target, and normalize the threat level to obtain the threat weight of different targets;

[0034] Step 3: Construct a target allocation model with multiple constraints based on the extracted target feature information and the target threat weight;

[0035] Step 4: When a target signal is received, the target signal is compared and analyzed with the known radar database. If it matches the database, the interference pattern of the target is assigned directly using the target allocation model in step 3 based on the correspondence between the radar database and the interference source database. If it does not match the database, it is determined to be a new target signal and the process goes to step 5.

[0036] Step 5: Cluster the new target signal with the signals in the radar database to obtain a similarity matrix between the new target signal and other signals in the radar database;

[0037] Step 6: Based on the similarity matrix after clustering, use the KNN algorithm to determine the signal category to which the new target signal belongs, use the interference pattern corresponding to the category to perform interference, and record the interference results;

[0038] Step 7: If the interference is effective, the interference pattern in step 6 is the optimal solution, and the radar database and interference source database are updated; if the interference is ineffective, proceed to step 8;

[0039] Step 8: Based on the similarity matrix in step 5, select the next interference pattern corresponding to the signal most similar to the signal to interfere, record the interference results, and proceed to step 7 to determine whether the interference is effective until the optimal solution is obtained. Based on the optimal solution, use the target allocation model in step 3 to allocate the interference pattern to the target.

[0040] As a specific implementation, in step 1, target information is acquired through a receiver, and characteristic parameters of each target are extracted, including target distance D, target direction W, target speed V, target type T, and target state C, as follows:

[0041] The target characteristic parameters are obtained to construct the target information set N, which is represented by a five-tuple N = {D, W, V, T, C}, where D is the target distance, W is the target direction, V is the target speed, T is the target type, and C is the target state.

[0042] As a specific implementation, the extracted target features are analyzed in step 2 to obtain the threat level of the target, and the threat level is normalized to obtain the threat weights of different targets, as follows:

[0043] Step 2.1: Use Bayesian or neural network methods to calculate the threat level of each target;

[0044] Step 2.2: Normalize the threat level to obtain the threat weight of the i-th target.

[0045]

[0046] Where N is the number of targets acquired, V i is the threat level of the i-th target.

[0047] As a specific implementation method, in step 3, a target allocation model with multiple constraints is constructed based on the extracted target feature information and the target threat weight, as follows:

[0048] When the jammer is working normally, the number of targets being jammed is less than or equal to the total number of targets that the jammer can jam. The number of jamming patterns of the jammer is M, and the number of target signals is N. The number of jamming patterns that can jam targets at the same time is R.j , where j = 1, 2...M; target N i Threat coefficient V i represents the threat value of the target to the interferer, where i = 1, 2..., N; p ij represents the probability that the jth interference pattern effectively interferes with the i-th target; x ij Indicates whether the jth interference pattern interferes with the i-th target, x ij =1 indicates interference, x ij =0 means no interference;

[0049] P represents the probability that the target is not interfered with, and the target allocation model is:

[0050]

[0051] The target allocation model satisfies three constraints:

[0052] ①The maximum number of R that can be allocated to the i-th target is j interference pattern, i.e.

[0053] ②The assigned interference patterns cannot exceed the total number of patterns M, that is,

[0054] ③A target can only be assigned one interference pattern at the same time, that is, x ij =0 or x ij =1.

[0055] As a specific implementation, step 5 clusters the new target signal with the signals in the radar database to obtain a similarity matrix between the new target signal and other signals in the radar database, as follows:

[0056] Step 5.1, set the radar database to {N1, N2, ..., N m}, where m represents the number of signals;

[0057] Step 5.2: Use the error sum squared SSE method to select k points as the initial clustering point centers:

[0058]

[0059] Among them, C i is the i-th cluster, p is C i Sample points, m i It is C i The mean centroid of , SSE is the clustering error of all samples, which represents the quality of clustering effect;

[0060] Step 5.3: According to the principle of minimum distance from the center point, all target points are divided into the class where each center point is located, as follows:

[0061] Calculate two signals N1=(x1,x2,…,x n ) and N2=(y1,y2,…,y n ) between them, where n is the number of features;

[0062] Step 5.4: There are multiple targets in each class. Calculate the mean of all sample points in each of the k classes and use it as the center point for the next iteration.

