A ballistic target recognition method and system based on long short-term memory neural network

Abstract

The present invention provides a ballistic target recognition method and system based on long short-term memory neural network, the method includes: obtaining radar sample data to be classified, recording echo data of targets to be classified with shell attributes, performing amplitude normalization, using the RCS value under the reference distance and pulse width to perform RCS calibration on the current echo data, obtaining pre-processed data, dividing it into a training set and a test set, and using the training set to train the LSTM network; pre-processing the subsequent newly obtained radar sample data to be classified, and inputting the subsequent pre-processed data as test data into the trained LSTM network for prediction, and obtaining the predicted value of shell attributes. The present invention reduces the complexity of algorithm implementation, is suitable for distinguishing rockets and grenades by artillery position detection radar, has high robustness, accurate recognition, helps our radar to grasp the enemy's military situation in a timely manner, provides convenience for confirming the enemy's firepower configuration, and the correct recognition rate of rockets and grenades reaches more than 90%.

Description

Technical Field

[0001] The present invention relates to the technical field of ballistic target recognition, and in particular to a ballistic target recognition method and system based on a long short-term memory neural network. Background Art

[0002] The artillery position detection radar confirms the enemy's firepower configuration by determining the coordinates of the enemy's launching position and identifying the type and attributes of the enemy's artillery targets.

[0003] At present, artillery shell type recognition mainly includes the distinction between mortars, grenades and rockets. Since mortars, grenades and rockets have great differences in ballistics, mortars can be directly distinguished by ballistic characteristics. However, since grenades and rockets have high similarities in ballistics, distinguishing between grenades and rockets is a major problem in target recognition for artillery position detection radar.

[0004] It is difficult to directly distinguish between grenades and rockets based on the echo amplitude. Existing identification methods usually distinguish between grenades and rockets through RCS (Radar cross-section) or target Doppler characteristics.

[0005] However, in actual application scenarios, due to problems such as meteorological clutter interference and the unstable distance between the radar and the ballistic target, the recognition performance of the gun position detection radar is seriously affected by the environment, making it difficult to achieve high-performance classification of grenades and rockets. Summary of the invention

[0006] In view of this, the purpose of the present invention is to develop a new projectile type recognition method for artillery position detection radar, and propose a ballistic target recognition method and system based on long short-term memory neural network. A rocket and grenade distinction model is constructed through a long short-term memory neural network LSTM. First, the echo data is preprocessed by RCS calibration, and then the LSTM network is trained using the training set. The trained LSTM network is used to predict the data at subsequent moments, reducing the complexity of the algorithm implementation, improving the recognition accuracy, and realizing high-performance projectile type classification, so as to help our radar grasp the enemy's military situation in a timely manner and provide convenience for confirming the enemy's firepower configuration.

[0007] The present invention provides a ballistic target recognition method based on a long short-term memory neural network, comprising the following steps:

[0008] S1. Obtain radar sample data to be classified, record the echo data of the target to be classified with projectile attributes, normalize the amplitude of the echo data, and use the radar cross-sectional area RCS value under the reference distance and pulse width to perform RCS calibration on the current echo data to obtain the pre-processed data sample;

[0009] Because the distance between the radar and the target is uncertain when the artillery position detection radar records rocket and grenade data, and the detection methods are different for different targets, it is necessary to calibrate the target RCS radar scattering cross-section according to the current working parameters (by distance and pulse width).

[0010] S2, dividing the preprocessed data samples into a training set and a test set, using the training set to train a long short-term memory neural network LSTM to obtain a trained LSTM network;

[0011] Preferably, the training set accounts for 80% of the total amount of preprocessed data samples, and the test set accounts for 20% of the total amount of preprocessed data samples.

[0012] LSTM solves the problem that RNN cannot learn too long sequences. The state update of LSTM units is controlled by input gate, forget gate and output gate. To build an LSTM network, you need to set the input dimension, time step and number of neurons. You can input and set it according to the working mode and data characteristics of the artillery position detection radar.

[0013] For example, the input dimension of the LSTM network is set to N k , the time step is P ; The number of hidden layers is 2, and the number of hidden nodes in each layer is set to 512 (the number of hidden nodes is about twice the number of input nodes); the pre-processed data after RCS calibration is used as the input of the LSTM network, and each input P Group pulse data;

[0014] S3, preprocess the newly obtained radar sample data to be classified according to step S1, use the preprocessed data obtained subsequently as test data, input the test data into the trained LSTM network for prediction, and obtain the predicted value of the shell attribute.

