UAV and flying bird classification method based on linear scoring model of radar multi-features

Abstract

The present invention belongs to the field of radar target recognition technology and relates to a method for classifying drones and flying birds based on a linear scoring model of radar multiple features. The method comprises extracting characteristic parameters of a flying target; calculating the bird probability value of the target corresponding to each characteristic parameter; using a linear scoring method to weight the bird probability value of the target corresponding to each characteristic parameter to calculate the final bird probability value of the target; and determining the type of the target based on the final bird probability value of the target. The present invention calculates the bird probability value by extracting multiple features of the target in the radar, and uses a linear scoring method to weight and fuse the multiple bird probability values ​​to obtain the final bird probability value. Finally, the final probability value is evaluated to determine the type of the target, thereby reducing the impact of the poor performance of a single feature of the target and being able to more accurately determine the target type.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar target recognition, and in particular relates to a method for classifying drones and flying birds based on a linear scoring model of radar multi-features. Background Art

[0002] With the rapid development of drone technology and the drone industry, the need to quickly identify non-cooperative drones and develop response strategies has become increasingly urgent. Targets with low altitude, low speed, and a small radar cross-section (RCS), of which birds and non-cooperative drones are the most representative, are easily confused during target identification due to limited radar observation capabilities. This can easily lead to numerous false alarms, severely impacting operator decision-making efficiency and even creating security vulnerabilities. Therefore, accurately classifying these two types of targets is crucial to the security and protection systems of airports and key areas.

[0003] Currently, algorithms for classifying drones and birds in radar fall into two main categories: single-feature classification algorithms and machine learning-based classification algorithms. The former typically rely on a single target feature (such as distance, speed, and radar reflection intensity) for classification. For example, they distinguish drones from birds based on the stability of the target's flight speed, assuming that drones have a more stable flight speed while birds have more volatile speeds. However, these methods have significant limitations: a single feature cannot fully characterize the target's motion characteristics. Misjudgments are easily made when this feature is interfered with by environmental noise or when the target is in a special motion state, such as a drone performing a hovering mission or a bird gliding. Furthermore, when a bird maintains a constant speed for a short period of time or a drone experiences speed fluctuations due to a powertrain failure, accurate classification based solely on speed features is impossible.

[0004] The latter is a machine learning-based classification algorithm that learns multiple target features by building neural networks or ensemble learning models (such as random forest models). For example, dynamic classification algorithms based on radar track sequences use a multi-headed Convolutional Neural Network (CNN) to extract position, velocity, and radiation features, combining long- and short-term confidence likelihood decision-making with a multi-factor metric module for classification. While this type of algorithm can improve classification accuracy to a certain extent, its performance is highly dependent on the quality and size of the training dataset.

[0005] Currently, there is an extremely scarce public dataset of radar bird and drone data. Actual projects often face problems such as insufficient data samples and uneven data distribution (for example, the proportion of drone data is much higher than that of bird data). This makes model training difficult and the generalization ability poor, making it difficult to apply stably in actual scenarios. Summary of the Invention

[0006] In order to solve the above technical problems, the present invention provides a method for classifying drones and flying birds based on a linear scoring model of multiple radar features, comprising:

[0007] Extracting characteristic parameters of the flight target; the characteristic parameters include at least two of the following: track distance variation characteristics, signal-to-noise ratio characteristics, heading fluctuation characteristics, and Doppler stability characteristics;

[0008] Calculate the flying bird probability value of the target corresponding to each characteristic parameter respectively;

[0009] The linear scoring method is used to weight the bird-flying probability value of the target corresponding to each characteristic parameter to calculate the final bird-flying probability value of the target;

[0010] The target type is determined based on the target's final bird probability value.

[0011] On the basis of the above technical solution, the present invention can also be improved as follows.