[0063] Step 5.5: Repeat steps 5.3 to 5.4 according to the center point updated in step 5.4 until convergence. At this time, the target signal will be classified into one category, and the distance matrix between the target and other targets, that is, the similarity matrix, can be obtained.

[0064] As a specific implementation, in step 6, based on the similarity matrix after clustering, the KNN algorithm is used to determine the signal category to which the new target signal belongs, and the interference pattern corresponding to the category is used for interference, and the interference results are recorded as follows:

[0065] Step 6.1: Use the KNN algorithm to sort the other signals from near to far according to their similarity based on the similarity matrix, with the clustered new target signal as the center.

[0066] Step 6.2: Select the K points closest to each other, determine the category of the new target signal based on the types of these K points, and assign the interference pattern corresponding to the category.

[0067] As a specific implementation, in step 5.3, two signals N1=(x1, x2, ..., x n ) and N2=(y1,y2,…,y n ) between them, using Euclidean distance, Manhattan distance or Chebyshev distance, the formula is:

[0068] The Euclidean distance calculation formula is:

[0069]

[0070] The formula for calculating Manhattan distance is:

[0071]

[0072] The Chebyshev distance calculation formula is:

[0073]

[0074] A machine learning-based adaptive interference target allocation device, comprising:

[0075] The characteristic parameter extraction module is used to obtain target information through the receiver and extract the characteristic parameters of each target, including target distance D, target direction W, target speed V, target type T and target state C;

[0076] The target threat level acquisition module is used to analyze the extracted target features to obtain the target threat level, and normalize the threat level to obtain the threat weight of different targets;

[0077] A target allocation model construction module with multiple constraints is used to construct a target allocation model with multiple constraints based on the extracted target feature information and the target's threat weight;

[0078] The target signal and radar database matching module is used to compare and analyze the target signal with the known radar database. If it matches the database, the target interference pattern is directly assigned using the target allocation model based on the correspondence between the radar database and the interference source database. If it does not match the database, it is determined to be a new target signal and transferred to the target signal clustering module.

[0079] The target signal clustering module is used to cluster the new target signal with the signals in the radar database to obtain the similarity matrix between the new target signal and other signals in the radar database;

[0080] The target signal category determination module is used to determine the signal category to which the new target signal belongs based on the clustered similarity matrix using the KNN algorithm, perform interference using the interference pattern corresponding to the category, and record the interference results;

[0081] The interference effect determination module is used to determine whether the interference is effective. If the interference is effective, the interference pattern in the target signal category determination module is the optimal solution, and the radar database and the interference source database are updated. If the interference is ineffective, the process proceeds to step 8.

[0082] The optimal solution determination module is used to select the next interference style corresponding to the signal most similar to the signal according to the similarity matrix, record the interference results, and transfer it to the interference effect judgment module to determine whether the interference is effective until the optimal solution is obtained. The target allocation model is used to allocate the interference style of the target according to the optimal solution.

[0083] A mobile terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, an adaptive interference target allocation method based on machine learning is implemented.

[0084] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the adaptive interference target allocation method based on machine learning.

[0085] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0086] Example

[0087] Combine Figure 1 、 Figure 2 The present invention provides an adaptive interference target allocation method based on machine learning, the steps are as follows:

[0088] Step 1: Obtain target information through the receiver and extract the characteristic parameters of each target, including target distance D, target direction W, target speed V, target type T and target state C, as follows:

[0089] The target information is obtained through the receiver, and the characteristic parameters of each signal are extracted, including the target distance D, target direction W, target speed V, target type T and target state C. A target information set N = {D, W, V, T, C} is constructed, where D is the target distance, W is the target direction, V is the target speed, T is the target type, and C is the target state.

[0090] Step 2: Analyze the extracted target features to obtain the threat level of the target, and normalize the threat level to obtain the threat weight of different targets, as follows:

[0091] According to the information set N obtained in step 1, a common target threat level estimation method is used to calculate the threat level of each target, such as using Bayesian, neural network and other methods to estimate the target threat level, and then the threat level is normalized to obtain the corresponding weight.

[0092] The target threat level is calculated as follows:

[0093]

[0094] Among them, N is the number of targets obtained, and the threat level of each target is V i .

