A multi-radar cooperative target recognition method

Abstract

This invention discloses a multi-radar cooperative target recognition method, relating to the field of radar target recognition technology. It solves the problems of complex target recognition algorithm system design, large data processing volume, and single-target recognition limitations in existing technologies. The method includes: acquiring multiple static RCS data of multiple targets to be identified; obtaining the motion trajectories of multiple targets using motion equations; calculating and adding variable micro-motion features to the corresponding static RCS data to obtain multiple dynamic RCS data; adding noise to the multiple dynamic RCS data; extracting statistical features from the multiple final RCS data and inputting them into an SVM classifier to obtain multiple recognition results; and fusing the multiple recognition results to obtain fused data. This method achieves a simple design, overcomes the curse of dimensionality and nonlinear separability problems when calculating large amounts of data, and significantly increases the target recognition rate.

Description

Technical Field

[0001] This invention relates to the field of radar target recognition technology, and in particular to a multi-radar cooperative target recognition method. Background Technology

[0002] Regarding radar target identification technology, many researchers have proposed technical solutions.

[0003] In their paper "A One-Dimensional Range Image Radar Target Recognition Method Based on Dynamic Time Warping Algorithm," Wu Zhao et al. utilized the characteristic of HRRP (Human Range Image Target) to reflect the target's structural distribution and geometry. They established various target attitude template databases through angular domain partitioning, used the dynamic time warping algorithm to estimate the similarity between HRRPs and different HRRPs in the target template database, and selected the target template with the highest similarity as the target recognition result. Experimental results demonstrate the effectiveness and robustness of this method.

[0004] For example, patent application CN114078214A, entitled "A Radar Target RCS Recognition Method and Apparatus Based on Complex Neural Networks," discloses a radar target RCS recognition method and apparatus based on complex neural networks. The complex RCS data of the target is identified and classified using a complex neural network. Specifically, the complex convolutional layer performs complex convolution on the complex RCS data, activates it using the Complex ReLU activation function, and then performs max pooling. Results show that this method can effectively improve the target recognition accuracy.

[0005] In the existing technology, the "one-dimensional range profile radar target recognition method based on dynamic time warping algorithm" combines one-dimensional range as a feature with dynamic time warping algorithm for recognition. This method requires an excessively large sample training set and has a complex algorithm design.

[0006] The method described in "A Radar Target RCS Recognition Method and Device Based on Complex Neural Networks" utilizes a complex neural network algorithm. However, this algorithm suffers from several drawbacks. First, it is a "black box" algorithm, meaning that it is unclear how and why the neural network produces certain outputs, making it difficult to understand what causes the complex neural network to make certain predictions. Second, the algorithm has an excessively long learning time, which may even prevent it from achieving its learning objectives. Finally, the recognition rate is poor when data is insufficient. Summary of the Invention

[0007] This invention provides a multi-radar cooperative target recognition method, which solves the problems of complex target recognition algorithm system design, large data processing volume, and single-target recognition in the prior art. It achieves a simple design, overcomes the curse of dimensionality and nonlinear separability when calculating large amounts of data, and greatly increases the target recognition rate.

[0008] This invention provides a multi-radar cooperative target identification method, the method comprising:

[0009] Acquire multiple static RCS data of multiple targets to be identified;

[0010] The motion trajectory of the target to be identified is simulated using the equation of motion to obtain multiple motion trajectories of the target to be identified.

[0011] The variation micro-motion features of each sampling point on the multiple motion trajectories are calculated respectively to obtain multiple variation micro-motion features of each motion trajectory;

[0012] The multiple micro-motion features are added to the corresponding static RCS data to obtain multiple dynamic RCS data.

[0013] Noise is added to the multiple dynamic RCS data to obtain multiple final RCS data;

[0014] Statistical features are extracted from the multiple final RCS data, and the statistical features are input into the trained SVM classifier to obtain multiple recognition results;

[0015] The multiple recognition results are fused using fusion rules to obtain fused data.