[0012] Further, the bird flying probability value of the target corresponding to the track distance change feature is calculated, including: assuming that the bird flying probability value of the target corresponding to the track distance change feature is , the maximum distance of the target track is , the minimum distance of the target track is , the target sailing distance is , the first distance difference threshold is , the second distance difference threshold is , the reference value of the first flying bird probability is , is a constant, and is a set positive real number, then:

[0013] ;

[0014] ;

[0015] and: .

[0016] Furthermore, when calculating the bird probability value of the target corresponding to the change in track distance, the judgment of the fluctuation size of the navigation distance is added to calculate the corresponding bird probability value, including: setting represents the target track length, Indicates the target The distance between the sampling points and the radar, the distance set of the target track is , the distance change direction matrix between the statistical target and the radar is , For the statistical purpose The direction of distance change between the sampling point and the radar, , The assignment range is , -1 means the The distance between the sampling point and the radar is relative to the The distance between the sampling point and the radar decreases, 0 means the distance between the target and the radar does not change, 1 means the distance between the target and the radar increases; the difference between the distances between the two sampling points and the radar is , ; is the threshold of the difference between the distances of two sampling points and the radar, and is used to determine the direction of change in the distances between two adjacent sampling points and the radar. The determination conditions are:

[0017] like , indicating that the distance fluctuates, whereas , indicating that the distance is stable;

[0018] when and When , it means that the distance between the target and the radar becomes larger. Let the number of targets whose distance from the radar increases in the track be , the distance between the target and the radar increases at the corresponding time position. Assign a value of 1, that is: ;

[0019] when and When , it means that the track distance becomes smaller. Let the number of targets whose distance from the radar becomes smaller in the track be , the direction of movement in which the distance between the target and the radar decreases at the corresponding time position Assign a value of -1, that is: ;

[0020] like , it means that the track distance has not changed at this time. Let the number of unchanged distances be , the distance between the target and the radar remains unchanged at the corresponding time position in the running direction Assign a value of 0, that is: ;

[0021] Suppose the probability value of the target bird flying corresponding to the change in the track distance after the update is The first threshold of the number of distance changes in the set track is The second threshold value of the number of distance changes in the track is set to , , the reference value of the second bird probability is , is a constant, and , for the total length of the target track, statistics are 、 and The value and movement direction matrix , in order to calculate the probability value of the bird flying corresponding to the distance change in the updated track ,but:

[0022] ;

[0023] .

[0024] Furthermore, considering the target trajectory, the corresponding bird probability value is updated, including: setting the target bird probability value corresponding to the updated track distance change to be The first threshold of the number of distance changes in the set track is The second threshold value of the number of distance changes in the track is set to , , the reference value of the second bird probability is , is a constant, and , for the total length of the target track, statistics are 、 and The value and movement direction matrix , in order to calculate the target bird probability value corresponding to the updated track distance change , The drone trajectory flag indicating the current situation, Indicates that the current situation is that the drone is executing the trajectory flag. Indicates that the current situation is not the drone executing the trajectory flag, judging the statistical distance change direction matrix Whether the value of is the same and non-zero within a period of time, that is: or , and the duration is greater than the first set time threshold, and after this period of time, there is a phenomenon of the distance change direction being opposite for a period of time, and the duration is the second set time threshold, then the target is a drone, and the target is not assigned a bird probability value. Calculate the bird probability value The calculation formula is updated to:

[0025] .