[0095] Step 3: Based on the extracted target feature information and the target threat weight, a target allocation model with multiple constraints is constructed as follows:

[0096] When the jammer is working normally, the number of targets being jammed is less than or equal to the total number of targets that the jammer can jam. The number of jamming patterns of the jammer is M, and the number of target signals is N. The number of jamming patterns that can jam targets at the same time is R. j, where j = 1, 2...M; target N i Threat coefficient V i represents the threat value of the target to the interferer, where i = 1, 2..., N; p ij represents the probability that the jth interference pattern effectively interferes with the i-th target; x ij Indicates whether the jth interference pattern interferes with the i-th target, x ij =1 indicates interference, x ij =0 means no interference;

[0097] P represents the probability that the target is not interfered with, and the target allocation model is:

[0098]

[0099] The target allocation model satisfies three constraints:

[0100] ①The maximum number of R that can be allocated to the i-th target is j interference pattern, i.e.

[0101] ②The assigned interference patterns cannot exceed the total number of patterns M, that is,

[0102] ③A target can only be assigned one interference pattern at the same time, that is, x ij =0 or x ij =1.

[0103] Step 4: Compare and analyze the target information with the known radar database. If it matches the database, the interference pattern of the target is directly assigned based on the correspondence between the radar database and the interference source database. If it does not match the database, it is determined to be a new target signal and proceeds to step 5.

[0104] Step 5: Cluster the new target signal with the signals in the radar database to obtain the similarity matrix between the signal and other signals, as follows:

[0105] Step 5.1, set the radar database to {N1, N2, ..., N m}, where m represents the number of signals, N i Indicates a certain signal;

[0106] Step 5.2: Use the error sum squared SSE method to select k points as the initial clustering point centers:

[0107]

[0108] Among them, C i is the i-th cluster, p is C i Sample points, mi It is C i The mean centroid of , SSE is the clustering error of all samples, which represents the quality of clustering effect;

[0109] Step 5.3: According to the principle of minimum distance from the center point, all target points are divided into the class where each center point is located, as follows:

[0110] Two signals N1=(x1,x2,…,x n ) and N2=(y1,y2,…,y n ), where n is the number of features, and the Euclidean distance between the two signals is:

[0111]

[0112] Step 5.4: There are several targets in each class. Calculate the mean of all sample points in k classes and use it as the center point for the next iteration.

[0113] Step 5.5: Repeat steps 5.3 to 5.4 according to the center point updated in step 5.4 until convergence. At this time, the target signal will be classified into a certain category, and the distance matrix between the target and other targets, that is, the similarity matrix, can be obtained.

[0114] like Figure 3 As shown in the figure, as the number of clusters k increases, SSE will gradually decrease. When k is less than the actual number of clusters, the increase in k will greatly increase the degree of aggregation of each cluster, so the decrease in SSE will be large. When k reaches the actual number of clusters, the decrease in SSE will drop sharply, and then tend to be flat as the k value continues to increase. This inflection point, that is, k = 4, corresponds to the actual number of clusters.

[0115] Step 6: Use the KNN algorithm to determine the signal category of the new target signal after clustering, use the interference pattern of the signal in that category to interfere, and record the interference results, as follows:

[0116] Step 6.1: Use the KNN algorithm to sort the other signals from near to far according to their similarity based on the similarity matrix, with the clustered new target signal as the center.

[0117] Step 6.2: Select the K points closest to each other, determine the category of the new target signal based on the types of these K points, and assign it to the interference pattern of that category;

[0118] The selection of K value is different from clustering. The larger the K value, the greater the error. Therefore, a smaller K value is generally selected here and gradually increased until the optimal solution is obtained. The similarity calculation can also use other similarity distance calculation formulas, such as Manhattan distance, Chebyshev distance, Hamming distance, angle cosine, and other distance formulas.

[0119] Manhattan distance calculation formula:

[0120]

[0121] Chebyshev distance calculation formula:

[0122]

[0123] Step 7: If the interference is effective, the interference pattern is the optimal solution, and the radar database and the interference source database are updated; if the interference is ineffective, proceed to step 8;

[0124] Step 8: Based on the similarity matrix in step 5, select the next interference pattern corresponding to the signal most similar to the signal to interfere, record the interference result, and go to step 7 to determine whether the interference is effective until the optimal solution is obtained. The target interference pattern is allocated according to the optimal solution.

