Track initiation method based on intuitive method and transformer

Abstract

The invention discloses a track initiation method based on an intuitive method and transformer, which belongs to the technical field of radar data processing, and comprises the steps of designing a target motion complex scene, constructing a target motion model, simulating radar detection data information of target motion, and generating measurement data about a target position received by a radar. Calculating speed, acceleration and angle information of a target through measurement data of multiple times, and preliminarily screening possible real track data of the target through a visual method; feature extraction is performed on the preliminarily screened track data, the feature track data is labeled to obtain a track feature data set with a label, and a neural network is trained by using the track feature data set with the label to complete discrimination of true and false target tracks, so that the accuracy of true target track initiation in a complex clutter environment is improved, and the accuracy of target track initiation is improved. And the number of false tracks is greatly reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of radar data processing, and in particular relates to a track initiation method based on an intuitive method and a transformer. Background Art

[0002] With the rapid popularization of low-cost, miniaturized, and highly maneuverable flying equipment such as drones, the airspace environment has become increasingly complex. How to detect and track aerial targets in a complex airspace electromagnetic environment has become a hot issue and is of great significance to ensuring the safety of our airspace.

[0003] Track initiation is one of the key technologies in radar systems. It involves the initial stage of detecting and tracking targets from radar data. According to certain rules, radar measurement data of multiple adjacent time periods are correlated and integrated to obtain a set of time series data, and it is judged whether the sequence matches the real target. An excellent track initiation algorithm should ensure the track of the real target is started, and should effectively suppress the start of false tracks to reduce the impact of false targets. It is the basis for subsequent track management, stable target tracking and identification.

[0004] In recent years, deep learning has made significant breakthroughs in the application of video, image, speech, natural language processing, and other fields, and has also made significant progress in the application of radar data processing. Among them, transformers have powerful parallel processing capabilities and long-range dependency capture capabilities. When dealing with the problem of track initiation, they can process the entire sequence in parallel and consider all measurement data at the same time, improving processing speed and efficiency; and they can capture long-range dependencies in the sequence and identify potential tracks from scattered measurement data.

[0005] In summary, by making full use of radar measurement data information, constructing a reasonable and effective track initiation data set, and training a neural network model with strong robustness, the real target track initiation in complex scenarios can be completed. Summary of the invention

[0006] The purpose of the present invention is to provide a track initiation method based on intuitive method and transformer, so as to solve the problem that the accuracy and real-time performance of track initiation are difficult to balance under the current complex background.

[0007] To achieve the above object, the present invention adopts the following technical solution:

[0008] On the one hand, the present specification provides a track initiation method based on an intuitive method and a transformer, comprising:

[0009] Step 102, obtaining radar measurement data of the target;

[0010] Step 104, screening the radar measurement data based on the target motion restriction condition to obtain track data that meets the initial condition;

[0011] Step 106, inputting the track data that meets the initial conditions into the trained neural network model to distinguish true from false tracks and obtain a track distinction result.

[0012] On the other hand, the present specification provides a track initiation device based on an intuitive method and a transformer, comprising:

[0013] A measurement data acquisition module, used to acquire radar measurement data of the target;

[0014] The preliminary track initiation module is used to filter the radar measurement data based on the target motion restriction conditions to obtain the track data that meets the starting conditions;

[0015] The true and false track discrimination module is used to input the track data that meets the initial conditions into the trained neural network model to perform true and false track discrimination and obtain the track discrimination result.

[0016] Based on the above technical solution, this specification can achieve the following technical effects:

[0017] This method designs a complex target motion scene, constructs a target motion model, simulates radar detection data information of target motion, and generates measurement data about the target position received by the radar. The target's velocity, acceleration and angle information are calculated through measurement data at multiple times, and the possible real target track data is preliminarily screened out through intuitive methods; the preliminarily screened track data is feature extracted, and the feature track data is labeled to obtain a track feature data set with labels. The neural network is trained using the track feature data set with labels to complete the discrimination of true and false target tracks, thereby improving the accuracy of the start of the real target track in a complex clutter environment and significantly reducing the number of false tracks. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 The figure is a flow chart of a track initiation method based on an intuitive method and a transformer in one embodiment of the present invention.

