A Radar True and False Target Recognition Method Based on Parameter Screening under Strong Interference

Abstract

The present invention discloses a method for identifying true and false radar targets based on parameter screening under strong interference conditions, belonging to the field of aircraft intelligent identification, and is used to solve the problem of low accuracy in judging the authenticity of fighter radar targets in the prior art. A method for identifying true and false targets based on parameter screening is proposed, including steps such as target data grouping, inter-frame target association, and true and false target identification. The present invention makes full use of the data attributes of radar true and false targets, explores the relationship between target motion characteristics and authenticity, and in 1000 simulations, the true and false target recognition rate is at least greater than 95% in 700 simulations, improving the accuracy of identifying the authenticity of fighter radar targets.

Description

Technical Field

[0001] The present invention belongs to the fields of target recognition and aviation technology, and particularly relates to a method for identifying true and false radar targets based on parameter screening under strong interference conditions. Background Art

[0002] During the air combat process, as an important detector, the radar plays an important role in detecting, tracking, and guiding the strike of targets. However, deceptive interference acts on the radar's detection and tracking systems with false target information. These radar false targets have characteristics similar to the echo signals of real targets, making it impossible for the receiver to correctly detect the target information or complete the correct tracking of the real target, thus confusing and deceiving the radar, preventing the radar from normally tracking effective targets, and further unable to perform air tasks such as reconnaissance and strike. Therefore, it is very important to identify the true and false of the radar-detected targets.

[0003] The identification of true and false radar targets is actually a classification process. According to the existing target motion trajectory data, its attribute characteristics are analyzed, and an algorithm is selected for discrimination based on its category. Currently, the main methods used for discriminating true and false targets in air combat radar are the true and false identification methods based on statistical characteristic differences, machine learning-related classification algorithms (such as SVM, KNN, etc.), and artificial neural network models (such as RNN, LSTM, etc.). Since the motion dimensions of true and false targets have a very strong correlation, each judgment method has certain applicability and limitations. Among them, the true and false identification method based on statistical characteristic differences constructs a statistical discriminant using the differences in the elevation angle and azimuth angle of the target, and utilizes the characteristic that the variance of false targets is greater than that of real targets to achieve the discrimination of true and false targets. This method is greatly affected by measurement errors and is prone to misjudgment. Traditional machine learning methods require manual feature extraction, and the ability of shallow machine learning to represent complex functions is limited in the case of limited samples. Its generalization ability is restricted for complex non-linear problems. Moreover, there is a very strong similarity in the motion of radar true and false targets, so its judgment accuracy is relatively low. The artificial neural network model has better feature extraction ability and stronger learning ability, but it relies on data-driven. The radar false targets are caused by various factors, including discrete and irregular as well as continuous and persistent false targets. There are false targets formed by distance, angle, and speed deception, as well as false targets formed by compound deception. The formation factors of false targets are complex and there is a very strong motion correlation between true and false targets, making it difficult for the artificial neural network to effectively identify radar true and false targets. Summary of the Invention

[0004] To solve the problem of low accuracy in identifying true and false targets by existing radars in air combat, the present invention proposes a method for identifying true and false radar targets under strong interference. Starting from the motion laws of true and false radar targets, considering the data differences between true and false targets under different influencing factors, and making full use of the attribute features in different dimensions, it effectively solves the problem of difficult distinction in the case of strong similarity in the motion laws of true and false targets, and improves the accuracy of judging true and false radar targets in air combat.

[0005] A method for identifying true and false radar targets based on parameter screening under strong interference, comprising the following steps:

[0006] (1) Data extraction and calibration

[0007] In an air combat simulation platform, a large amount of simulation data is collected, the radar observation data and the corresponding true values are extracted, and the true and false nature of the radar observation targets in the simulation data is calibrated using the true values. The specific steps are as follows:

[0008] (1a) Conduct 2V2 fighter air combat simulation using a distributed simulation environment, set the fighter to turn on the electronic countermeasure system, set the step size to 1 second, and collect radar target data in no less than 1000 battles;

[0009] (1b) The simulation includes three different types of sensor data: radar, optoelectronic, and electronic warfare. Since the algorithm only generates false targets for radar data, the radar observation data is extracted for subsequent algorithm input. The attributes of the radar observation data are selected and divided into three categories: target number, position, and speed. When the target is not interrupted, the same number can identify the same target. When the target is lost, a new number is assigned to the new aircraft. The position parameters include: target distance, target azimuth angle, target elevation angle, target horizontal heading angle, target northward position, target celestial position, and target eastward position, a total of 7 parameters. The speed parameters include: target northward speed, target celestial speed, target eastward speed, target radial speed, target range rate, and target Mach number, a total of 6 parameters. There are a total of 14 parameters.