[0016] In one possible implementation, obtaining the static RCS data of the target to be identified includes: establishing a target model and simulating the target model using electromagnetic simulation software to obtain the target's static RCS data.

[0017] In one possible implementation, the equation of motion is specifically expressed as:

[0018]

[0019] Where r represents the geocentric distance vector; μ represents the Earth's gravitational constant, μ = 3.986005 × 10⁻⁶. 4 m 3 / s 2 |r| represents the magnitude of the geocentric distance vector; ρ represents the Earth's radius; (x,y,z) T This represents the coordinates of the target to be identified.

[0020] In one possible implementation, simulating the motion trajectory of the target to be identified using motion equations to obtain the motion trajectory of the target to be identified includes:

[0021] The initial velocity, azimuth angle, and tilt angle of the target to be identified are obtained, and the current velocity component is obtained according to the motion equation.

[0022] Based on the current velocity component and the initial position of the target to be identified, the position of the target to be identified at the next moment is obtained;

[0023] The acceleration at the current moment is obtained using the two-body equation, and the current velocity component is corrected based on the current acceleration. This correction is then used to determine the position of the target to be identified at the next moment, thus obtaining the trajectory of the target.

[0024] In one possible implementation, the two-body motion equations are specifically expressed as:

[0025]

[0026] Where r represents the geocentric distance vector; μ represents the Earth's gravitational constant, μ = 3.986005 × 10⁻⁶. 4 m 3 / s 2 |r| represents the magnitude of the geocentric distance vector; This represents the second derivative with respect to the geocentric distance vector.

[0027] In one possible implementation, the micro-motion feature is specifically represented as:

[0028] θ = θ(t) = θ1sinω1t

[0029] Where θ1 represents the oscillation angle and ω1 represents the oscillation angular frequency.

[0030] In one possible implementation, the step of fusing the multiple recognition results using fusion rules to obtain fused data specifically includes:

[0031] Obtain an identification framework including the plurality of targets to be identified, and determine a plurality of pairwise exclusive subsets of the identification framework;

[0032] The recognition probability of each target to be identified in the recognition framework is made to satisfy the probability condition, and then the basic probability distribution of the multiple targets to be identified is obtained by solving the problem.

[0033] The basic probability distributions are fused using combination rules to determine the fused data.

[0034] In one possible implementation, the probability condition is specifically expressed as:

[0035]

[0036] Where m(A) represents the recognition result of the target to be identified; m(Φ) represents the recognition rate of impossible propositions; A represents multiple targets to be identified; Θ represents the empty set, i.e., impossible propositions.

[0037] In one possible implementation, the combination rule is specifically expressed as:

[0038]

[0039] Where m(A) represents the recognition result of the target to be identified; m(Φ) represents the recognition rate of impossible propositions; A represents multiple targets to be identified; Θ represents the empty set, i.e., impossible propositions; A i Proposition A i , is a subset of the recognition framework; B j Proposition B j , is a subset of the identification framework; m1 represents the basic probability assignment function 1 of evidence body 1; m2 represents the basic probability assignment function 2 of evidence body 2; k represents the conflict coefficient.

[0040] One or more technical solutions provided in this invention have at least the following technical effects or advantages:

