Radar anti-ISRJ method and device based on two-dimensional complex convolutional network

Abstract

The present application relates to a radar anti-ISRJ method and device based on a two-dimensional complex convolutional network. The method obtains a sample data set of radar echo signals by simulation means, uses a signal transformation method based on matched filter design to match and transform the echo signal samples in the sample data set to obtain corresponding matching matrix samples, and generates corresponding ideal enhanced matching matrix samples, and then constructs training data to train the matching transformation enhancement network, converts the measured echo data into the corresponding matching matrix and inputs it into the trained matching transformation enhancement network to obtain a matching probability prediction matrix, and then judges the target in the echo signal based on the matching probability prediction matrix to achieve anti-ISRJ target recognition. The method can be applied to different radar signal systems, and the parameter estimation accuracy of the signal component is higher, solving the problem of difficulty in radar identification of real targets and false targets caused by intermittent sampling and forwarding interference.

Description

Technical Field

[0001] The present application relates to the technical field of anti-intermittent sampling and forwarding interference signal processing for pulse radar, and in particular to a radar anti-ISRJ method and device based on a two-dimensional complex convolutional network. Background Art

[0002] Interrupted-Sampling Repeater Jamming (ISRJ) is a radar main lobe jamming method that has emerged in recent years and poses a great threat to radar detection, tracking and target identification.

[0003] The ISRJ jammer intercepts and samples a portion of the radar's transmission signals, does not process the sampled signals or forwards them after a certain modulation, so that the jammer can reach the radar receiver in the same wave gate as the real target echo. The signal forwarded by the jammer is coherent with the radar transmission signal. The jammer can obtain a high signal processing gain after pulse compression processing, and the forwarded jammer signal is only a one-way attenuation in space relative to the target echo, so the ISRJ jammer can flexibly form deception or suppression interference in the radar with very small transmission energy.

[0004] However, there are still major problems with existing radar anti-ISRJ methods, mainly reflected in the fact that most passive anti-ISRJ methods are only applicable to radars under the linear frequency modulation (LFM) signal system and their performance is greatly dependent on the accuracy of ISRJ working parameter estimation. With the development of radar technology, the radar transmission waveform will not be limited to LFM signals. Therefore, it is necessary to study anti-ISRJ methods that can be applied to different radar transmission signal systems and have higher ISRJ working parameter estimation accuracy. Summary of the invention

[0005] Based on this, it is necessary to provide a radar anti-ISRJ method and device based on a two-dimensional complex convolutional network that can be applied to different radar signal systems and has higher parameter estimation accuracy for signal components in order to address the above technical problems.

[0006] A radar anti-ISRJ method based on a two-dimensional complex convolutional network, the method comprising:

[0007] A sample data set of radar echo signals is obtained based on simulation means, wherein the sample data set includes echo signal samples that are not disturbed and echo signal samples that are polluted by intermittent sampling and forwarding interference;

[0008] Using a signal transformation method based on matched filtering design and a reference signal to perform matching transformation on the echo signal samples in the sample data set to obtain corresponding matching matrix samples;

[0009] Generate corresponding ideal enhanced matching matrix samples according to the simulation parameters of the echo signal samples, and construct a set of training data according to the matching matrix samples and the corresponding ideal enhanced matching matrix samples;

[0010] Using the training data to train a matched transform enhancement network to obtain a trained matched transform enhancement network, wherein the trained matched transform enhancement network has a signal component matching probability prediction capability;

[0011] Acquire the echo signal received by the radar system, convert it into a corresponding matching matrix, and then input it into the trained matching transformation enhancement network to obtain a matching probability prediction matrix;

[0012] The target in the echo signal is judged based on the matching probability prediction matrix to achieve anti-ISRJ target identification.

[0013] In one embodiment, when the signal transformation method based on matched filtering design and the reference signal are used to perform matching transformation on the echo signal samples in the sample data set:

[0014] Sampling the reference signal and the echo signal samples to obtain discrete reference signal and discrete echo signal samples;

[0015] According to the signal matching transformation method based on matched filter design, the discrete reference signal is multiplied point by point by the discrete echo signal samples from left to right to obtain the corresponding matching matrix samples;

[0016] The signal matching transformation method is expressed as:

[0017]

[0018] In the above formula, Represented as echo signal samples, where is the number of echo sampling points within the receiving gate, is represented as the reference signal, where is the number of reference signal points.