[0010] (1c) Calibrate the true and false nature of the radar observation targets. Set the target with the closest observed target position to the actual target position and an error not exceeding 3000m as the true target, and the rest of the targets as false targets.

[0011] (2) Analysis of the characteristics of true and false targets

[0012] Explore the generation principles of true and false targets in air combat, analyze the track characteristics, angle characteristics, and speed characteristics of true and false targets, extract discriminative data features, and provide theoretical support for formulating true and false target recognition strategies. Under different interference modes (0: automatic; 1: noise false target; 2: noise towing interference; 3: all), and different maneuver types (taking values from 0 to 9), visualize the true and false observation data, analyze the track, angle, and speed characteristics of true and false targets, and find discriminative feature dimensions. The specific steps are as follows:

[0013] (2a) Analyze the three-dimensional track points of the target, the northward, upward, and eastward coordinate values, and the differences in the observed values of the true and false targets of the one-dimensional distance information, extract the characteristics of the true and false targets, and observe the similarities and differences between the observed values and the true values, considering the influence brought by the observation error.

[0014] (2b) Analyze the elevation angle and azimuth angle information of the radar data, combine with the change of the one-dimensional observation distance, observe the differences between the true and false targets, and the data characteristics of the true and false targets corresponding to the same target.

[0015] (2c) Analyze the northward, upward, and eastward speeds of the true and false targets and observe the similarities and differences.

[0016] (3) Grouping of target data

[0017] In the current simulation environment, there are at most 2 true targets observed by the radar, and there are several false targets for each true target. Analyze the observed data of the true and false targets of the same target and the observed data of different targets. Use information such as the azimuth angle, distance, and altitude of the target to group the observed data of different targets, that is, group a true target of the same target and several false targets corresponding to this true target into one group, which is convenient for subsequent true and false target recognition.

[0018] (4) Inter-frame target association

[0019] Use the target position information and target number to perform inter-frame association on the grouped data. And update the frequency of each group of data according to the target number. After the step (3) groups the intra-frame data, the grouped data only saves the current observation information and lacks historical information. In order to use the historical information of the target, perform inter-frame association on the currently grouped targets. Given the inter-frame time interval and the maximum flight speed of the aircraft, it can be inferred that there is a distance constraint for the same target between frames, and when the target is not interrupted, the same target number can mark the same target. Using the above information, inter-frame target association can be achieved.

[0020] (5) Radar true and false target recognition

[0021] Identify true and false radar targets by taking advantage of the conditions that true targets have longer continuity and the movement of targets is restricted by the maximum speed. Among them, continuity can effectively filter out discrete false targets, and false targets causing jumps can be filtered out through movement constraints. The specific steps are as follows:

[0022] (5a) Discrete false targets only exist in a few frames. Utilize the target continuity information to filter out discrete false targets.

[0023] (5b) The movement of true targets is restricted by spatial positions. When there are large jumps in positions, it can be determined that the jumping targets are false targets.

[0024] (5c) It can be observed from data analysis that from the perspective of our aircraft, enemy false targets are distributed behind true targets, that is, the observed distance of false targets of the same target is farther than that of true targets. This characteristic can be used to distinguish true and false targets.

[0025] (5d) When there is no interruption of the target, the same target number can mark the same target. When a target is identified as a false target, it can be inferred that all subsequent targets with the same number are false targets.

[0026] (6) Model effect evaluation

[0027] Integrate the algorithm model into the simulation environment, make real-time true / false judgments on the radar observed targets, and evaluate the recognition effect of the algorithm model according to the evaluation criteria. When there are radar targets within 3 km of the true value, the true targets are the radar targets that meet the distance within 3 km; when there are no radar targets within 3 km of the true value, the true targets are the closest radar targets or output that there are no true targets.