[0026] Furthermore, the probability of flying birds is calculated by using the statistical distance rasterization ratio, including:

[0027] Assume that all distance positions in the target track are , the horizontal coordinate in the statistical track is , the vertical axis is , Indicates the sampling points, Indicates the The horizontal coordinate of the sampling point, Indicates the The vertical coordinate of the sampling point, The maximum value is , The minimum value is , The maximum value is , The minimum value is , set the grid size to , the number of horizontal grids is , the number of vertical grids is , the grid occupancy matrix is , the number of grids occupied by the point cloud is , the grid occupancy ratio is , assign the point cloud to the grid and count the grid occupancy to obtain the grid occupancy matrix, and count the number of grids occupied by the point cloud. The size of the grid occupancy ratio reflects the target track fluctuation. The larger the value of , the smaller the point cloud occupies, then:

[0028] ;

[0029] ;

[0030] ;

[0031] set up Indicates the set grid ratio threshold, and uses the statistical distance rasterization ratio method to get the bird probability value , 、 and is a set positive real number; the target direction is one-way, the probability of the target being a drone is greater than the probability of the target being a bird, and vice versa, the probability of the target being a bird is greater than the probability of the target being a drone. The calculation formula is:

[0032] ;

[0033] ;

[0034] Suppose the probability that the target is a flying bird is , the probability value of the target bird corresponding to the change of track distance is , the probability value of the target bird corresponding to the updated track distance change , for the track distance change characteristics, the probability that the target is a flying bird type The calculation formula is:

[0035] .

[0036] Furthermore, the flying bird probability value of the target corresponding to the signal-to-noise ratio feature is calculated, including:

[0037] Assume that the probability value of the target bird corresponding to the signal-to-noise ratio feature is , the corresponding signal-to-noise ratio in the target track is , target track The signal-to-noise ratio at time is , represents the target track length, represents the first signal-to-noise ratio threshold set by experimental data, represents the second signal-to-noise ratio threshold set by experimental data, and By positive real numbers, is the corresponding mean signal-to-noise ratio in the target track, then:

[0038] ;

[0039] ;

[0040] .

[0041] Furthermore, the probability value of the target bird flying corresponding to the heading fluctuation characteristic is calculated, including: setting the heading of the target to , the heading difference between two adjacent times is , Indicates the total number of turns during the target's flight time. Indicates the threshold value of the total number of turns within the set target flight time. represents the probability value of the flying bird in the experimental setting, Indicates the target track length, and the target bird probability value corresponding to the heading fluctuation characteristic is ,but:

[0042] .

[0043] Furthermore, the bird probability value of the target corresponding to the Doppler stability characteristic is calculated, including:

[0044] Assume that the probability value of the target bird corresponding to the Doppler stability characteristic is , the Doppler value of the target during flight is expressed as: , Indicates the sampling points, Indicates the target track length, and the Doppler value in a time period is expressed as , The Doppler threshold set for the experiment, is the maximum Doppler value, is the minimum Doppler value, if , then the target is a drone, is a negative real number, otherwise .

[0045] Furthermore, when the characteristic parameters include track distance change characteristics, signal-to-noise ratio characteristics, heading fluctuation characteristics and Doppler stability characteristics, the probability value of the target corresponding to the track distance change characteristics is: , the flying bird probability value of the target corresponding to the signal-to-noise ratio feature is The flying bird probability value of the target corresponding to the heading fluctuation characteristic is The probability value of the target bird corresponding to the Doppler stability characteristic is , The weight is , The weight is , The weight is , The weight is ,but:

[0046] .

[0047] Furthermore, let the target type be , The value of 1 indicates that the target is a drone. The value of 0 indicates that the target is a flying bird, and the final probability value of the flying bird is , is the threshold value of the bird probability value, then:

[0048] .

[0049] The beneficial effects of the present invention are as follows: the present invention calculates a bird probability value by extracting multiple features of the target in the radar, including track distance change, signal-to-noise ratio median, heading fluctuation characteristics and Doppler stability characteristics, and uses a linear scoring method to weightedly fuse multiple bird probability values ​​to obtain a final bird probability value. Finally, the final probability value is evaluated to determine the type of target, thereby reducing the impact of poor performance of a single feature of the target and being able to more accurately determine the target type. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Schematic diagram of the drone and bird classification method based on the linear scoring model of radar multi-features provided by the present invention;

[0051] Figure 2 Schematic diagram of the flow of the drone and bird classification method based on the linear scoring model of radar multi-features of the present invention;