[0019] Figure 2 The figure is a flow chart of a track initiation method based on the intuitive method and transformer in one embodiment of the present invention.

[0020] Figure 3 Schematic diagram of the structure of a neural network model in one embodiment of the present invention.

[0021] Figure 4The figure is a schematic diagram of the structure of a track initiation device based on the intuitive method and transformer in one embodiment of the present invention.

[0022] Figure 5 The figure is a schematic diagram of an electronic device according to the present invention. DETAILED DESCRIPTION

[0023] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description and claims. It should be noted that the drawings are all in very simplified form and are not in precise proportions, and are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention.

[0024] It should be noted that, in order to clearly explain the content of the present invention, the present invention specifically cites multiple embodiments to further illustrate different implementations of the present invention, wherein the multiple embodiments are enumerated rather than exhaustive. In addition, for the sake of brevity of explanation, the contents mentioned in the previous embodiments are often omitted in the subsequent embodiments. Therefore, the contents not mentioned in the subsequent embodiments can refer to the previous embodiments accordingly.

[0025] Example 1

[0026] Please refer to Figure 1 , Figure 1 The figure shows a track initiation method based on the intuitive method and transformer provided in this embodiment. In this embodiment, the method includes:

[0027] Step 102, obtaining radar measurement data of the target;

[0028] Step 104, screening the radar measurement data based on the target motion restriction condition to obtain track data that meets the initial condition;

[0029] In this embodiment, the target motion restriction condition includes a target speed restriction condition, a target acceleration restriction condition and a target yaw angle restriction condition; the target speed restriction condition is that the measured speed in the radar measurement data should be greater than or equal to a preset minimum speed and less than or equal to a preset maximum speed; the target acceleration restriction condition is that the measured acceleration in the radar measurement data should be less than or equal to a preset maximum acceleration; the target yaw angle restriction condition is that the measured yaw angle in the radar measurement data should be less than or equal to a preset maximum yaw angle.

[0030] Step 106, inputting the track data that meets the initial conditions into the trained neural network model to distinguish true from false tracks and obtain a track distinction result.

[0031] In this embodiment, obtaining the trained neural network model includes:

[0032] Step 202, constructing a target motion model in a preset radar measurement area to perform motion simulation and obtain target measurement data;

[0033] In this embodiment, the target measurement data includes: target real track measurement data and target false track measurement data;

[0034] In this embodiment, one implementation of step 202 is:

[0035] Step 302, constructing a target motion model in a preset radar measurement area and performing several radar scans based on the target motion parameters to obtain the target's real track measurement data;

[0036] Step 304, generating corresponding clutter measurement data based on the clutter data of each radar scanning cycle within a preset radar measurement area;

[0037] In this embodiment, before step 304, the following steps are also included:

[0038] Step 303: Determine clutter data for each radar scanning cycle based on Poisson distribution and clutter density parameters.

[0039] Step 306 , filtering the clutter measurement data based on the target motion restriction condition to obtain the target false track measurement data.

[0040] Step 204, extracting features and normalizing the target measurement data to obtain track feature data;

[0041] In this embodiment, before step 204, the following steps are also included:

[0042] Labels are added to the target real track measurement data and the target false track measurement data respectively, wherein the label of the target real track measurement data is set to 1, and the label of the target false track measurement data is set to 0.

[0043] In this embodiment, one implementation of step 204 is:

[0044] Step 402, respectively extracting speed features, acceleration features and angle features from the target real track measurement data and the target false track measurement data, and obtaining speed feature values, acceleration feature values ​​and angle feature values ​​of each measurement data;

[0045] Step 404, concatenating the velocity eigenvalue, the acceleration eigenvalue and the angle eigenvalue into a one-dimensional vector to obtain a one-dimensional eigenvector;

[0046] Step 406, normalize each eigenvalue in the one-dimensional eigenvector to obtain track characteristic data.

[0047] Step 206: Train the neural network model based on the track feature data to obtain a trained neural network model.

[0048] In this embodiment, one implementation of step 206 is:

[0049] Step 502, inputting the track feature data into the neural network model to calculate and obtain the track prediction result;

[0050] Step 504, using a cross entropy loss function to calculate the difference between the track prediction result and the actual result, and obtaining a loss function calculation result;

[0051] Step 506, based on the loss function calculation result, calculate the gradient of the neural network model parameters;

[0052] Step 508: Based on the gradient, an optimizer is used to update the parameters of the neural network model to obtain a trained neural network model.