[0041] This invention employs a multi-radar cooperative target recognition method, which includes: acquiring multiple static RCS data of multiple targets to be identified; using the static RCS data of the targets to be identified as the basis for subsequent addition of micro-motion features; obtaining the motion trajectories of the multiple targets to be identified using motion equations, and correcting the velocity of the targets to be identified using the acceleration in the motion trajectories; calculating the changing micro-motion features of each sampling point on the multiple motion trajectories to obtain multiple changing micro-motion features for each motion trajectory; adding the multiple changing micro-motion features to the corresponding static RCS data to obtain multiple dynamic RCS data; adding noise to the multiple dynamic RCS data to obtain multiple final RCS data; and adding different micro-motion features of the targets to be identified to the static RCS data to achieve the desired result. In subsequent target recognition, there is a clear feature classification. Statistical features of multiple final RCS data are extracted and input into a trained SVM classifier to obtain multiple recognition results. The SVM classification algorithm is simple in design and overcomes the curse of dimensionality and nonlinear separability by using kernel functions. Multiple recognition results are fused using fusion rules to obtain fused data. Combining data fusion and target recognition, the probability of multi-radar collaborative target recognition is greatly increased compared to single-radar target recognition. This effectively solves the problems of complex target recognition algorithm system design, large data processing volume, and single-target recognition in existing technologies. It achieves a simple design that can overcome the curse of dimensionality and nonlinear separability when calculating large amounts of data, and greatly increases the target recognition rate. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments of the present invention or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 A flowchart illustrating the steps of a multi-radar cooperative target recognition method provided in an embodiment of the present invention;

[0044] Figure 2 The three-dimensional Fire2 model provided in this embodiment of the invention is created using the modeling function of the electromagnetic simulation software FEKO.

[0045] Figure 3 The three-dimensional W78 model provided in this embodiment of the invention is created using the modeling function of the electromagnetic simulation software FEKO.

[0046] Figure 4 This invention provides a three-dimensional slotted spherical cone model created using the modeling function of the electromagnetic simulation software FEKO.

[0047] Figure 5 The three-dimensional seamless spherical cone model provided in this embodiment of the invention is created using the modeling function of the electromagnetic simulation software FEKO.

[0048] Figure 6 The three-dimensional flat-bottomed cone model provided in this embodiment of the invention is created using the modeling function of the electromagnetic simulation software FEKO.

[0049] Figure 7 The static RCS database for Fire2 provided in this embodiment of the invention has the horizontal axis representing the azimuth angle, ranging from 0 to 180° in 1° increments, and the vertical axis representing the RCS of Fire2, in dBsm.

[0050] Figure 8 The static RCS database of W78 provided in this embodiment of the invention has the horizontal axis as the azimuth angle, ranging from 0 to 180° in 1° increments, and the vertical axis as the RCS of W78, in dBsm.

[0051] Figure 9 The static RCS database for the slotted spherical cone provided in this embodiment of the invention has the horizontal axis representing the azimuth angle, ranging from 0 to 180° in 1° increments, and the vertical axis representing the RCS of the slotted spherical cone, in dBsm.

[0052] Figure 10The static RCS database for the seamless spherical cone provided in this embodiment of the invention has the horizontal axis representing the azimuth angle, ranging from 0 to 180° in 1° increments, and the vertical axis representing the RCS of the seamless spherical cone, in dBsm.

[0053] Figure 11 The static RCS database for flat-bottomed cones provided in this embodiment of the invention has the horizontal axis representing the azimuth angle, ranging from 0 to 180° in 1° increments, and the vertical axis representing the RCS of the flat-bottomed cone, in dBsm.

[0054] Figure 12 The bar chart shows the single radar identification results and multi-radar fusion identification results provided in the embodiments of the present invention;

[0055] Figure 13 This is a flowchart illustrating a specific step in using the present invention, as provided in an embodiment of the invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0057] Radar detects targets and determines their positions in space using radio waves. Over the years, radar's information acquisition capabilities have continuously improved. However, in practical applications of radar systems, due to the increasing complexity of modern environments and the further development of various target feature control technologies, target observability is decreasing. A key challenge is how to reliably detect and identify air, ground, and maritime targets at greater distances. Target recognition based on information fusion, which integrates the performance advantages of various radars, offers numerous benefits, including improving the stability and reliability of automatic target recognition systems, enhancing their anti-interference capabilities and environmental adaptability, increasing the accuracy of target recognition, and reducing uncertainty. It has become one of the main research directions in the field of automatic target recognition. Therefore, research combining information fusion technology with radar target recognition technology has significant application implications.