[0052] Figure 3 Schematic diagram of target track distance;

[0053] Figure 4 Schematic diagram of target track distance changing with time. DETAILED DESCRIPTION

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0055] As an example, as shown in the attached Figure 1 As shown, to solve the above technical problems, this embodiment provides a method for classifying drones and flying birds based on a linear scoring model of multiple radar features, including:

[0056] Extracting characteristic parameters of the flight target; the characteristic parameters include at least two of the following: track distance variation characteristics, signal-to-noise ratio characteristics, heading fluctuation characteristics, and Doppler stability characteristics;

[0057] Calculate the flying bird probability value of the target corresponding to each characteristic parameter respectively;

[0058] The linear scoring method is used to weight the bird-flying probability value of the target corresponding to each characteristic parameter to calculate the final bird-flying probability value of the target;

[0059] The target type is determined based on the target's final bird probability value.

[0060] When performing flight missions, drones often need to meet requirements such as energy conservation and smooth motion. Consequently, drones tend to maintain relatively stable motion and a smoother trajectory. However, birds, whose movements are often related to foraging and avoiding danger, exhibit greater maneuverability and greater velocity and trajectory fluctuations. To address these motion characteristics, the present invention classifies targets by extracting target track distance variation, median signal-to-noise ratio, heading fluctuation, and Doppler value.

[0061] Compared with drones, the distance of bird tracks fluctuates greatly, resulting in a relatively small distance difference relative to the radar, and the distance fluctuates greatly. Figure 2The flowchart of the present invention is shown as follows. The present invention calculates the probability value of a flying bird by extracting multiple features of the target in the radar, such as: track distance change features (including: track distance, navigation distance fluctuation size and distance rasterization ratio), signal-to-noise ratio features, heading fluctuation features and Doppler stability features, and uses a linear scoring method to weightedly fuse multiple flying bird probability values ​​to obtain a final flying bird probability value. Finally, the final probability value is evaluated to determine the type of target. In the radar classification of drones and flying birds, when the individual features of some targets perform poorly, for example, the target trajectory is short, the present invention extracts multiple features and uses a linear scoring method to weightedly fuse each feature, thereby reducing the impact of the poor performance of a single feature of the target and being able to more accurately determine the target type.

[0062] Optionally, calculating the bird flying probability value of the target corresponding to the track distance change feature includes: setting the bird flying probability value of the target corresponding to the track distance change feature to be , the maximum distance of the target track is , the minimum distance of the target track is , the target sailing distance is , the first distance difference threshold is , the second distance difference threshold is , the reference value of the first flying bird probability is , is a constant, and is a set positive real number, then:

[0063] ;

[0064] ;

[0065] and: However, in reality, some birds have long flight distances, so relying solely on flight distance to determine target type is not ideal. Therefore, we add a method to determine the fluctuation of flight distance.

[0066] Optionally, when calculating the bird probability value of the target corresponding to the track distance change, add a judgment on the size of the navigation distance fluctuation and calculate the corresponding bird probability value, including: setting represents the target track length, Indicates the target The distance between the sampling points and the radar, the distance set of the target track is , the distance change direction matrix between the statistical target and the radar is , For the statistical purpose The direction of distance change between the sampling point and the radar, , The assignment range is , -1 means the The distance between the sampling point and the radar is relative to the The distance between the sampling point and the radar decreases, 0 means the distance between the target and the radar does not change, 1 means the distance between the target and the radar increases; the difference between the distances between the two sampling points and the radar is , ; is the threshold of the difference between the distances of two sampling points and the radar, and is used to determine the direction of change in the distances between two adjacent sampling points and the radar. The determination conditions are:

[0067] like , indicating that the distance fluctuates, whereas , indicating that the distance is stable;