[0053] In this embodiment, reference Figure 2 , a track initiation method based on intuitive method and transformer, comprising the following steps:

[0054] Step 1: Build a target motion model and simulate the radar measurement data containing the target position information. Calculate the target's speed, acceleration, angle and other data information through the measurement data at multiple times.

[0055] Step 2: By setting multiple constraints such as target speed, acceleration, heading angle, etc., use the intuitive method to start the preliminary track and preliminarily screen out the possible target real track data.

[0056] Step 3: Select the target track position points at four consecutive moments for feature extraction. The extracted features are the target's speed information, acceleration information, and heading angle information.

[0057] Step 4: Label the suspected real track data that has been initiated and build a data set including real tracks and false tracks.

[0058] Step 5: Use the labeled track feature dataset to train the neural network to determine whether the track is true or false.

[0059] Specifically, a radar measurement area of ​​50 km × 50 km is set, and a target motion model that performs uniformly accelerated linear motion is constructed within the radar measurement area. The initial position and motion direction are random, and the target motion speed range is limited to [50 m / s, 500 m / s], and the acceleration range is [0, 50 m / s 2 ], the ranging error and angle measurement error are 30m and 0.1° respectively, the radar scanning period is 5s, and a total of 4 scans are performed.

[0060] Specifically, based on the target motion speed range, acceleration value range, ranging error and angle measurement error, 5000 target real track measurement combinations are generated through simulation.

[0061] Specifically, the number of clutter in each radar scanning cycle is determined according to the Poisson distribution, which can be expressed as

[0062]

[0063] Where λ represents the clutter density parameter, γ is a random number that obeys a uniform distribution in [0,1], and J is the number of clutter measurements obtained.

[0064] Specifically, in the radar measurement area, in the generated clutter measurement data, according to the limitation of target speed, acceleration, angle and other information, the false target track combination that meets the initial conditions is screened out. In this embodiment, a total of 5000 false track measurement combinations are simulated and generated.

[0065] Among them, the restriction condition on the target speed is that the measured speed of the target should be within the appropriate value range, which can be expressed as v min and v max Respectively represent the minimum and maximum speeds allowed.

[0066] Among them, the restriction condition on the target acceleration is that the measured acceleration of the target should be in an appropriate value range, which can be expressed as a max Indicates the maximum acceleration allowed.

[0067] The restriction condition on the target yaw angle is that the measured yaw angle of the target should be within a suitable value range, which can be expressed as in is the maximum heading deviation angle set. Usually, to ensure a high probability start of the track, The value of is π.

[0068] Specifically, labels are added to the real track and the false track, the label of the real track data is set to 1, and the label of the false track data is set to 0. Specifically, feature extraction is performed on the false track measurement combination and the target real track measurement combination respectively.

[0069] The features extracted from each track data include the speed (vx t1 ,vx t2 ,vx t3 ,vy t1 ,vy t2 ,vy t3), the acceleration of the points in three consecutive scanning cycles (ax t1 ,ax t2 ,ay t1 ,ay t2 ), and the angle between the lines connecting the points of three consecutive scanning cycles There are 12 eigenvalues ​​in total. Specifically, the 12 eigenvalues ​​of the four time periods are concatenated into a one-dimensional vector, and each eigenvalue is normalized, that is, Among them, X(k) represents the kth eigenvalue in the track data, X max and X min are the maximum and minimum values ​​of the eigenvalue range, respectively. Represents the result after normalization of the eigenvalue.

[0070] The normalized track feature data is used as the input of the neural network. The dimension of the input layer of the neural network is 12×1, and the dimension of the output layer is 2×1 to represent the real track and the false track.

[0071] Specifically, 80% of the data in the dataset is randomly selected as a training set and 20% is randomly selected as a test set for training the neural network model.