[0015] According to existing data, there is no effective method to distinguish rockets and grenades using the long short-term memory neural network LSTM network at this stage. The present invention constructs a rocket and grenade distinction model for the first time through the long short-term memory neural network LSTM, predicts the properties of artillery shells, and achieves a correct recognition rate of more than 90% for rockets and grenades.

[0016] Furthermore, the method for obtaining radar sample data to be classified in step S1 includes: subjecting the echo data to pulse compression, moving target detection MTD, constant false alarm detection CFAR, and target tracking to obtain point trace data of the projectile target, and using the original echo pulse compressed data corresponding to the point trace data as the radar sample data to be classified.

[0017] Specifically, after a series of signal processing such as pulse compression, moving target detection MTD, constant false alarm detection CFAR, target tracking, etc., the moving target can be successfully detected and the point trace data of the moving target can be obtained.

[0018] Furthermore, the method of using the original echo pulse compression data corresponding to the point trace data as radar sample data to be classified includes:

[0019] Assume that the point data of the projectile target is k Group, k The group trace data contains M k Group pulse, M k The number of pulses in each group is N k , intercept M k The first 8 seconds of data in the group pulse, extract the last P The IQ data of the group pulse (orthogonal amplitude components In-phase and Quadrature) is obtained K Group P × N k The dimensional IQ data, i.e. the original echo pulse compressed data, is used as the radar sample data to be classified;

[0020] No. k The IQ data of the group is expressed as:

[0021] (1)

[0022] In formula (1), Is P × N k indivual s ij Arranged P OK N k A matrix of columns; s ij Indicates i The first observation j The sampling value of a pulse on the target, s ij is a plural number.

[0023] Since the gun position detection radar mainly focuses on the ascending phase data, after extracting the first 8 seconds of data, the last P A group of pulses are extracted to obtain the IQ data of rocket and grenade pulse pressure. k Group track data M k × Nk IQ data, intercept all pulses within 8 seconds after the first pulse starts, and take the last pulse of this segment P The group pulse is used as the input of the LSTM network to obtain P × N k dimensional data, in practical applications, P Can be set according to the radar working mode. In TAS scanning mode, P It can be taken as 12.

[0024] Furthermore, the method for normalizing the amplitude of the echo data in step S1 includes:

[0025] Set up k Group IQ data includes p The first pulse in the group p The IQ data of the group pulse is , the target signal-to-noise ratio under the current observation is SNR p , then the normalized target amplitude is expressed as:

[0026] (2)

[0027] In formula (2), For the k The first p Detection point i The echo amplitude of a pulse; For the k The first p The average value of all pulse echo amplitudes at a detection point; Indicates converting the target signal-to-noise ratio under the current observation into a linear amplitude.

[0028] Normalizing the amplitude of echo data can improve the stability of signal processing, so that target echo data of different intensities and distances have the same scale when processing, thereby reducing processing errors caused by amplitude differences. When performing pulse compression, target detection and other processing, normalization can ensure the consistency and effectiveness of the algorithm and improve the accuracy and reliability of the processing results. Normalized data is easier to perform further data analysis and processing, such as feature extraction and classification, thereby improving the overall processing efficiency and effect.

[0029] Furthermore, the method of performing RCS calibration on the current echo data in step S1 includes:

[0030] Assume the pulse width of the current echo data is T p , the target distance is R p , the reference pulse width is set toT ref , the reference distance is set to R ref , the reference signal-to-noise ratio is SNR ref , the RCS under the reference distance and reference pulse width is RCS ref , calibrate the RCS according to the radar equation and get the calibrated data .

[0031] RCS calibration provides the reflection coefficient of the target, allowing the radar to simulate the target's echo signal, which helps evaluate the performance and detection capabilities of the radar system. In the signal processing process, RCS calibration helps the radar process the received echo signal more accurately and improve the signal-to-noise ratio and resolution. With the target characteristic data obtained by RCS calibration, the radar can more effectively identify the type, size and shape of the target, thereby improving the accuracy of target recognition.

[0032] Furthermore, the method for training the long short-term memory neural network LSTM in step S2 includes:

[0033] The training set is input into the LSTM network to obtain an output value, and a loss function value is calculated based on a comparison between the output value and the actual value. The loss function value is used to optimize the parameters of the LSTM network so that the loss function value is reduced to within a set value.