[0068] when and When , it means that the distance between the target and the radar becomes larger. Let the number of targets whose distance from the radar increases in the track be , the distance between the target and the radar increases at the corresponding time position. Assign a value of 1, that is: ;

[0069] when and When , it means that the track distance becomes smaller. Let the number of targets whose distance from the radar becomes smaller in the track be , the direction of movement in which the distance between the target and the radar decreases at the corresponding time position Assign a value of -1, that is: ;

[0070] like , it means that the track distance has not changed at this time. Let the number of unchanged distances be , the distance between the target and the radar remains unchanged at the corresponding time position in the running direction Assign a value of 0, that is: ;

[0071] Suppose the probability value of the target bird flying corresponding to the change in the track distance after the update is The first threshold of the number of distance changes in the set track is The second threshold value of the number of distance changes in the track is set to , , the reference value of the second bird probability is , is a constant, and , for the total length of the target track, statistics are 、 and The value and movement direction matrix , in order to calculate the probability value of the bird flying corresponding to the distance change in the updated track ,but:

[0072] ;

[0073] .

[0074] To avoid the following Figure 3 The target with the track shown above can be easily judged as a flying bird using the above method. Considering the target track, the corresponding flying bird probability value is updated. Figure 3 In the target track distance diagram shown, the horizontal axis is Coordinate, vertical axis is Coordinates. The schematic diagram of the target track distance changing with time is shown in the attached figure. Figure 4 As shown, the horizontal axis is time, unit: s, and the vertical axis is target track distance, unit: km.

[0075] Optionally, considering the target trajectory, the corresponding bird probability value is updated, including: setting the target bird probability value corresponding to the updated track distance change to be The first threshold of the number of distance changes in the set track is The second threshold value of the number of distance changes in the track is set to , , the reference value of the second bird probability is , is a constant, and , for the total length of the target track, statistics are 、 and The value and movement direction matrix , in order to calculate the target bird probability value corresponding to the updated track distance change , The drone trajectory flag indicating the current situation, Indicates that the current situation is that the drone is executing the trajectory flag. Indicates that the current situation is not the drone executing the trajectory flag, judging the statistical distance change direction matrix Whether the value of is the same and non-zero within a period of time, that is: or , and the duration is greater than the first set time threshold, and after this period of time, there is a phenomenon of the distance change direction being opposite for a period of time, and the duration is the second set time threshold, then the target is a drone, and the target is not assigned a bird probability value. Calculate the bird probability value The calculation formula is updated to:

[0076] .

[0077] Optionally, the bird flying probability value is calculated using the statistical distance rasterization ratio, including:

[0078] Assume that all distance positions in the target track are , the horizontal coordinate in the statistical track is , the vertical axis is , Indicates the sampling points, Indicates the The horizontal coordinate of the sampling point, Indicates the The vertical coordinate of the sampling point, The maximum value is , The minimum value is , The maximum value is , The minimum value is , set the grid size to , the number of horizontal grids is , the number of vertical grids is , the grid occupancy matrix is , the number of grids occupied by the point cloud is , the grid occupancy ratio is , assign the point cloud to the grid and count the grid occupancy to obtain the grid occupancy matrix, and count the number of grids occupied by the point cloud. The size of the grid occupancy ratio reflects the target track fluctuation. The larger the value of , the smaller the point cloud occupies, then:

[0079] ;

[0080] ;

[0081] ;

[0082] set up Indicates the set grid ratio threshold, and uses the statistical distance rasterization ratio method to get the bird probability value , 、 and is a set positive real number; the target direction is one-way, the probability of the target being a drone is greater than the probability of the target being a bird, and vice versa, the probability of the target being a bird is greater than the probability of the target being a drone. The calculation formula is:

[0083] ;

[0084] ;

[0085] Suppose the probability that the target is a flying bird is , the probability value of the target bird corresponding to the change of track distance is , the probability value of the target bird corresponding to the updated track distance change , for the track distance change characteristics, the probability that the target is a flying bird type The calculation formula is:

[0086] .