[0028] The present invention makes full use of the different dimensional attribute characteristics of radar motion data, considers the data differences between true and false targets under different influencing factors, and effectively solves the problem of difficult distinction in the case of strong similarity in the motion laws of true and false targets. The simulation results show that the recognition rate of true and false targets in 1000 simulations of the present invention is at least greater than 95% in 700 simulations, improving the accuracy of true / false recognition of fighter radar targets. Description of the drawings

[0029] Figure 1 is the flow chart of the present invention.

[0030] Figure 2 is the result graph of trajectory analysis of the present invention.

[0031] Figure 3 is the result graph of angle analysis of the present invention.

[0032] Figure 4 a is the result graph of grouping of the present invention.

[0033] Figure 4b is the recognition result diagram of the present invention. Detailed implementation manners

[0034] The following will be described in detail in combination with the radar targets in a certain air combat simulation platform. It should be noted that when this method is applied to different simulation systems or real environments, only the data sources are different. In addition, the described embodiments are only intended to facilitate the understanding of the present invention and do not limit any actual applications.

[0035] A method for identifying true and false radar targets based on parameter screening under strong interference conditions, and the implementation process is as Figure 1 shown, including the following steps:

[0036] (1) Data extraction and calibration

[0037] In the air combat simulation platform, a large amount of simulation data is collected, the radar observation data and the corresponding true values are extracted, and the true and false nature of the radar observation targets in the simulation data is calibrated using the true values. Specifically as follows:

[0038] (1a) Use a distributed simulation environment to conduct 2V2 fighter air combat simulations, set the fighter to turn on the electronic countermeasure system, set the step size to 1 second, and collect the radar target data in no less than 1000 battles;

[0039] (1b) The simulation includes three different types of sensor data: radar, optoelectronic, and electronic warfare. Since the algorithm only generates false targets for radar data, the radar observation data is extracted for subsequent algorithm input. Select the attributes of the radar observation data, which are divided into 3 categories: target number, position, and speed; when the target is not interrupted, the same number can mark the same target, and when the target is lost, a new number is assigned to the new aircraft; the position parameters include: target distance (meters), target azimuth angle (radians), target elevation angle (radians), target horizontal heading angle (radians), target northward position (meters), target skyward position (meters), target eastward position (meters), a total of 7 parameters; the speed parameters include: target northward speed (meters per second), target skyward speed (meters per second), target eastward speed (meters per second), target radial speed (meters per second), target range rate of change (meters per second), target Mach number (Mach), a total of 6 parameters. A total of 14 parameters.

[0040] (1c) Calibrate the true and false nature of the radar observation targets, and set the target with the closest observed target position to the actual target position and an error not exceeding 3000m as the true target, and the rest of the targets are false targets.

[0041] (2) Analysis of the characteristics of true and false targets

[0042] Explore the generation principles of true and false targets in air combat, analyze the track characteristics, angle characteristics, and speed characteristics of true and false targets, extract discriminative data features, and provide theoretical support for formulating true and false target recognition strategies. Under different interference modes (0: automatic; 1: noise false target; 2: noise drag interference; 3: all available) and different maneuver types (values from 0 to 9), visualize the true and false observation data, analyze the track, angle, and speed characteristics of true and false targets, and find discriminative feature dimensions. The steps are as follows:

[0043] (2a) Analyze the differences in the observed values of the true and false targets of the three-dimensional track points of the target, the northward, skyward, and eastward coordinate values, and the one-dimensional distance information, extract the characteristics of the true and false targets, observe the similarities and differences between the observed values and the true values, and consider the influence brought by the observation value error.

[0044] (2b) Analyze the pitch angle and azimuth angle information of the radar data, and combine the change of the one-dimensional observation distance to observe the differences between the true and false targets, as well as the data characteristics of the true and false targets corresponding to the same target.

[0045] (2c) Analyze the northward, skyward, and eastward speeds of the true and false targets and observe the similarities and differences.

[0046] In this implementation case, the visualization diagrams are as shown in Figure 2 、 Figure 3 It can be observed that the rules of false targets are as follows: there are track characteristics of similar distance false targets generated parallel to the true track, there are tracks of false targets that are not parallel to the true track but are parallel to each other, and there are also some discrete and irregular false targets. In most cases, the azimuth angles of different targets have different values, but there are cross values. And for several false targets of the same target and the true target at the same moment, they are basically the same in the azimuth angle dimension. In the pitch angle dimension, the observed values have a similar trend to the true values, but the overall change in the pitch angle is less than 1°, so the trend is not very important and it can be considered that the pitch angles are not discriminative because they are very close. It can be seen from the distance that most of the observed true and false targets are farther from the distance of the true value points to our aircraft, but there are also a small number of cases where the points are closer. The true and false targets are discriminable in terms of distance.