[0034] The loss function of the long short-term memory neural network LSTM is also called the generalized linear state loss function, which can help the model achieve better performance when solving different tasks. The application of the LSTM loss function can more effectively handle classification and regression problems, as well as serialization problems. In addition, the LSTM loss function can also better handle unstructured data and help the model better learn more complex data representations. The LSTM loss function is more durable, safer and more reliable, can avoid a large number of algorithmic errors, better avoid errors in learning, and can more effectively apply variables and parameters, which helps to improve the stability and reliability of the model. At the same time, it can help the model reduce training time and improve model training efficiency.

[0035] Furthermore, the method of inputting the test data into the trained LSTM network for prediction in step S3 includes:

[0036] The newly obtained radar sample data to be classified: M k × N k The IQ data is preprocessed to obtain P × N kdimensional test data, the P × N k The dimensional test data is input into the trained LSTM network to predict the shell attributes.

[0037] The invention predicts the shell properties of rockets and grenades by training a long short-term memory neural network (LSTM). Compared with the traditional target Doppler feature differentiation method, the recognition accuracy is greatly improved.

[0038] The present invention also provides a ballistic target recognition system based on a long short-term memory neural network, which executes the ballistic target recognition method based on a long short-term memory neural network as described above, including:

[0039] Obtain radar sample data module to be classified: used to record the echo data of the target to be classified with shell attributes, normalize the amplitude of the echo data, calibrate the RCS of the current echo data using the radar cross-sectional area RCS value under the reference distance and pulse width, and obtain the pre-processed data sample;

[0040] LSTM network training module: used for dividing the preprocessed data samples into a training set and a test set, and using the training set to train the long short-term memory neural network LSTM to obtain a trained LSTM network;

[0041] Shell attribute prediction module: It is used to preprocess the newly obtained radar sample data to be classified according to step S1, use the preprocessed data obtained later as test data, input the test data into the trained LSTM network for prediction, and obtain the predicted value of the shell attribute.

[0042] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, the steps of the ballistic target recognition method based on the long short-term memory neural network as described above are implemented.

[0043] The present invention also provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the ballistic target recognition method based on the long short-term memory neural network as described above are implemented.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] The ballistic target recognition method and system based on long short-term memory neural network provided by the present invention adopts long short-term memory neural network to construct a rocket and grenade differentiation model. First, the echo data is preprocessed by RCS calibration, and then the LSTM network is trained by using the training set. The trained LSTM network is used to predict the data at subsequent moments, so as to realize a rocket and grenade differentiation method suitable for artillery position detection and calibration radar, reduce the complexity of algorithm implementation, have the characteristics of high robustness and accurate recognition, effectively improve the projectile type recognition capability of artillery position detection and calibration radar, can realize high-performance projectile type classification, help our radar to grasp the enemy's military situation in time, provide convenience for confirming the enemy's firepower configuration, and the correct recognition rate of rockets and grenades reaches more than 90%. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiment.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.

[0047] In the attached picture:

[0048] Figure 1 This is a flow chart of a method for distinguishing between rockets and grenades based on LSTM according to an embodiment of the present invention;

[0049] Figure 2 A flowchart of obtaining radar sample data to be classified according to an embodiment of the present invention;

[0050] Figure 3 A flowchart of preprocessing echo data according to an embodiment of the present invention;

[0051] Figure 4 This is a flowchart of training an LSTM network according to an embodiment of the present invention;

[0052] Figure 5 This is a flowchart of using an LSTM network to perform predictions in an embodiment of the present invention;

[0053] Figure 6 It is a flow chart of a ballistic target recognition method based on long short-term memory neural network of the present invention;

[0054] Figure 7 The figure is a schematic diagram of the structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0055] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and products consistent with some aspects of the present disclosure as detailed in the appended claims.

[0056] The terms used in this disclosure are for the purpose of describing specific embodiments only and are not intended to limit the disclosure. The singular forms of "a", "said" and "the" used in this disclosure and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0057] It should be understood that although the terms first, second, third, etc. may be used in the present disclosure to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present disclosure, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0058] The embodiments of the present invention are further described in detail below with reference to the accompanying drawings.