[0058] The present invention provides a multi-radar cooperative target identification method, which includes the following steps S101 to S107.

[0059] S101, acquire multiple static RCS data of multiple targets to be identified. Acquiring the static RCS data of the targets to be identified includes: establishing a target model and simulating the target model using electromagnetic simulation software to obtain the target's static RCS data.

[0060] In a specific embodiment of this invention, the full-angle static RCS database of the target is obtained using the electromagnetic simulation software FEKO. The targets to be identified are a flat-bottomed cone, W78, Fire2, a seamless spherical-bottomed cone, and a slotted spherical-bottomed cone. First, the physical model of the target to be identified is imported into FEKO. The radar frequency is selected as 10 GHz, the wave source is selected as a planar linearly polarized wave, multiple wave sources are used, the elevation angle is 90°, the azimuth angle is 0–180°, the step is 1°, observation is performed along the radar line of sight, and the solution algorithm is the physical optics method. Figure 2 The image shows a 3D Fire2 model created using the modeling function of the electromagnetic simulation software FEKO. Figure 3 The image shows a 3D W78 model created using the modeling function of the electromagnetic simulation software FEKO. Figure 4 The image shows a three-dimensional slotted spherical cone model created using the modeling function of the electromagnetic simulation software FEKO. Figure 5 The image shows a three-dimensional seamless spherical cone model created using the modeling function of the electromagnetic simulation software FEKO. Figure 6 The image shows a three-dimensional flat-bottomed cone model created using the modeling function of the electromagnetic simulation software FEKO. Figure 7 The image shows the static RCS database for Fire2. The horizontal axis represents the azimuth angle, ranging from 0 to 180° in 1° increments, and the vertical axis represents the RCS of Fire2, in dBsm. Figure 8 The image shows the static RCS database for W78. The horizontal axis represents the azimuth angle, ranging from 0 to 180° in 1° increments, and the vertical axis represents the RCS of W78, in dBsm. Figure 9 The image shows a static RCS database for a seamed spherical cone. The horizontal axis represents the azimuth angle, ranging from 0 to 180° in 1° increments, and the vertical axis represents the RCS of the seamed spherical cone, in dBsm. Figure 10 The image shows the static RCS database for a seamless spherical cone. The horizontal axis represents the azimuth angle, ranging from 0 to 180° in 1° increments, and the vertical axis represents the RCS of the seamless spherical cone, in dBsm. Figure 11 The image shows the static RCS database for a flat-bottomed cone. The horizontal axis represents the azimuth angle, ranging from 0 to 180° in 1° increments, and the vertical axis represents the RCS of the flat-bottomed cone, in dBsm.

[0061] S102, the motion trajectory of the target to be identified is simulated using the equations of motion to obtain the motion trajectories of multiple targets. Specifically, this includes the following steps:

[0062] (1) Obtain the initial velocity, azimuth angle, and tilt angle of the target to be identified, and obtain the current velocity component according to the motion equation; the motion equation is specifically expressed as:

[0063]

[0064] Where r represents the geocentric distance vector; μ represents the Earth's gravitational constant, μ = 3.986005 × 10⁻⁶. 4 m 3 / s 2 |r| represents the magnitude of the geocentric distance vector; ρ represents the Earth's radius; (x,y,z) T This represents the coordinates of the target to be identified.

[0065] (2) Based on the current velocity component and the initial position of the target to be identified, the position of the target to be identified at the next moment is obtained.

[0066] (3) The acceleration at the current moment is obtained using the two-body equations, and the current velocity component is corrected based on the current acceleration. This correction is then used to determine the position of the target to be identified at the next moment, thus obtaining the target's trajectory. The two-body motion equations are specifically expressed as follows:

[0067]

[0068] Where r represents the geocentric distance vector; μ represents the Earth's gravitational constant, μ = 3.986005 × 10⁻⁶. 4 m 3 / s 2 |r| represents the magnitude of the geocentric distance vector; This represents the second derivative of the distance vector from the Earth's center, i.e., the absolute acceleration.