[0125] This embodiment provides a method for allocating interference targets based on machine learning, which transforms the interference target allocation problem into a multi-objective optimization problem. While ensuring that the interference method satisfies all signal types, a clustering algorithm based on signal type and a KNN classification algorithm are used to obtain the optimal solution. The method adaptively allocates the most effective interference method to the detected multi-target signals, solving the error and time loss problems that currently exist in manual operation of electronic warfare equipment. This method can provide decision makers with a faster and more accurate solution, enabling them to quickly and accurately determine the most effective interference method in actual combat.

Claims

1. An adaptive interference target allocation method based on machine learning, It is characterized in that The following steps are involved: Step 1: Obtain target information through the receiver and extract characteristic parameters of each target, including target distance D, target direction W, target speed V, target type T and target state C; Step 2: Analyze the extracted target features to obtain the threat level of the target, and normalize the threat level to obtain the threat weights of different targets; Step 3: construct a target allocation model with multiple constraints based on the extracted target feature information and the target threat weight; Step 4: When a target signal is received, the target signal is compared and analyzed with the known radar database. If it matches the database, the interference pattern of the target is directly allocated using the target allocation model in step 3 through the corresponding relationship between the radar database and the interference source database; if it does not match the database, it is determined to be a new target signal and the process goes to step 5; Step 5: Cluster the new target signal with the signals in the radar database to obtain a similarity matrix between the new target signal and other signals in the radar database; Step 6: Based on the clustered similarity matrix, use the KNN algorithm to determine the signal category to which the new target signal belongs, use the interference style corresponding to the category to perform interference, and record the interference results; Step 7: If the interference is effective, the interference pattern in step 6 is the optimal solution, and the radar database and the interference source database are updated; If the interference is ineffective, go to step 8; Step 8. According to the similarity matrix of step 5, select the next interference style corresponding to the signal most similar to the signal for interference, record the interference result, and go to step 7 to determine whether the interference is effective until the optimal solution is obtained. According to the optimal solution, use the target allocation model in step 3 to allocate the interference style of the target.

2. The method for adaptive interference target allocation based on machine learning according to claim 1, It is characterized in that The target information is obtained through the receiver as described in step 1, and the characteristic parameters of each target are extracted, including the target distance D, the target direction W, the target speed V, the target type T and the target state C, as follows: The target feature parameters are obtained to construct the target information set N, which is represented by a five-tuple N = {D, W, V, T, C}, where D is the target distance, W is the target direction, V is the target speed, T is the target type, and C is the target state.

3. The method for adaptive interference target allocation based on machine learning according to claim 1, It is characterized in that The extracted target features described in step 2 are analyzed to obtain the threat level of the target, and the threat level is normalized to obtain the threat weights of different targets, as follows: Step 2.1, using Bayesian or neural network methods to calculate the threat level of each target; Step 2.2: Normalize the threat level to obtain the threat weight of the i-th target Where N is the number of targets acquired, V i is the threat level of the i-th target.

4. The method for adaptive interference target allocation based on machine learning according to claim 1, It is characterized in that According to the extracted target feature information and the target threat weight described in step 3, a target allocation model with multiple constraints is constructed, as follows: When the jammer is working normally, the number of targets being jammed is less than or equal to the total number of targets that the jammer can jam. The number of jamming patterns of the jammer is M, and the number of target signals is N. The number of jamming patterns that can jam targets at the same time is R. j , where j = 1, 2, ..., M; target N i Threat coefficient V i represents the threat value of the target to the interferer, where i = 1, 2, ..., N; p ij represents the probability that the jth interference pattern effectively interferes with the i-th target; x ij Indicates whether the jth interference pattern interferes with the i-th target, x ij =1 means interference, x ij =0 means no interference; P represents the probability that the target is not disturbed, so the target allocation model is: The target allocation model satisfies three constraints: ① The i-th target can be allocated at most R j interference pattern, namely ②The number of interference patterns assigned cannot exceed the total number of patterns M, that is, ③A target can only be assigned one interference pattern at the same time, that is, x ij =0 or x ij =1.