[0059] The present invention provides a method for ballistic target recognition based on long short-term memory neural network. Figure 6 As shown, the following steps are included:

[0060] S1. Obtain radar sample data to be classified, record the echo data of the target to be classified with shell attributes, normalize the amplitude of the echo data, and use the radar cross-sectional area RCS value under the reference distance and pulse width to calibrate the current echo data to obtain the pre-processed data sample (such as Figure 3 shown);

[0061] Because the distance between the radar and the target is uncertain when the artillery position detection radar records rocket and grenade data, and the detection methods are different for different targets, it is necessary to calibrate the target RCS according to the current working parameters (by distance and pulse width).

[0062] The method for obtaining radar sample data to be classified includes: subjecting the echo data to pulse compression, moving target detection MTD, constant false alarm detection CFAR, and target tracking to obtain point trace data of the projectile target, and using the original echo pulse compressed data corresponding to the point trace data as the radar sample data to be classified.

[0063] Specifically, after a series of signal processing such as pulse compression, moving target detection MTD, constant false alarm detection CFAR, target tracking, etc., the moving target can be successfully detected and the point trace data of the moving target can be obtained.

[0064] The method of using the original echo pulse compressed data corresponding to the point trace data as radar sample data to be classified includes:

[0065] Assume that the point data of the projectile target is k Group, k The group trace data contains M k Group pulse, M k The number of pulses in each group is N k , intercept M k The first 8 seconds of data in the group pulse, extract the last P The IQ data of the group pulse (orthogonal amplitude components In-phase and Quadrature) is obtained K Group P × N k The IQ data of the dimension is the original echo pulse compression data, which is used as the radar sample data to be classified (such as Figure 2 shown);

[0066] No. k The IQ data of the group is expressed as:

[0067] (1)

[0068] In formula (1), Is P × N k indivual s ij Arranged P OK N k A matrix of columns; s ij Indicates i The first observation j The sampling value of a pulse on the target, s ij is a plural number.

[0069] Since the gun position detection radar mainly focuses on the ascending phase data, after extracting the first 8 seconds of data, the last P A group of pulses are extracted to obtain the IQ data of rocket and grenade pulse pressure. k Group track data M k × N k IQ data, intercept all pulses within 8 seconds after the first pulse starts, and take the last pulse of this segment P The group pulse is used as the input of the LSTM network to obtain P × N k dimensional data, in practical applications, P Can be set according to the radar working mode. In TAS scanning mode, P It can be taken as 12.

[0070] Methods for amplitude normalization of echo data include:

[0071] Set up k Group IQ data includes p The first pulse in the group p The IQ data of the group pulse is , the target signal-to-noise ratio under the current observation is SNR p , then the normalized target amplitude is expressed as:

[0072] (2)

[0073] In formula (2), For the k The first p Detection point i The echo amplitude of a pulse; For the k The first p The average value of all pulse echo amplitudes at a detection point; Indicates converting the target signal-to-noise ratio under the current observation into a linear amplitude.

[0074] The method of performing RCS calibration on the current echo data includes:

[0075] Assume the pulse width of the current echo data is T p , the target distance is R p , the reference pulse width is set to T ref , the reference distance is set to R ref , the reference signal-to-noise ratio is SNR ref, the RCS under the reference distance and reference pulse width is RCS ref , calibrate the RCS according to the radar equation and get the calibrated data .

[0076] S2, dividing the pre-processed data samples into a training set and a test set, using the training set to train a long short-term memory neural network LSTM to obtain a trained LSTM network (such as Figure 4 shown);

[0077] The training set accounts for 80% of the total data samples after preprocessing, and the test set accounts for 20% of the total data samples after preprocessing.

[0078] Set the input dimension of the LSTM network to N k , the time step is P ; The number of hidden layers is 2, and the number of hidden nodes in each layer is set to 512; the pre-processed data after RCS calibration is used as the input of the LSTM network, and each input P Group pulse data;

[0079] Methods for training long short-term memory neural networks LSTM include:

[0080] The training set is input into the LSTM network to obtain an output value, and a loss function value is calculated based on a comparison between the output value and the actual value. The loss function value is used to optimize the parameters of the LSTM network so that the loss function value is reduced to within a set value.

[0081] S3, preprocess the newly obtained radar sample data to be classified according to step S1, use the preprocessed data obtained later as test data, input the test data into the trained LSTM network for prediction, and obtain the predicted value of the shell attribute (such as Figure 5 as shown).