[0047] (3) Grouping of target data

[0048] In the current simulation environment, there are at most 2 true targets observed by the radar and there are several false targets for each true target. As can be seen from step (2), although different targets are discriminable in the azimuth angle dimension, there are also a small number of frames with equal values. Therefore, there is an error when grouping only based on the azimuth angle information. In this implementation case, the observed data of different targets are grouped using information such as the azimuth angle, distance, and altitude of the target. The same target group contains one true target and several false targets corresponding to this true target. The grouping results are as shown in Figure 4 a, and the specific steps are as follows:

[0049] (3a) Based on the multiple radar target data collected in the current frame, calculate the maximum difference in azimuth angles between the targets. If the difference in azimuth angles is greater than 2 degrees, it is considered that the azimuth dimension is separable. Select the maximum and minimum azimuth angles, and group the two observed targets based on the two endpoints.

[0050] (3b) If the azimuth is not separable, determine whether the distances between the current observed targets are distinguishable, that is, whether the maximum distance between the targets is greater than 100,000 meters. If the distances are separable, select the two points with the maximum distance as the reference for grouping and partitioning.

[0051] (3c) Otherwise, determine whether the altitude values of the observed targets are distinguishable, that is, whether the maximum difference in altitude values between the targets is greater than 800 meters. If the altitudes are separable, group and partition according to the maximum and minimum altitudes. Otherwise, it is considered that only one group of aircraft data is observed in the current frame, that is, there is only one real target and several false targets.

[0052] (4) Inter-frame target association

[0053] Perform inter-frame association on the grouped data. After grouping the data in step (3), the grouped data only stores the current observed information and lacks historical information. To use the historical information of the targets, perform inter-frame association on the currently grouped targets. The specific steps are as follows:

[0054] (4a) When there is no interruption of the target, the same target number can mark the same target. According to the historical grouping information, search whether the target number belongs to a certain group. If the result is found, directly group. Otherwise, make the next judgment.

[0055] (4b) For the distances between the current frame's observed groups and the two grouped true targets in the previous frame, perform inter-frame association. The closer the distance, the higher the correlation is considered.

[0056] (4c) When there is only one target in the previous frame and the current frame, the strategy in step (4b) does not take effect. At this time, according to the inter-frame time interval of 1 s and the maximum flight speed of the aircraft (less than the speed of sound of 340 m / s), it can be inferred that there is a distance constraint for the same target between frames. Considering the observation error, the constraint threshold is taken as 500 m.

[0057] (4d) Update the frequency of each group of data according to the target number, and delete the expired data.

[0058] (5) Radar true and false target identification

[0059] Utilize the conditions that true targets have long continuity and the target motion is restricted by the maximum speed to identify radar true and false targets. Among them, continuity can effectively filter out discrete false targets, and motion constraints can filter out false targets that cause jumps.

[0060] (5a) Discrete false targets only exist for a few frames. By using the target continuity information, discrete false targets can be filtered out. That is, when a target appears less than 5 frames, it can be considered a false target.

[0061] (5b) The movement of real targets is restricted by spatial positions. When there is a large jump in position, it can be determined that the jumping target is a false target.

[0062] (5c) It can be observed from data analysis that from the perspective of our aircraft, enemy false targets are distributed behind real targets, that is, for the same target, the observation distance of the false target is farther than that of the real target. This feature can be used to distinguish between true and false targets.

[0063] (5d) When there is no interruption of the target, the same target number can mark the same target. When a target is identified as a false target, it can be inferred that all subsequent targets with the same number are false targets.

[0064] (6) Judgment of the authenticity of radar targets

[0065] Integrate the algorithm model into the simulation environment to make real-time judgments on the authenticity of radar observed targets, and evaluate the recognition effect of the algorithm model according to the evaluation criteria. When there are radar targets within 3 km of the true value, the true targets are the radar targets that meet the distance within 3 km; when there are no radar targets within 3 km of the true value, the true targets are the closest radar targets or output that there is no real target.