[0019] In one embodiment, generating corresponding ideal enhanced matching matrix samples according to simulation parameters of the echo signal samples comprises:

[0020] Generate a corresponding all-zero matrix according to the simulation parameters of the echo signal samples;

[0021] Calculate the time sampling point according to the real target distance during simulation, set all the elements in the row corresponding to the time sampling point in the all-zero matrix to "1", and obtain the intermediate ideal enhanced matching matrix sample;

[0022] According to the jammer parameters during simulation, the time sampling point of the signal component forwarded each time by the ISRJ jammer appears at the radar receiving wave gate, and the duration of all slices forwarded each time is set to "1" for the corresponding elements in the intermediate ideal enhanced matching matrix sample to obtain the ideal enhanced matching matrix sample.

[0023] In one embodiment, when the matching transform enhancement network is trained using the training data, the matching matrix of the echo signal samples in the training data is used as the input data of the matching transform enhancement network, and the corresponding ideal enhanced matching matrix samples are used as label data. Before the matching matrix of the echo signal samples is input into the matching transform enhancement network, the real part and the imaginary part of the echo signal samples are split into two data channels.

[0024] In one embodiment, the matching transformation enhancement network is constructed based on a channelized two-dimensional complex convolutional layer, and the two-dimensional complex convolutional network is composed of a stack of multiple layers of two-dimensional complex convolutional network layers;

[0025] The data processing process of the two-dimensional complex convolution layer in the matching transformation enhancement network is expressed as:

[0026]

[0027] In the above formula, ,in, , , is the function for extracting the real part of a complex number, is the function for extracting the imaginary part of a complex number, Indicates that the data is added in the new dimension To stack, It represents the matching matrix obtained by multiplying the discrete reference signal by the discrete echo signal samples from left to right point by point. Indicates the number of channels, and They represent the real convolution layer representing the real part parameters and the real convolution layer representing the imaginary part, respectively. Indicates that the data is in the existing dimension Make stitching.

[0028] In one embodiment, judging the target in the echo signal based on the matching probability prediction matrix includes:

[0029] Using a preset threshold to judge the probability prediction value of each pixel in the matching probability prediction matrix to generate an enhanced matching matrix;

[0030] Summing the enhanced matching matrix row by row to obtain a signal duration vector;

[0031] Calculate the signal component duty cycle based on the signal duration vector and the radar transmission signal duration;

[0032] The duty cycle of the signal component is judged according to the true and false target judgment threshold, thereby achieving the recognition of true and false targets.

[0033] In one embodiment, the radar signal structure of the echo signal is a pulse signal with arbitrary intra-pulse modulation, including a linear frequency modulation signal, a phase coded signal, a frequency coded signal, and a frequency-phase integrated coded signal.

[0034] The present application also provides a radar anti-ISRJ device based on a two-dimensional complex convolutional network, the device comprising:

[0035] A sample data set construction module, used to obtain a sample data set of radar echo signals based on simulation means, wherein the sample data set includes echo signal samples that are not disturbed and echo signal samples that are polluted by intermittent sampling and forwarding interference;

[0036] A matching matrix sample generation module, used to perform matching transformation on the echo signal samples in the sample data set using a signal transformation method based on matched filter design and a reference signal to obtain corresponding matching matrix samples;

[0037] A training data construction module, used to generate corresponding ideal enhanced matching matrix samples according to the simulation parameters of the echo signal samples, and to construct a set of training numbers according to the matching matrix samples and the corresponding ideal enhanced matching matrix samples;

[0038] A neural network training module, used to train a matched transform enhancement network using the training data to obtain a trained matched transform enhancement network, wherein the trained matched transform enhancement network has a signal component matching probability prediction capability;

[0039] A matching probability prediction matrix obtaining module is used to obtain the echo signal received by the radar system, convert it into a corresponding matching matrix, and then input it into the trained matching transformation enhancement network to obtain a matching probability prediction matrix;

[0040] The anti-ISRJ target recognition module is used to judge the target in the echo signal based on the matching probability prediction matrix to achieve anti-ISRJ target recognition.