5. The method for adaptive interference target allocation based on machine learning according to claim 1, It is characterized in that In step 5, the new target signal is clustered with the signals in the radar database to obtain a similarity matrix between the new target signal and other signals in the radar database, as follows: Step 5.1, set the radar database to {N 1 , N 2 ,…,N m }, where m represents the number of signals; Step 5.2: Use the error square sum SSE method to select k points as the initial clustering point centers: Among them, C i is the i-th cluster, p is C i The sample points, m i It is C i The mean centroid of , SSE is the clustering error of all samples, which represents the quality of clustering effect; Step 5.3: According to the principle of minimum distance from the center point, all target points are divided into the classes where the center points are located, as follows: Calculate two signals N 1 =(x 1 ,x 2 ,…,x n ) and N 2 =(y 1 ,y 2 ,…,y n ) between 1 ,N 2 ), where n is the number of features; Step 5.4: There are multiple targets in each class. Calculate the mean of all sample points in k classes respectively as the center point of the next iteration; Step 5.5: Repeat steps 5.3 to 5.4 according to the center point updated in step 5.4 until convergence. At this time, the target signal will be classified into one category, and the distance matrix between the target and other targets, that is, the similarity matrix, can be obtained.

6. The method for adaptive interference target allocation based on machine learning according to claim 1, It is characterized in that According to the similarity matrix after clustering, the KNN algorithm is used in step 6 to determine the signal category to which the new target signal belongs, and the interference pattern corresponding to the category is used for interference, and the interference results are recorded, as follows: Step 6.1, using the KNN algorithm, with the clustered new target signal as the center, and arranging other signals from near to far according to their similarity according to the similarity matrix; Step 6.2: Select the K points closest to each other, determine the category of the new target signal based on the types of these K points, and assign the interference pattern corresponding to the category.

7. The method for adaptive interference target allocation based on machine learning according to claim 5, It is characterized in that In step 5.3, two signals N are calculated 1 =(x 1 ,x 2 ,…,x n ) and N 2 =(y 1 ,y 2 ,…,y n ) between 1 ,N 2 ), using Euclidean distance, Manhattan distance or Chebyshev distance, the formula is: The Euclidean distance calculation formula is: The Manhattan distance calculation formula is: The Chebyshev distance calculation formula is:

8. An adaptive interference target allocation device based on machine learning, It is characterized in that The device comprises: The characteristic parameter extraction module is used to obtain target information through the receiver and extract characteristic parameters of each target, including target distance D, target direction W, target speed V, target type T and target state C; The target threat level acquisition module is used to analyze the extracted target features to obtain the target threat level, and normalize the threat level to obtain the threat weights of different targets; A target allocation model construction module with multiple constraints is used to construct a target allocation model with multiple constraints based on the extracted target feature information and the target threat weight; The target signal and radar database matching module is used to compare and analyze the target signal with the known radar database. If it matches the database, the interference pattern of the target is directly allocated using the target allocation model through the corresponding relationship between the radar database and the interference source database; if it does not match the database, it is determined to be a new target signal and transferred to the target signal clustering module; A target signal clustering module is used to cluster the new target signal with the signals in the radar database to obtain a similarity matrix between the new target signal and other signals in the radar database; The target signal category determination module is used to determine the signal category to which the new target signal belongs based on the clustered similarity matrix using the KNN algorithm, perform interference using the interference style corresponding to the category, and record the interference results; The interference effect determination module is used to determine whether the interference is effective. If the interference is effective, the interference pattern in the target signal category determination module is the optimal solution, and the radar database and the interference source database are updated; if the interference is ineffective, proceed to step 8; The optimal solution determination module is used to select the interference style corresponding to the next signal that is most similar to the signal according to the similarity matrix, record the interference results, and transfer to the interference effect determination module to determine whether the interference is effective until the optimal solution is obtained. According to the optimal solution, the target allocation model is used to allocate the interference style of the target.

9. A mobile terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the program, an adaptive interference target allocation method based on machine learning is implemented as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the program is executed by a processor, the steps in the adaptive interference target allocation method based on machine learning as described in any one of claims 1 to 7 are implemented.