[0072] Specifically, a neural network model is constructed, and the detailed structure is as follows Figure 3 , mainly including input layer, embedding layer, position encoding layer, encoding layer, fully connected layer and output layer. The number of attention heads in the encoding layer is set to 3, the number of encoder layers is set to 1, the constructed training set is input into the neural network, the cross entropy loss function and adam optimizer are selected, the number of training times is set to 50, the number of training batch samples is set to 64, the learning rate is set to 0.03, and the neural network model is trained.

[0073] The training process can be divided into four steps: (1) Forward propagation: pass the track feature value data through the neural network model to calculate the output result of track prediction; (2) Calculate the loss: use the cross entropy loss function to calculate the difference between the predicted output and the true label; (3) Backward propagation: calculate the gradient of the model parameters according to the loss function; (4) Parameter update: use the optimizer to update the model parameters according to the gradient value. Specifically, the optimal neural network model is obtained through training, and the neural network model is used to verify the real track recognition accuracy of the test set data.

[0074] At this point, a neural network model for identifying the real and false tracks of a target can be obtained.

[0075] By using the intuitive method to preliminarily screen the tracks that meet the initial conditions, and then using the trained neural network model to complete the discrimination of true and false tracks, the number of false tracks can be greatly reduced and the recognition probability of true tracks can be increased.

[0076] The beneficial effects of the present invention are:

[0077] (1) Capable of initiating target tracks in complex background environments;

[0078] (2) Increase the probability of the track starting of the real target and reduce the probability of the track starting of the false target;

[0079] (3) Improve the efficiency of track initiation while ensuring the quality of track initiation.

[0080] In summary, this method designs a complex target motion scene, constructs a target motion model, simulates the radar detection data information of the target motion, and generates the measurement data about the target position received by the radar. The target's velocity, acceleration and angle information are calculated through the measurement data of multiple times, and the possible real target track data is preliminarily screened out through the intuitive method; the preliminarily screened track data is feature extracted, and the feature track data is labeled to obtain a track feature data set with labels. The neural network is trained using the track feature data set with labels to complete the discrimination of true and false target tracks, thereby improving the accuracy of the start of the real target track in a complex clutter environment and greatly reducing the number of false tracks.

[0081] Example 2

[0082] Please refer to Figure 4 , Figure 4 The figure shows a track initiation device based on the intuitive method and transformer provided in this embodiment. In this embodiment, the device includes:

[0083] A measurement data acquisition module, used to acquire radar measurement data of the target;

[0084] The preliminary track initiation module is used to filter the radar measurement data based on the target motion restriction conditions to obtain the track data that meets the starting conditions;

[0085] The true and false track discrimination module is used to input the track data that meets the initial conditions into the trained neural network model to perform true and false track discrimination and obtain the track discrimination result.

[0086] Optionally, the target motion constraint conditions include target speed constraint conditions, target acceleration constraint conditions and target yaw angle constraint conditions; the target speed constraint condition is that the measured speed in the radar measurement data should be greater than or equal to a preset minimum speed and less than or equal to a preset maximum speed; the target acceleration constraint condition is that the measured acceleration in the radar measurement data should be less than or equal to a preset maximum acceleration; the target yaw angle constraint condition is that the measured yaw angle in the radar measurement data should be less than or equal to a preset maximum yaw angle.

[0087] Optionally, also include:

[0088] The motion simulation module is used to construct a target motion model in a preset radar measurement area to perform motion simulation and obtain target measurement data;

[0089] The data processing module is used to extract features and normalize the target measurement data to obtain track feature data;

[0090] The model training module is used to train the neural network model based on the track feature data to obtain the trained neural network model.

[0091] Optionally, the target measurement data includes: target real track measurement data and target false track measurement data;

[0092] Optional motion simulation modules include:

[0093] A radar scanning unit is used to construct a target motion model in a preset radar measurement area and perform several radar scans based on the target motion parameters to obtain the target's real track measurement data;

[0094] A clutter measurement data generating unit, configured to generate corresponding clutter measurement data based on clutter data of each radar scanning cycle within a preset radar measurement area;

[0095] The false track acquisition unit is used to filter the clutter measurement data based on the target motion restriction condition to obtain the target false track measurement data.

[0096] Optionally, the motion simulation module further includes:

[0097] The clutter data determination unit is used to determine the clutter data of each radar scanning cycle based on Poisson distribution and clutter density parameters.