[0041] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0042] A sample data set of radar echo signals is obtained based on simulation means, wherein the sample data set includes echo signal samples that are not disturbed and echo signal samples that are polluted by intermittent sampling and forwarding interference;

[0043] Using a signal transformation method based on matched filtering design and a reference signal to perform matching transformation on the echo signal samples in the sample data set to obtain corresponding matching matrix samples;

[0044] Generate corresponding ideal enhanced matching matrix samples according to the simulation parameters of the echo signal samples, and construct a set of training data according to the matching matrix samples and the corresponding ideal enhanced matching matrix samples;

[0045] Using the training data to train a matched transform enhancement network to obtain a trained matched transform enhancement network, wherein the trained matched transform enhancement network has a signal component matching probability prediction capability;

[0046] Acquire the echo signal received by the radar system, convert it into a corresponding matching matrix, and then input it into the trained matching transformation enhancement network to obtain a matching probability prediction matrix;

[0047] The target in the echo signal is judged based on the matching probability prediction matrix to achieve anti-ISRJ target identification.

[0048] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0049] A sample data set of radar echo signals is obtained based on simulation means, wherein the sample data set includes echo signal samples that are not disturbed and echo signal samples that are polluted by intermittent sampling and forwarding interference;

[0050] Using a signal transformation method based on matched filtering design and a reference signal to perform matching transformation on the echo signal samples in the sample data set to obtain corresponding matching matrix samples;

[0051] Generate corresponding ideal enhanced matching matrix samples according to the simulation parameters of the echo signal samples, and construct a set of training data according to the matching matrix samples and the corresponding ideal enhanced matching matrix samples;

[0052] Using the training data to train a matched transform enhancement network to obtain a trained matched transform enhancement network, wherein the trained matched transform enhancement network has a signal component matching probability prediction capability;

[0053] Acquire the echo signal received by the radar system, convert it into a corresponding matching matrix, and then input it into the trained matching transformation enhancement network to obtain a matching probability prediction matrix;

[0054] The target in the echo signal is judged based on the matching probability prediction matrix to achieve anti-ISRJ target identification.

[0055] The above-mentioned radar anti-ISRJ method and device based on two-dimensional complex convolutional network obtains a sample data set of radar echo signals through simulation means, uses a signal transformation method based on matched filter design to match and transform the echo signal samples in the sample data set to obtain corresponding matching matrix samples, and generates corresponding ideal enhanced matching matrix samples, and then constructs training data to train the matching transformation enhancement network, converts the measured echo data into the corresponding matching matrix and inputs it into the trained matching transformation enhancement network to obtain the matching probability prediction matrix, and then judges the target in the echo signal based on the matching probability prediction matrix to achieve anti-ISRJ target recognition. This method can be applied to different radar signal systems, and the parameter estimation accuracy of the signal component is higher, solving the problem of difficulty in radar identification of real targets and false targets caused by intermittent sampling and forwarding interference. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 1 is a flow chart of a radar anti-ISRJ method based on a two-dimensional complex convolutional network in one embodiment;

[0057] Figure 2 A block diagram of specific steps of a radar anti-ISRJ method based on a two-dimensional complex convolutional network in one embodiment;

[0058] Figure 3 In one embodiment, a radar is simulated to receive an echo and basic information of a target and ISRJ interference contained in the echo, wherein: Figure 3 (a) is a schematic diagram of the time domain waveform of the radar receiving echo. Figure 3 (b) is a schematic diagram of the one-dimensional image of the radar receiving echo. Figure 3 (c) is a schematic diagram of the time-dimensional image of the radar receiving echo corresponding to the target echo. Figure 3 (d) is a schematic diagram of the time one-dimensional image of the radar receiving echo corresponding to the ISRJ interference echo;

[0059] Figure 4 Schematic diagram of matching transformation results of simulated radar receiving echo in one embodiment, wherein: Figure 4 (a) is the matching transformation phase diagram of the radar receiving echo, Figure 4 (b) is the matching transformation amplitude diagram of the radar receiving echo. Figure 4 (c) is the ideal enhanced matching matrix for radar receiving echo;

[0060] Figure 5 FIG. 1 is a schematic diagram of an enhancement result of a matching matrix using a matching transformation enhancement network in one embodiment, wherein: Figure 5(a) is a schematic diagram of the matching matrix result of radar receiving echo. Figure 5 (b) is a schematic diagram of the enhanced matching transformation of radar received echo;

[0061] Figure 6 FIG. 1 is a schematic diagram of an enhancement result of a matching matrix using a matching transformation enhancement network in one embodiment, wherein: Figure 6 (a) is a schematic diagram of the duration estimation of the signal component of the radar received echo. Figure 6 (b) is a schematic diagram of the duty cycle estimation of the signal component of the radar received echo. Figure 6 (c) is a schematic diagram showing the duration of the actual signal component of the radar received echo. Figure 6 (d) is a schematic diagram of the duty cycle of the actual signal component of the radar received echo;