[0087] In general, the heading of a drone is relatively stable, while the heading of a bird often changes greatly during flight due to movements such as turning and circling. Optionally, the probability value of the bird flying corresponding to the signal-to-noise ratio feature is calculated, including:

[0088] The signal-to-noise ratio feature is the median of the signal-to-noise ratio. Let the probability of the target bird flying corresponding to the signal-to-noise ratio feature be , the corresponding signal-to-noise ratio in the target track is , target track The signal-to-noise ratio at time is , represents the target track length, represents the first signal-to-noise ratio threshold set by experimental data, represents the second signal-to-noise ratio threshold set by experimental data, and By positive real numbers, is the corresponding mean signal-to-noise ratio in the target track, then:

[0089] ;

[0090] ;

[0091] .

[0092] Considering that the flying speed of UAV is relatively stable during flight, while the speed of birds is often uncontrollable and fluctuates greatly during flight, it is optional to calculate the probability value of the bird flying corresponding to the heading fluctuation characteristics of the target, including: setting the heading of the target to , the heading difference between two adjacent times is , Indicates the total number of turns during the target's flight time. Indicates the threshold value of the total number of turns within the set target flight time. represents the probability value of the flying bird in the experimental setting, Indicates the target track length, and the target bird probability value corresponding to the heading fluctuation characteristic is ,but:

[0093] .

[0094] The specific calculation method is: Set the target heading to , calculate the heading difference between two adjacent times ,in ; For heading difference For each heading difference in , determine whether these heading differences meet the Within the range (wherein the present invention experiment , ). If the heading difference is Within the range, The value of is added by 1, and the total number of turns during the target flight time is calculated. .when When it is greater than or equal to the set value, Assign values ​​as follows:

[0095] .

[0096] Considering that the flight speed of UAVs is relatively stable during flight, while the speed of birds is often uncontrollable and fluctuates greatly during flight, the present invention utilizes this feature to classify targets by judging the Doppler stability of the targets and assign corresponding bird probability values.

[0097] Optionally, calculate the bird probability value of the target corresponding to the Doppler stability characteristic, including:

[0098] Assume that the probability value of the target bird corresponding to the Doppler stability characteristic is , the Doppler value of the target during flight is expressed as: , Indicates the sampling points, Indicates the target track length, and the Doppler value in a time period is expressed as , The Doppler threshold set for the experiment, is the maximum Doppler value, is the minimum Doppler value, if , then the target is a drone, is a negative real number, otherwise .

[0099] Optionally, when the characteristic parameters include track distance change characteristics, signal-to-noise ratio characteristics, heading fluctuation characteristics and Doppler stability characteristics, the probability value of the target corresponding to the track distance change characteristics is , the flying bird probability value of the target corresponding to the signal-to-noise ratio feature is The flying bird probability value of the target corresponding to the heading fluctuation characteristic is The probability value of the target bird corresponding to the Doppler stability characteristic is , The weight is , The weight is , The weight is , The weight is ,but:

[0100] .

[0101] Optionally, set the target type to , The value of 1 indicates that the target is a drone. The value of 0 indicates that the target is a flying bird, and the final probability value of the flying bird is , is the threshold value of the bird probability value, then:

[0102] .