Claims

1. A method for identifying true and false radar targets based on parameter screening under strong interference, characterized in that: It includes the following steps: (1) Data extraction and calibration A large amount of simulation data is collected in the air combat simulation platform, the radar observation data and the corresponding true values are extracted, and the true and false nature of the radar observation targets in the simulation data is calibrated using the true values; the specific steps are as follows: (1a) Conduct 2V2 fighter air combat simulation using a distributed simulation environment, set the fighter to turn on the electronic countermeasure system, set the step size to 1 second, and collect radar target data in no less than 1000 battles; (1b) The simulation includes three different types of sensor data: radar, optoelectronic, and electronic warfare. Extract the radar observation data for subsequent algorithm input; select the attributes of the radar observation data, which are divided into 3 categories: target number, position, and speed; when the target is not interrupted, the same number indicates the same target, and if the target is lost, a new number is assigned to the new aircraft; The position parameters include: 7 parameters of target distance, target azimuth angle, target elevation angle, target horizontal course angle, target northward position, target celestial position, and target eastward position; The speed parameters include: 6 parameters of target northward speed, target celestial speed, target eastward speed, target radial speed, target range rate of change, and target Mach number; a total of 14 parameters; (1c) Calibrate the true and false nature of the radar observation targets. Set the target with the closest observed target position to the actual target position and an error not exceeding 3000m as the true target, and the rest of the targets as false targets; (2) Analysis of the characteristics of true and false targets Under different interference modes and different maneuver types, visualize the true and false observation data, analyze the track, angle, and speed characteristics of true and false targets, and find the discriminative feature dimensions; The specific steps are as follows: (2a) Analyze the differences in the observed values of the three-dimensional space track points, northward, celestial, and eastward coordinate values, and one-dimensional distance information of true and false targets, extract the characteristics of true and false targets, and observe the similarities and differences between the observed values and the true values, considering the influence of the observed value error; (2b) Analyze the elevation angle and azimuth angle information of the radar data, combine the one-dimensional observed distance change, observe the differences between true and false targets, and the data characteristics of true and false targets corresponding to the same target; (2c) Analyze the northward, celestial, and eastward speeds of true and false targets and observe the similarities and differences; (3) Grouping of target data In the current simulation environment, there are at most 2 true targets among the radar observation targets, and there are several false targets for each true target. Analyze the observed data of the true and false targets of the same target and the observed data of different targets, and use the target azimuth angle, distance, and height information to group the observed data of different targets, that is, group one true target of the same target and several false targets corresponding to the true target for subsequent true and false target identification; (4) Inter-frame target association Using the target position information and the target number, perform inter-frame association on the grouped data, and update the frequency of each group of data according to the target number; after grouping the intra-frame data in step (3), the grouped data only stores the current observation information and lacks historical information. To use the historical information of the target, perform inter-frame association on the currently grouped targets. Given the inter-frame time interval and the maximum flight speed of the aircraft, infer the distance constraint for the same target in different frames. When there is no interruption of the target, the same target number indicates the same target. Use the above information to achieve inter-frame target association; (5) Radar true and false target identification Use the continuity of true targets and the condition that the target motion is constrained by the maximum speed to identify radar true and false targets. Among them, the continuity effectively filters out discrete false targets, and the motion constraint filters out false targets that cause jumps; the specific steps are as follows: (5a) Discrete false targets only exist in a few frames. Use the target continuity information to filter out discrete false targets; (5b) The motion of true targets is constrained by spatial position. When there is a large jump in position, determine that the jumping target is a false target; (5c) It is observed from data analysis that from the perspective of our aircraft, enemy false targets are distributed behind true targets, that is, for the same target, the observed distance of the false target is farther than that of the true target; use this characteristic to achieve the identification of true and false targets; (5d) When there is no interruption of the target, the same target number indicates the same target. When a target is identified as a false target, infer that all subsequent occurrences of this target are false targets; (6) Model effect evaluation Integrate the algorithm model into the simulation environment, perform real-time true and false judgment on the radar observed targets, and evaluate the recognition effect of the algorithm model according to the evaluation criteria; when there is a radar target within 3 km of the true value, the true target is the radar target that meets the distance within 3 km; when there is no radar target within 3 km of the true value, the true target is the closest radar target or output no true target.

Images