[0069] In a specific embodiment of this invention, the selected target to be identified is an aerial target whose flight trajectory is in the outer atmosphere. All air resistance is negligible, and it is only affected by Earth's gravity to undergo inertial motion. A spherical Earth model is selected, with the northeast celestial coordinate system as the reference. The coordinates of the target to be identified are assumed to be: r1 = (x, y, z). T ,speed acceleration The Earth's radius is ρ = 6371 km, and the geocentric distance vector is r = (x, y, z + ρ). T In this coordinate system, the distance between the Earth's centers is Based on the above conditions, establish the motion equations of the target to be identified.

[0070] The trajectory of a target can be simulated using differential equations of motion. During the simulation, the magnitude of each velocity component is first obtained based on the initial velocity, azimuth, and tilt angle of the target to be identified. Then, the velocity and initial position coordinates are combined to calculate the target's position at the next moment. Then, according to the two-body motion equation, the acceleration of the target at a certain moment can be obtained. The target velocity is corrected based on this acceleration. This process is repeated. Since the time interval selected each time is very small, the target motion within each time interval can be regarded as uniform motion.

[0071] S103, calculate the change micro-motion features of each sampling point on multiple motion trajectories, and obtain multiple change micro-motion features for each motion trajectory. The change micro-motion features are specifically represented as follows:

[0072] θ = θ(t) = θ1sinω1t

[0073] Where θ1 represents the oscillation angle and ω1 represents the oscillation angular frequency.

[0074] In a specific embodiment of this invention, the targets to be identified are a flat-bottomed cone, W78, Fire2, a seamless spherical-bottomed cone, and a slotted spherical-bottomed cone. The five types of targets have motion trajectories derived from two-body motion, and the micro-motion characteristics of each type are different. This invention identifies targets based on extracted dynamic RCS data. As shown in step S101, the angle between the target and the radar is different at each moment, resulting in different static RCS data. The micro-motion characteristics affect the attitude angle of different targets relative to the radar at the same moment, thus affecting the static RCS data, which is ultimately fed back to the dynamic RCS data. The micro-motion characteristics of the five types of targets to be identified are divided into two categories: precession corresponding to Fire and W78, and oscillation corresponding to the flat-bottomed cone, slotted decoy, and seamless decoy. Precession is the target spinning first, while simultaneously rotating around the velocity tangential angle axis. The angle between the target's central axis and the velocity tangential vector is the precession angle, and the frequency of the precession rotation is the precession frequency. Oscillation refers to the periodic up-and-down oscillation of the target to be identified in the vertical plane, with the maximum angle being the oscillation angle and the variation pattern being the micro-motion characteristic.

[0075] S104, add multiple variable micro-motion features to the corresponding static RCS data to obtain multiple dynamic RCS data.

[0076] By combining the target position and radar position at each moment, the attitude angle at each moment is obtained. Then, by combining the micro-motion features obtained in step S102, the micro-motion angle is added to the attitude angle sequence.

[0077] This invention selects dynamic RCS data of the target to be identified as the identification feature. Compared with one-dimensional range image as the identification feature, dynamic RCS sequence not only takes into account static RCS data, but also reflects the target's current motion trajectory and micro-motion characteristics. Moreover, the similar shapes of the target group to be identified greatly reduce the classifiability when reflected in the one-dimensional range image; furthermore, the one-dimensional range image is an image feature, and the amount of data required to use it as the identification feature is too large, greatly increasing the identification time.

[0078] S105 adds noise to multiple dynamic RCS data to obtain multiple final RCS data.

[0079] Most targets to be identified have complex shapes and are much larger than the radar wavelength. Therefore, the target echo is a vector synthesis of the echo signals from the various parts (scatterers) of the target. When the target's position relative to the radar changes, the parameters of the reflected echo signal (amplitude, frequency, phase, etc.) also change. This characteristic of change is similar to the statistical characteristics of general noise, hence it is called radar target noise. Here, log-normally distributed multiplicative and additive noise is added, with a signal-to-noise ratio uniformly distributed between 5 and 10 dB. Adding noise makes the simulation data as close as possible to the actual data.