[0082] Methods for inputting test data into the trained LSTM network for prediction include:

[0083] The newly obtained radar sample data to be classified: M k × N k The IQ data is preprocessed to obtain P × N k dimensional test data, the P × N k The dimensional test data is input into the trained LSTM network to predict the shell attributes.

[0084] Figure 1 The basic process of the LSTM-based method for distinguishing between rockets and grenades in this embodiment is shown.

[0085] The embodiment of the present invention predicts the shell properties of rockets and grenades by training a long short-term memory neural network LSTM. Compared with the traditional distinction method using target Doppler characteristics, the recognition accuracy is greatly improved, and the correct recognition rate of rockets and grenades reaches more than 90%.

[0086] The embodiment of the present invention further provides a ballistic target recognition system based on a long short-term memory neural network, which executes the ballistic target recognition method based on a long short-term memory neural network as described above, including:

[0087] Obtain radar sample data module to be classified: used to record the echo data of the target to be classified with shell attributes, normalize the amplitude of the echo data, calibrate the RCS of the current echo data using the radar cross-sectional area RCS value under the reference distance and pulse width, and obtain the pre-processed data sample;

[0088] LSTM network training module: used for dividing the preprocessed data samples into a training set and a test set, and using the training set to train the long short-term memory neural network LSTM to obtain a trained LSTM network;

[0089] Shell attribute prediction module: It is used to preprocess the newly obtained radar sample data to be classified according to step S1, use the preprocessed data obtained later as test data, input the test data into the trained LSTM network for prediction, and obtain the predicted value of the shell attribute.

[0090] The embodiment of the present invention further provides a computer device, Figure 7 is a schematic diagram of the structure of a computer device provided by an embodiment of the present invention; see the accompanying drawings Figure 7 As shown, the computer device includes: an input device 23, an output device 24, a memory 22 and a processor 21; the memory 22 is used to store one or more programs; when the one or more programs are executed by the one or more processors 21, the one or more processors 21 implement the ballistic target recognition method based on the long short-term memory neural network provided in the above embodiment; wherein the input device 23, the output device 24, the memory 22 and the processor 21 can be connected by a bus or other means, Figure 7 The example of connecting through bus is taken in the following.

[0091] The memory 22 is a readable and writable storage medium of a computing device, which can be used to store software programs and computer executable programs, such as program instructions corresponding to the ballistic target recognition method based on the long short-term memory neural network described in the embodiment of the present invention; the memory 22 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required for a function; the data storage area can store data created according to the use of the device, etc.; in addition, the memory 22 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device; in some instances, the memory 22 can further include a memory remotely arranged relative to the processor 21, and these remote memories can be connected to the device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0092] The input device 23 may be used to receive input digital or character information, and to generate key signal input related to user settings and function control of the device; the output device 24 may include a display device such as a display screen.

[0093] The processor 21 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 22, that is, realizes the above-mentioned ballistic target recognition method based on long short-term memory neural network.

[0094] The computer device provided above can be used to execute the ballistic target recognition method based on long short-term memory neural network provided in the above embodiment, and has corresponding functions and beneficial effects.

[0095] The embodiment of the present invention also provides a storage medium containing computer executable instructions, which are used to execute the ballistic target recognition method based on long short-term memory neural network as provided in the above embodiment when executed by a computer processor. The storage medium is any of various types of memory devices or storage devices, and the storage medium includes: installation media, such as CD-ROM, floppy disk or tape device; computer system memory or random access memory, such as DRAM, DDRRAM, SRAM, EDORAM, Rambus RAM, etc.; non-volatile memory, such as flash memory, magnetic media (such as hard disk or optical storage); registers or other similar types of memory components, etc.; the storage medium may also include other types of memory or combinations thereof; in addition, the storage medium may be located in a first computer system in which the program is executed, or may be located in a different second computer system, which is connected to the first computer system via a network (such as the Internet); the second computer system may provide program instructions to the first computer for execution. The storage medium includes two or more storage media that may reside in different locations (for example, in different computer systems connected via a network). The storage medium may store program instructions (for example, specifically implemented as a computer program) that may be executed by one or more processors.

[0096] Of course, the storage medium containing computer executable instructions provided in an embodiment of the present invention is not limited to the ballistic target recognition method based on long short-term memory neural network as described in the above embodiment, and can also execute related operations in the ballistic target recognition method based on long short-term memory neural network provided in any embodiment of the present invention.