[0062] Figure 7 It is a structural block diagram of a radar anti-ISRJ device based on a two-dimensional complex convolutional network in one embodiment;

[0063] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0065] In view of the problems existing in the existing technology, radar anti-ISRJ interference method, such as Figure 1 As shown, the radar anti-ISRJ method based on a two-dimensional complex convolutional network provided by the present application specifically includes the following steps:

[0066] Step S100, obtaining a sample data set of radar echo signals based on simulation means, wherein the sample data set includes echo signal samples that are not interfered with and echo signal samples that are polluted by intermittent sampling and forwarding interference.

[0067] Step S110 , using a signal transformation method based on matched filter design and a reference signal to perform matching transformation on the echo signal samples in the sample data set to obtain corresponding matching matrix samples.

[0068] Step S120: Generate corresponding ideal enhanced matching matrix samples according to the simulation parameters of the echo signal samples, and construct a set of training data according to the matching matrix samples and the corresponding ideal enhanced matching matrix samples.

[0069] Step S130, using the training data to train the matched transform enhancement network to obtain a trained matched transform enhancement network, wherein the trained matched transform enhancement network has the ability to predict signal component matching probabilities.

[0070] Step S140, obtaining the echo signal received by the radar system, converting it into a corresponding matching matrix and inputting it into the trained matching transformation enhancement network to obtain a matching probability prediction matrix.

[0071] Step S150, judging the target in the echo signal based on the matching probability prediction matrix to achieve anti-ISRJ target recognition.

[0072] In the present application, a pulse radar anti-ISRJ interference method based on matching transformation and two-dimensional complex convolution network is provided, which can be applied to different radar signal systems, and the parameter estimation accuracy of signal components is higher, solving the problem of difficulty in radar identification of real targets and intermittent sampling forwarding interference false targets. The method obtains the matching matrix of the echo signal by using a signal transformation method based on matched filtering design. The matching matrix can reflect the phase accumulation of different lengths in the echo signal, and then calculates the ideal enhanced matching matrix as a label for deep learning model training, thereby obtaining a deep learning model that can output an enhanced matching matrix containing only the duration of the time component. When the model is applied, the pre-processed matching matrix is ​​directly input to output the enhanced result. Finally, the output result is threshold judged, the value is converted into a 01 distribution, and then the sum is obtained to obtain the signal duration of each time component, and the real target and the intermittent sampling forwarding interference false target are judged by calculating the duty cycle. The method can perceive ISRJ interference without relying on the radar transmission signal system, and improves the accuracy of parameter estimation of different signal components through deep learning methods.

[0073] Specifically, the method actually includes the neural network training part, i.e., the contents of step S100 to step S130, the matching matrix enhancement part using the trained neural network, i.e., the contents of step S140, and the anti-ISRJ target recognition part based on the enhanced matching matrix, i.e., the contents of step S150. The latter two parts can also be regarded as one part.

[0074] In this method, the training data generation in the training part of the neural network and the post-training neural network application part both convert the echo signal samples and the measured echo signals into corresponding matching matrix samples and matching matrices based on the signal transformation method designed by matched filtering.

[0075] After analyzing the similarities and differences between the ISRJ signal and the target echo signal, a signal transformation method derived from matched filtering is used to transform the echo into a two-dimensional plane in the time domain and the delay domain. This transformation can be called a matching transformation, and the plane where the signal components exist after the transformation is called a matching transformation domain. In this transformation domain, the ISRJ signal component and the target echo component will be separated in the delay domain, and their distribution in the time domain will be described by the complex value curve in the relative delay domain. The specific transformation method is described as follows:

[0076] Assuming that the radar's transmitted waveform satisfies the constant modulus constraint, the pulse waveform emitted by any radar can be expressed by the following formula:

[0077]

[0078] In formula (1), is the radar transmitting waveform pulse width, is the function of the radar transmitting waveform phase changing with time, and j represents the imaginary unit.

[0079] Assume that the ISRJ jammer generates Sampling pulse, The time width of a sampling pulse is , for A sampling pulse is performed If the signal is forwarded for times, the intermittent sampling pulse signal of the ISRJ jammer can be written as:

[0080]

[0081] In formula (2), Before The total time required for sampling and forwarding.