[0080] S106: Extract statistical features from multiple final RCS data, and input the statistical features into the trained SVM classifier to obtain multiple recognition results.

[0081] Statistical characteristics include: mean, mode, range, median, and variance.

[0082] (1) Mean, which describes the average positional information of the RCS sequence. The formula is as follows:

[0083]

[0084] (2) Mode: The mode reflects the distribution of the target's RCS sequence within the sampling time period. A smaller target size results in a smaller mode. The mode is represented as:

[0085]

[0086]

[0087]

[0088] (3) Range, the difference between the maximum and minimum values, reflects the extreme values ​​of the sample values ​​within the statistical time of the RCS sequence. This characteristic is related to the target's motion pattern and external dimensions. The formula is as follows:

[0089]

[0090] (4) Median: Sort the sequence by size, and the middle number is the median.

[0091] (5) Variance, which represents the degree to which sample values ​​deviate from their expected value, is expressed by the following formula:

[0092]

[0093] Where N represents the size of the RCS sequence; x i x j This represents the value in the RCS sequence.

[0094] Typically, the extracted data is not linearly separable in low-dimensional space. Therefore, it is necessary to use a kernel function to map the data to a high dimension to make it linearly separable. Here, a Gaussian kernel function is chosen, with a kernel parameter of 0.001 and a penalty coefficient of 1000. Then, the statistical features of multiple final RCS data are fed into a pre-configured SVM classifier for training, and then recognition is performed.

[0095] The classification algorithm selected in this invention is support vector machine. Compared with neural network algorithm, it is simple to design and overcomes the problems of dimensionality curse and nonlinear separability by using kernel function.

[0096] S107, Multiple recognition results are fused using fusion rules to obtain fused data, specifically including:

[0097] (1) Obtain an identification frame including multiple targets to be identified, and determine multiple mutually exclusive subsets of the identification frame; In a specific embodiment provided by the present invention, these five types of targets to be identified constitute an identification frame, Θ = {W78, Fire2, flat-bottomed cone, seamless spherical-bottomed cone, seamed spherical-bottomed cone}, and the subsets of the identification frame are mutually exclusive.

[0098] (2) Assume that the recognition probability of each target in the recognition framework satisfies the probability condition, and then solve for the probability of multiple targets to be recognized to obtain the basic probability distribution of multiple targets. For any proposition A in the recognition framework, the probability m(A) of each proposition occurring satisfies the following condition:

[0099]

[0100] Where m(A) represents the recognition result of the target to be identified; m(Φ) represents the recognition rate of impossible propositions; A represents multiple targets to be identified; Θ represents the empty set, i.e., impossible propositions.

[0101] (3) Use combination rules to fuse the basic probability distributions and determine the fused data.

[0102] After obtaining the basic probability distribution of each type of target, they are fused using combination rules. For multi-evidence fusion, this is generally done using... Let represent the combination operation; the overall fusion recognition result is:

[0103]

[0104] Where m(A) represents the recognition result of the target to be identified; m(Φ) represents the recognition rate of impossible propositions; A represents multiple targets to be identified; Θ represents the empty set, i.e., impossible propositions; A i Proposition A i , is a subset of the recognition framework; Bj Proposition B j , is a subset of the identification framework; m1 represents the basic probability assignment function 1 of evidence body 1; m2 represents the basic probability assignment function 2 of evidence body 2; k represents the conflict coefficient.

[0105] Regarding the two pieces of evidence A i B j The fusion rules are as follows:

[0106]

[0107] The fusion of multiple evidence bodies involves transforming the above formula to obtain a new evidence body after fusing two existing ones. This new evidence body is then fused with other evidence bodies.