[0103] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for classifying drones and flying birds based on a linear scoring model with multiple radar features, characterized by: include: Extracting characteristic parameters of the flight target; the characteristic parameters include at least two of the following: track distance variation characteristics, signal-to-noise ratio characteristics, heading fluctuation characteristics, and Doppler stability characteristics; Calculate the flying bird probability value of the target corresponding to each characteristic parameter respectively; Calculate the flying probability value of the target corresponding to the heading fluctuation characteristic, including: setting the target heading to , the heading difference between two adjacent times is , Indicates the total number of turns during the target's flight time. Indicates the threshold value of the total number of turns within the set target flight time. represents the probability value of the flying bird in the experimental setting, Indicates the target track length, and the target bird probability value corresponding to the heading fluctuation characteristic is ,but: ; , calculate the heading difference between two adjacent times ,in ; For heading difference For each heading difference in , determine whether these heading differences meet the If the heading difference is within Within the range, The value of is added by 1, and the total number of turns during the target flight time is calculated. ;when When it is greater than or equal to the set value, Assignment: ； The linear scoring method is used to weight the bird-flying probability value of the target corresponding to each characteristic parameter to calculate the final bird-flying probability value of the target; The target type is determined based on the target's final bird probability value.

2. The method for classifying drones and flying birds based on a linear scoring model of radar multi-features according to claim 1 is characterized in that: Calculate the bird probability value of the target corresponding to the track distance change feature, including: assuming the bird probability value of the target corresponding to the track distance change feature is , the maximum distance of the target track is , the minimum distance of the target track is , the target sailing distance is , the first distance difference threshold is , the second distance difference threshold is , the reference value of the first flying bird probability is , is a constant, and is a set positive real number, then: ； ； and: .