[0097] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0098] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A ballistic target recognition method based on long short-term memory neural network, characterized in that: The following steps are involved: S1. Obtain radar sample data to be classified, record the echo data of the target to be classified with projectile attributes, normalize the amplitude of the echo data, and use the radar cross-sectional area RCS value under the reference distance and pulse width to perform RCS calibration on the current echo data to obtain the pre-processed data sample; S2, dividing the preprocessed data samples into a training set and a test set, using the training set to train a long short-term memory neural network LSTM to obtain a trained LSTM network; S3, preprocessing the newly obtained radar sample data to be classified according to step S1, using the preprocessed data obtained later as test data, inputting the test data into the trained LSTM network for prediction, and obtaining the predicted value of the shell attribute; The method for obtaining radar sample data to be classified in step S1 includes: subjecting the echo data to pulse compression, moving target detection MTD, constant false alarm detection CFAR, and target tracking to obtain point trace data of the projectile target, and using the original echo pulse compressed data corresponding to the point trace data as the radar sample data to be classified; The method of using the original echo pulse compressed data corresponding to the point trace data as radar sample data to be classified includes: Assume that the point data of the projectile target is k Group, k The group trace data contains M k Group pulse, M k The number of pulses in each group is N k , intercept M k The first 8 seconds of data in the group pulse, extract the last P The IQ data of the group pulse is obtained K Group P × N k dimensional IQ data as radar sample data to be classified; No. k The IQ data of the group is expressed as: （1） In formula (1), Is P × N k indivual s ij Arranged P OK N k A matrix of columns; s ij Indicates i The first observation j The sampling value of a pulse on the target, s ij is a plural number; The method for normalizing the amplitude of the echo data in step S1 includes: Set up k Group IQ data includes p The first pulse in the group p The IQ data of the group pulse is , the target signal-to-noise ratio under the current observation is SNR p , then the normalized target amplitude is expressed as: （2） In formula (2), For the k The first p Detection point i The echo amplitude of a pulse; For the k The first p The average value of all pulse echo amplitudes at a detection point; Indicates converting the target signal-to-noise ratio under the current observation into a linear amplitude.

2. The ballistic target recognition method based on long short-term memory neural network according to claim 1 is characterized in that: The method for performing RCS calibration on the current echo data in step S1 includes: Assume the pulse width of the current echo data is T p , the target distance is R p , the reference pulse width is set to T ref , the reference distance is set to R ref , the reference signal-to-noise ratio is SNR ref , the RCS under the reference distance and reference pulse width is RCS ref , calibrate the RCS according to the radar equation and get the calibrated data .

3. The ballistic target recognition method based on long short-term memory neural network according to claim 1 is characterized in that: The method for training the long short-term memory neural network LSTM in step S2 includes: The training set is input into the LSTM network to obtain an output value, and a loss function value is calculated based on a comparison between the output value and the actual value. The loss function value is used to optimize the parameters of the LSTM network so that the loss function value is reduced to within a set value.

4. The ballistic target recognition method based on long short-term memory neural network according to claim 1 is characterized in that: The method of inputting the test data into the trained LSTM network for prediction in step S3 includes: The newly obtained radar sample data to be classified: M k × N k The IQ data is preprocessed to obtain P × N k dimensional test data, the P × N k The dimensional test data is input into the trained LSTM network to predict the shell attributes.

5. A ballistic target recognition system based on long short-term memory neural network, executing the ballistic target recognition method based on long short-term memory neural network as claimed in any one of claims 1 to 4, characterized in that: include: Obtain radar sample data module to be classified: used to record the echo data of the target to be classified with shell attributes, normalize the amplitude of the echo data, calibrate the RCS of the current echo data using the radar cross-sectional area RCS value under the reference distance and pulse width, and obtain the pre-processed data sample; LSTM network training module: used for dividing the preprocessed data samples into a training set and a test set, and using the training set to train the long short-term memory neural network LSTM to obtain a trained LSTM network; Shell attribute prediction module: It is used to preprocess the newly obtained radar sample data to be classified according to step S1, use the preprocessed data obtained later as test data, input the test data into the trained LSTM network for prediction, and obtain the predicted value of the shell attribute.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the ballistic target recognition method based on long short-term memory neural network described in any one of claims 1 to 4 are implemented.

7. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the ballistic target recognition method based on long short-term memory neural network as described in any one of claims 1 to 4 are implemented.

Images