[0108]

[0109]

[0110] in, denoted by , where is the normalization factor; n represents the number of events; and k represents the conflict coefficient, which reflects the nature of the evidence A. i B j The degree of conflict between the evidence is determined by the conflict coefficient, which ranges from [0,1]. Furthermore, the DS combination rule satisfies the associative and commutative laws of mathematics, meaning the result of multi-evidence combination operations is independent of the order of calculation. This formula allows for the fusion of identification data from multiple radars.

[0111] In a specific embodiment of the present invention, the method further includes providing a training set and a test set for training the SVM algorithm.

[0112] In step S102, the motion trajectories of multiple targets to be identified are obtained, and the motion trajectory of each target is cropped. 2000 data points are randomly cropped from the motion trajectory (sampling rate 200Hz, so the time length is 2000 / 200 = 10s), resulting in 500 samples. During training, there are a total of 500 samples for each target class, with 2000 points per sample, for a total of 2500 samples across the five target classes. 200 samples from each target class are randomly selected as training data, and the remaining 300 samples are used as test data.

[0113] Configure the SVM parameters. Typically, the extracted data is not linearly separable in low-dimensional space, so a kernel function is needed to map the data to a higher dimension to make it linearly separable. Here, a Gaussian kernel is chosen, with a kernel parameter of 0.001 and a penalty coefficient of 1000. The acquired training set is then fed into the configured SVM classifier for training, and finally, the test data is used for recognition.

[0114] This invention combines data fusion and target recognition, and compared with single-radar target recognition, the multi-radar collaborative target recognition rate is greatly increased.

[0115] This invention first obtains the static RCS data of the target using electromagnetic simulation software, then combines it with the target's motion trajectory and micro-motion features to obtain the dynamic RCS sequence to be identified. Then, it obtains the identification results under multiple radars using the support vector machine method, and combines this identification result with DS evidence theory to greatly improve the target identification rate.

[0116] In a specific embodiment provided by this invention, Table 1 shows the identification results of a single radar and the fusion identification results of multiple radars. The table shows that a total of five features were selected as the parameters to be identified: mean, variance, median, mode, and range. The table also shows the identification results of a single radar and the fusion identification results of three radars. It can be seen that the fusion identification results of multiple radars are significantly better than the single radar identification results. When the statistical feature parameter is the mean, the average identification rate of the three radars is 69.6%, while the fusion identification result is 82.3%, improving the average identification rate by 12.7%. When the statistical feature parameter is the variance, the three radars... The average recognition rate was 78.1%, while the fusion recognition result was 99.3%, an improvement of 21.2%; when the median was extracted as the statistical feature parameter, the average recognition rate of the three radars was 71.2%, while the fusion recognition result was 86.3%, an improvement of 15.1%; when the mode was extracted as the statistical feature parameter, the average recognition rate of the three radars was 76.8%, while the fusion recognition result was 96.3%, an improvement of 19.5%; when the range was extracted as the statistical feature parameter, the average recognition rate of the three radars was 75.3%, while the fusion recognition result was 94.3%, an improvement of 19%.

[0117] Figure 12 This is a bar chart showing the results of single-radar identification and multi-radar fusion identification. The horizontal axis represents the statistical feature parameters extracted and processed through the target's dynamic RCS sequence, namely mean, variance, median, mode, and range; the vertical axis represents the recognition rate. Figure 12 Visualizing the data in Table 1 provides a more intuitive demonstration of the advantages of the fusion recognition algorithm.

[0118] Table 1

[0119]

[0120]

[0121] like Figure 13The flowchart shown is a specific embodiment of the present invention. When using the method provided by the present invention, static RCS data of multiple targets to be identified are first acquired from multiple radars, and then multiple motion trajectories are obtained using motion equations. Then, the micro-motion features at each moment are calculated based on the motion trajectories, and the micro-motion features are added to the static RCS data to obtain dynamic RCS data.