[0082] The signal sampled by the ISRJ jammer is:

[0083]

[0084] The total signal forwarded by ISRJ is:

[0085]

[0086] Assume that the radar illuminates The time delays corresponding to the distances of the targets are , and the backscattering coefficients are , m represents the number to be summed in the summation expression, and the value of m ranges from 1 to , Indicates the number of ISRJ jammer retransmissions, including the delay of the target carrying the ISRJ jammer , the complex amplitude of the ISRJ interference signal is , then the radar echo can be written as:

[0087]

[0088] In formula (5), is noise, usually Gaussian white noise, Indicates the target echo signal received by the radar.

[0089] Based on the above radar receiving echo expression, in this embodiment, the matching transformation proposed It can be described as follows:

[0090]

[0091] In formula (6), is the space of complex functions of one real independent variable, is the space of complex functions of two real independent variables, is the signal to be transformed, is the reference signal.

[0092] After matching transformation, the echo signal can be written as:

[0093]

[0094] In the radar signal processing process, analog-to-digital conversion is usually required to obtain discrete digital signals. For the case of discrete-time digital echo, assume that the echo received by the radar is :

[0095]

[0096] In formula (8), is the number of echo sampling points within the receiving gate

[0097]

[0098] In formula (9), is the number of reference signal points. Generally, the number of reference signal points is smaller than the number of received signal points. .

[0099] Then the matrix obtained after matching transformation can be described as follows:

[0100]

[0101] Formula (10) can be used to calculate the matching matrix of the echo signal very conveniently and intuitively.

[0102] Furthermore, when the signal transformation method based on matched filter design and the reference signal are used to match the echo signal samples in the sample data set: the reference signal and the echo signal samples are sampled to obtain discrete reference signals and discrete echo signal samples, and the discrete reference signal and the discrete echo signal samples are multiplied point by point from left to right according to the signal matching transformation method based on matched filter design, i.e., formula (10), to obtain the corresponding matching matrix samples. The result of each multiplication calculation is used as the row of the matching matrix, and each calculation is moved to the right a little bit until the rightmost sampling point of the reference signal is aligned with the rightmost sampling point of the direct wave vector.

[0103] In this embodiment, the reference signal may be a direct wave signal or a radar transmission channel bypass signal. For a non-cooperative dual-station radar, the reference signal may be a direct wave signal; for a cooperative dual-station radar, the reference signal may be a radar transmission signal transmitted by communication means or a signal reconstructed based on transmission signal parameters; for a single-station radar, the reference signal is generally a transmission signal obtained by bypassing the radar transmission channel, or may be a reference signal pre-stored in the radar waveform design stage.

[0104] Next, considering that in an ideal situation, the length of the in-phase accumulation of the matching transform can reflect the duration of different signal components, the duration of the target echo is the same as the radar transmission pulse width, and the ISRJ signal component must have a short duration or discontinuity in the time domain due to its intermittent sampling and forwarding mechanism. Therefore, by calculating the duration of each signal component on the matching matrix, the real target and the intermittent sampling and forwarding interference false target can be distinguished. However, it is difficult to directly observe the position and duration of all signal components from the phase diagram or amplitude diagram of the matching transform alone. Based on this, there is a large error in obtaining the duration of the time component, especially in the presence of noise and when the SNR is low and the JNR is high, it is difficult to see the phase accumulation even if the phase diagram is directly observed.

[0105] Therefore, in this embodiment, the matching matrix is ​​enhanced by a deep learning method to directly obtain the duration of the received echo corresponding to the signal at each time delay.

[0106] Since the neural network needs to be trained, a corresponding training data set needs to be constructed before training. In step S100, the echo signal of the radar system is simulated by simulation means, including the echo signal without interference and the echo signal interfered by ISRJ. In step S110, the echo signal samples are transformed into corresponding matching matrix samples using the above-mentioned signal transformation method based on matched filter design as the input data of the neural network.

[0107] Furthermore, training deep learning models also requires clarifying the form of the ideal enhanced matching matrix, that is, the label of the data set, which is defined as follows: the enhanced matching matrix has the same shape as the original matching matrix, each row represents the time component signal of the corresponding delayed sampling point, the point with signal matching takes the value of 1, and the point without signal matching takes the value of 0.

[0108] In the ideal enhanced matching matrix, the signal component corresponding to the real target only needs to calculate the corresponding time sampling point according to the target distance set during simulation, and then set the corresponding row in the enhanced matching matrix to 1; while the ISRJ interference signal component needs to combine the target distance carrying the jammer, the jammer slice width and the forwarding mode to calculate the corresponding time sampling delay point when the interference signal component appears and the duration corresponding to the reference signal.