[0098] Optionally, the data processing module includes:

[0099] A feature extraction unit is used to extract velocity features, acceleration features and angle features from the target real track measurement data and the target false track measurement data, respectively, to obtain velocity feature values, acceleration feature values ​​and angle feature values ​​of each measurement data;

[0100] An eigenvalue concatenation unit is used to concatenate the velocity eigenvalue, the acceleration eigenvalue and the angle eigenvalue into a one-dimensional vector to obtain a one-dimensional eigenvector;

[0101] The normalization unit is used to normalize each eigenvalue in the one-dimensional eigenvector to obtain track characteristic data.

[0102] Optionally, the velocity characteristic value is the velocity value between target measurement data points of two adjacent radar scanning cycles; the acceleration characteristic value is the acceleration value between target measurement data points of three consecutive radar scanning cycles; and the angle characteristic value is the angle between the lines connecting the target measurement data points of three consecutive radar scanning cycles.

[0103] In this embodiment, it also includes:

[0104] The label adding module is used to add labels to the target real track measurement data and the target false track measurement data respectively, wherein the label of the target real track measurement data is set to 1, and the label of the target false track measurement data is set to 0.

[0105] Optionally, the model training module includes:

[0106] A track prediction unit, used to input track feature data into a neural network model and calculate and obtain a track prediction result;

[0107] A loss function calculation unit, used to calculate the difference between the track prediction result and the actual result using the cross entropy loss function to obtain the loss function calculation result;

[0108] A parameter gradient calculation unit, used to calculate the gradient of the neural network model parameters based on the loss function calculation result;

[0109] The parameter updating unit is used to update the parameters of the neural network model based on the gradient using an optimizer to obtain a trained neural network model.

[0110] Based on this, this device designs a complex scene of target motion, constructs a target motion model, simulates radar detection data information of target motion, and generates measurement data about the target position received by the radar. The target's speed, acceleration and angle information are calculated through measurement data at multiple times, and the possible real track data of the target is preliminarily screened out through intuitive methods; the features of the preliminarily screened track data are extracted, and the feature track data are labeled to obtain a track feature data set with labels. The neural network is trained using the track feature data set with labels to complete the discrimination of true and false target tracks, thereby improving the accuracy of the start of the real target track in a complex clutter environment and greatly reducing the number of false tracks.

[0111] Example 3

[0112] Please refer to Figure 5, this embodiment provides an electronic device, which includes a processor, an internal bus, a network interface, a memory and a non-volatile memory, and of course may also include hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a track initiation method based on the intuitive method and transformer at the logical level. Of course, in addition to software implementation methods, this specification does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0113] The network interface, processor and memory can be connected to each other through a bus system. The above bus can be divided into an address bus, a data bus, a control bus, etc.

[0114] The memory is used to store programs. Specifically, the program may include program code, and the program code includes computer operation instructions. The memory may include a read-only memory and a random access memory, and provides instructions and data to the processor.

[0115] The processor is used to execute the program stored in the above memory, and specifically perform:

[0116] Step 102, obtaining radar measurement data of the target;

[0117] Step 104, screening the radar measurement data based on the target motion restriction condition to obtain track data that meets the initial condition;

[0118] Step 106, inputting the track data that meets the initial conditions into the trained neural network model to distinguish true from false tracks and obtain a track distinction result.

[0119] The processor may be an integrated circuit chip with signal processing capability. In the implementation process, each step of the above method may be completed by an integrated logic circuit of the processor hardware or by instructions in the form of software.

[0120] Based on the same invention, the embodiment of this specification also provides a computer-readable storage medium, which stores one or more programs. When the one or more programs are executed by an electronic device including multiple application programs, the electronic device executes Figure 1-Figure 3 The corresponding embodiment provides a track initiation method based on the intuitive method and transformer.

[0121] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-readable storage media containing computer-usable program code.

[0122] In addition, for the above-mentioned specific implementation of the system, since it is basically similar to the method implementation, the description is relatively simple, and the relevant parts can be referred to the partial description of the method implementation. Moreover, it should be noted that in each module of the system of the present application, the components therein are logically divided according to the functions to be implemented, but the present application is not limited thereto, and the components can be re-divided or combined as needed.

[0123] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other. Each embodiment focuses on the differences between it and other embodiments.