[0122] Noise is added to the dynamic RCS data to make it closer to the real data, resulting in multiple final RCS data. Then, features are extracted from the multiple final RCS data, and the extracted features are input into a trained SVM classifier to obtain the recognition result.

[0123] The recognition results are fused using fusion rules to obtain fused data.

[0124] The various embodiments described in this specification are presented in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. All or part of this invention can be used in numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, mobile communication terminals, multiprocessor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.

[0125] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the present invention.

Claims

1. A multi-radar cooperative target identification method, characterized in that, include: Acquire multiple static RCS data of multiple targets to be identified; The motion trajectory of the target to be identified is simulated using the equation of motion, and multiple motion trajectories of the target to be identified are obtained. The variation micro-motion features of each sampling point on the multiple motion trajectories are calculated respectively to obtain multiple variation micro-motion features of each motion trajectory; The multiple micro-motion features are added to the corresponding static RCS data to obtain multiple dynamic RCS data. Noise is added to the multiple dynamic RCS data to obtain multiple final RCS data; Statistical features are extracted from the multiple final RCS data, and the statistical features are input into the trained SVM classifier to obtain multiple recognition results; The multiple recognition results are fused using fusion rules to obtain fused data.

2. The method according to claim 1, characterized in that, Obtaining static RCS data of the target to be identified includes: establishing a target model and simulating the target model using electromagnetic simulation software to obtain the target's static RCS data.

3. The method according to claim 1, characterized in that, The equation of motion is specifically expressed as follows: in, Represents the geocentric distance vector; Represents the Earth's gravitational constant. ; The magnitude of the geocentric distance vector; Indicates the Earth's radius; This represents the coordinates of the target to be identified.

4. The method according to claim 1, characterized in that, The step of simulating the motion trajectory of the target to be identified using motion equations to obtain the motion trajectory of the target to be identified includes: The initial velocity, azimuth angle, and tilt angle of the target to be identified are obtained, and the current velocity component is obtained according to the motion equation. Based on the current velocity component and the initial position of the target to be identified, the position of the target to be identified at the next moment is obtained; The acceleration at the current moment is obtained using the two-body motion equation, and the current velocity component is corrected based on the current acceleration. This correction is then used to determine the position of the target to be identified at the next moment, thus obtaining the trajectory of the target to be identified.

5. The method according to claim 4, characterized in that, The two-body motion equations are specifically expressed as follows: in, Represents the geocentric distance vector; Represents the Earth's gravitational constant. ; The magnitude of the geocentric distance vector; This represents the second derivative with respect to the geocentric distance vector.

6. The method according to claim 1, characterized in that, The aforementioned micro-motion characteristics are specifically represented as follows: in, Indicates the swing angle. This represents the angular frequency of the oscillation.

7. The method according to claim 1, characterized in that, The process of fusing the multiple recognition results using fusion rules to obtain fused data specifically includes: Obtain an identification framework including the plurality of targets to be identified, and determine a plurality of pairwise exclusive subsets of the identification framework; The recognition probability of each target to be identified in the recognition framework is made to satisfy the probability condition, and then the basic probability distribution of the multiple targets to be identified is obtained by solving the problem. The basic probability distributions are fused using combination rules to determine the fused data.

8. The method according to claim 7, characterized in that, The probability condition is specifically expressed as follows: in, This refers to the recognition result of the target to be identified; The recognition rate of propositions that are impossible to occur; This represents multiple targets to be identified; It represents the empty set, i.e., a proposition that cannot occur.

9. The method according to claim 7, characterized in that, The combination rule is specifically expressed as follows: in, This refers to the recognition result of the target to be identified; The recognition rate of propositions that are impossible to occur; This represents multiple targets to be identified; This represents the empty set, i.e., propositions that cannot occur. Expressing a proposition , which is a subset of the recognition framework; Expressing a proposition , which is a subset of the recognition framework; The basic probability assignment function 1 represents evidence body 1; The basic probability assignment function 2 represents evidence body 2; This represents the conflict coefficient.

Images