3. The method for classifying drones and flying birds based on a linear scoring model using multiple radar features according to claim 2, wherein: When calculating the bird probability value of the target corresponding to the change in track distance, add the judgment of the fluctuation size of the navigation distance and calculate the corresponding bird probability value, including: setting represents the target track length, Indicates the target The distance between the sampling points and the radar, the distance set of the target track is , the distance change direction matrix between the statistical target and the radar is , For the statistical purpose The direction of distance change between the sampling point and the radar, , The assignment range is , -1 means the The distance between the sampling point and the radar is relative to the The distance between the sampling point and the radar decreases, 0 means the distance between the target and the radar does not change, 1 means the distance between the target and the radar increases; the difference between the distances between the two sampling points and the radar is , ; is the threshold of the difference between the distances of two sampling points and the radar, and is used to determine the direction of change in the distances between two adjacent sampling points and the radar. The determination conditions are: like , indicating that the distance fluctuates, whereas , indicating that the distance is stable; when and When , it means that the distance between the target and the radar becomes larger. Let the number of targets whose distance from the radar increases in the track be , the distance between the target and the radar increases at the corresponding time position. Assign a value of 1, that is: ; when and When , it means that the track distance becomes smaller. Let the number of targets whose distance from the radar becomes smaller in the track be , the direction of movement in which the distance between the target and the radar decreases at the corresponding time position Assign a value of -1, that is: ; like , it means that the track distance has not changed at this time. Let the number of unchanged distances be , the distance between the target and the radar remains unchanged at the corresponding time position in the running direction Assign a value of 0, that is: ; Suppose the probability value of the target bird flying corresponding to the change in the track distance after the update is The first threshold of the number of distance changes in the set track is The second threshold value of the number of distance changes in the track is set to , , the reference value of the second bird probability is , is a constant, and , for the total length of the target track, statistics are 、 and The value and movement direction matrix , in order to calculate the probability value of the bird flying corresponding to the distance change in the updated track ,but: ； 。 4. The method for classifying drones and flying birds based on a linear scoring model of radar multi-features according to claim 3 is characterized in that: Considering the target trajectory, update the corresponding bird probability value, including: setting the target bird probability value corresponding to the updated track distance change to be The first threshold of the number of distance changes in the set track is The second threshold value of the number of distance changes in the track is set to , , the reference value of the second bird probability is , is a constant, and , for the total length of the target track, statistics are 、 and The value and movement direction matrix , in order to calculate the target bird probability value corresponding to the updated track distance change , The drone trajectory flag indicating the current situation, Indicates that the current situation is that the drone is executing the trajectory flag. Indicates that the current situation is not the drone executing the trajectory flag, judging the statistical distance change direction matrix Whether the value of is the same and non-zero within a period of time, that is: or , and the duration is greater than the first set time threshold, and after this period of time, there is a phenomenon of the distance change direction being opposite for a period of time, and the duration is the second set time threshold, then the target is a drone, and the target is not assigned a bird probability value. Calculate the bird probability value The calculation formula is updated to: 。 5. The method for classifying drones and flying birds based on a linear scoring model of radar multi-features according to claim 4 is characterized in that: The bird flying probability value is calculated using the statistical distance rasterization ratio method, including: Assume that all distance positions in the target track are , the horizontal coordinate in the statistical track is , the vertical axis is , Indicates the sampling points, Indicates the The horizontal coordinate of the sampling point, Indicates the The vertical coordinate of the sampling point, The maximum value is , The minimum value is , The maximum value is , The minimum value is , set the grid size to , the number of horizontal grids is , the number of vertical grids is , the grid occupancy matrix is , the number of grids occupied by the point cloud is , the grid occupancy ratio is , assign the point cloud to the grid and count the grid occupancy to obtain the grid occupancy matrix, and count the number of grids occupied by the point cloud. The size of the grid occupancy ratio reflects the target track fluctuation. The larger the value of , the smaller the point cloud occupies, then: ； ； ； set up Indicates the set grid ratio threshold, and uses the statistical distance rasterization ratio method to get the bird probability value , 、 and is a set positive real number; the target direction is one-way, the probability of the target being a drone is greater than the probability of the target being a bird, and vice versa, the probability of the target being a bird is greater than the probability of the target being a drone. The calculation formula is: ； ； Suppose the probability that the target is a flying bird is , the probability value of the target bird corresponding to the change of track distance is , the probability value of the target bird corresponding to the updated track distance change , for the track distance change characteristics, the probability that the target is a flying bird type The calculation formula is: 。 6. The method for classifying drones and flying birds based on a linear scoring model of radar multi-features according to claim 1, characterized in that: Calculate the bird probability value of the target corresponding to the signal-to-noise ratio feature, including: Assume that the probability value of the target bird corresponding to the signal-to-noise ratio feature is , the corresponding signal-to-noise ratio in the target track is , target track The signal-to-noise ratio at time is , represents the target track length, represents the first signal-to-noise ratio threshold set by experimental data, represents the second signal-to-noise ratio threshold set by experimental data, and By positive real numbers, is the corresponding mean signal-to-noise ratio in the target track, then: ； ； 。 7. The method for classifying drones and flying birds based on a linear scoring model of radar multi-features according to claim 1, characterized in that: Calculate the bird probability value of the target corresponding to the Doppler stability characteristics, including: Assume that the probability value of the target bird corresponding to the Doppler stability characteristic is , the Doppler value of the target during flight is expressed as: , Indicates the sampling points, Indicates the target track length, and the Doppler value in a time period is expressed as , The Doppler threshold set for the experiment, is the maximum Doppler value, is the minimum Doppler value, if , then the target is a drone, is a negative real number, otherwise .

8. The method for classifying drones and flying birds based on a linear scoring model of radar multi-features according to claim 1, characterized in that: When the characteristic parameters include track distance change characteristics, signal-to-noise ratio characteristics, heading fluctuation characteristics and Doppler stability characteristics, the probability value of the target bird corresponding to the track distance change characteristics is , the flying bird probability value of the target corresponding to the signal-to-noise ratio feature is The flying bird probability value of the target corresponding to the heading fluctuation characteristic is The probability value of the target bird corresponding to the Doppler stability characteristic is , The weight is , The weight is , The weight is , The weight is ,but: 。 9. The method for classifying drones and flying birds based on a linear scoring model of radar multi-features according to claim 1 is characterized in that The target type is , The value of 1 indicates that the target is a drone. The value of 0 indicates that the target is a flying bird, and the final probability value of the flying bird is , is the threshold value of the bird probability value, then: 。

Images