[0124] The above is a description of a specific embodiment of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the specific order or sequential order shown in the process depicted in the drawings is not necessarily required to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0125] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A track initiation method based on intuitive method and transformer, characterized in that: include: Obtain radar measurement data of the target; The radar measurement data is screened based on the target motion restriction conditions to obtain the track data that meets the starting conditions; The track data that meets the initial conditions is input into the trained neural network model to distinguish true and false tracks and obtain the track discrimination result.

2. The method according to claim 1, characterized in that The target motion restriction condition includes a target speed restriction condition, a target acceleration restriction condition and a target yaw angle restriction condition; the target speed restriction condition is that the measured speed in the radar measurement data should be greater than or equal to a preset minimum speed and less than or equal to a preset maximum speed; the target acceleration restriction condition is that the measured acceleration in the radar measurement data should be less than or equal to a preset maximum acceleration; The target yaw angle restriction condition is that the measured yaw angle in the radar measurement data should be less than or equal to a preset maximum yaw angle.

3. The method according to claim 1, characterized in that Obtaining the trained neural network model includes: Construct a target motion model in a preset radar measurement area to perform motion simulation and obtain target measurement data; Perform feature extraction and normalization on the target measurement data to obtain track feature data; The neural network model is trained based on the track feature data to obtain a trained neural network model.

4. The method according to claim 3, characterized in that The target measurement data includes: target real track measurement data and target false track measurement data; The step of constructing a target motion model in a preset radar measurement area to perform motion simulation and obtain target measurement data includes: Construct a target motion model in a preset radar measurement area and perform several radar scans based on the target motion parameters to obtain the target's real track measurement data; In a preset radar measurement area, based on the clutter data of each radar scanning cycle, corresponding clutter measurement data is generated; The clutter measurement data is screened based on the target motion restriction conditions to obtain the target false track measurement data.

5. The method according to claim 4, characterized in that Before generating corresponding clutter measurement data in the preset radar measurement area based on the clutter data of each radar scanning cycle, the method further includes: determining the clutter data of each radar scanning cycle based on Poisson distribution and clutter density parameters.

6. The method according to claim 5, characterized in that The feature extraction and normalization processing of the target measurement data to obtain the track feature data includes: The velocity feature extraction, acceleration feature extraction and angle feature extraction are performed on the target real track measurement data and the target false track measurement data respectively, and the velocity feature value, acceleration feature value and angle feature value of each measurement data are obtained; The velocity eigenvalue, the acceleration eigenvalue and the angle eigenvalue are concatenated into a one-dimensional vector to obtain a one-dimensional eigenvector; Each eigenvalue in the one-dimensional eigenvector is normalized to obtain the track characteristic data.

7. The method according to claim 6, characterized in that The velocity characteristic value is the velocity value between target measurement data points of two adjacent radar scanning cycles; the acceleration characteristic value is the acceleration value between target measurement data points of three consecutive radar scanning cycles; and the angle characteristic value is the angle between the lines connecting the target measurement data points of three consecutive radar scanning cycles.

8. The method according to claim 7, characterized in that Before the target measurement data is subjected to feature extraction and normalization processing to obtain track feature data, the method further includes: adding labels to the target real track measurement data and the target false track measurement data, respectively, wherein the label of the target real track measurement data is set to 1, and the label of the target false track measurement data is set to 0.

9. The method according to claim 8, characterized in that The training of the neural network model based on the track feature data to obtain the trained neural network model includes: Input the track characteristic data into the neural network model and calculate the track prediction result; Use the cross entropy loss function to calculate the difference between the track prediction result and the actual result to obtain the loss function calculation result; Based on the loss function calculation results, the gradient of the neural network model parameters is calculated; Based on the gradient, the optimizer is used to update the parameters of the neural network model to obtain the trained neural network model.

10. A track initiation device based on intuitive method and transformer, characterized in that: include: A measurement data acquisition module, used to acquire radar measurement data of the target; The preliminary track initiation module is used to filter the radar measurement data based on the target motion restriction conditions to obtain the track data that meets the starting conditions; The true and false track discrimination module is used to input the track data that meets the initial conditions into the trained neural network model to perform true and false track discrimination and obtain the track discrimination result.

Images