[0109] Assume that the radar illuminates The distances to the targets are , the radar receiver sampling rate is , the pulse width of the transmitting signal is , the radar receiving gate time width is , the target number carrying the jammer is , the corresponding target distance is , the jammer slice width is , the forwarding mode is intermittent sampling and repeated forwarding, and the number of forwarding times is , for the speed of light express.

[0110] For The time sampling delay point on the enhanced matching matrix corresponding to the real target is:

[0111]

[0112] Then, The time sampling delay point at which the interference signal component generated by the second forwarding appears is:

[0113]

[0114] Calculate the number of slices based on the jammer forwarding mode and forwarding times:

[0115]

[0116] In formula (13), is the ceiling function, To take the minimum function.

[0117] Then, The starting time sampling point of each slice is:

[0118]

[0119] The end time sampling point is:

[0120]

[0121] In this embodiment, generating the corresponding ideal enhanced matching matrix sample according to the simulation parameters of the echo signal sample includes: first generating the corresponding all-zero matrix according to the matching matrix sample, and then calculating the time sampling delay point according to the actual target distance during simulation. , set all the elements in the row corresponding to the time sampling delay point in the all-zero matrix to "1", and obtain the intermediate ideal enhanced matching matrix sample. According to the jammer parameters during simulation, calculate the time sampling point of the signal component forwarded by the ISRJ jammer at the radar receiving wave gate each time , and the duration of all slices forwarded each time , The corresponding elements in the middle ideal enhanced matching matrix sample are set to "1" to obtain the ideal enhanced matching matrix sample.

[0122] In step S130, when the matching transform enhancement network is trained using the training data, the echo signal samples in the training data are used as input data of the matching transform enhancement network, and the corresponding ideal enhanced matching matrix samples are used as label data. Before the echo signal samples are input into the matching transform enhancement network, the real and imaginary parts of the echo signal sample matching matrix are split into two data channels.

[0123] Considering that the traditional real network cannot perform complex operations, the input and output of the network layer can only be real values, and usually can only process the amplitude information of the complex signal, but cannot obtain the phase information. For radar signal processing, the use of a complex network can first extract the amplitude and phase information from the complex input data, and then further learn deeper features to obtain better feature representation capabilities. However, different versions of the neural network framework have different support for complex data. Considering the versatility problem in model deployment, in this embodiment, a channelized complex convolutional network is designed, and different real channels are used to save the real and imaginary information of the complex number. Two corresponding real network layers are used in the complex network layer to receive the real channel and the complex channel to calculate the real and imaginary parts of the output data respectively, thereby realizing complex operations. In the network construction, the final output layer needs to use a real convolutional layer to fuse the real and imaginary parts of the final feature map to obtain a matching probability prediction matrix.

[0124] Assume a batch of matching matrices Shape ,in is the number of signal frames contained in a batch of data, that is, the number of matching matrices, is the height of the matching matrix, which can be obtained by calculate, is the matching matrix, whose value is equal to the pulse width of the transmitted signal , then the input data required by the channelized complex convolutional network can be calculated by formula (16-18):

[0125]

[0126] In formula (16)-formula (18), is the function for extracting the real part of a complex number, is the function for extracting the imaginary part of a complex number, Indicates that the data is added in the new dimension To stack, It represents the matching matrix obtained by multiplying the discrete reference signal by the discrete echo signal samples from left to right point by point.

[0127] So, we get Shape As the input data propagates forward through the network, the number of channels changes, but the matrix shape remains constant due to the padding setting. unchanged, then for each network layer's input data , a more general shape expression is , Represents the number of channels. The first half of the channels store the real part information, and the second half of the channels store the imaginary part information.

[0128] Taking the channelized two-dimensional complex convolution layer as an example, assume and Respectively represent the real convolution layer representing the real part parameter and the real convolution layer representing the imaginary part. Then in the channelized complex convolution network, the basic calculation principle can be expressed by formula (19-21):

[0129]

[0130] In formula (19)-formula (21), Indicates that the data is in the existing dimension Make stitching.

[0131] The essence of matching transformation that can characterize the duration of signal components is in-phase accumulation, which is reflected in the matching matrix as horizontal lines with the same phase at corresponding positions in the phase diagram. The amplitude diagram combined with the phase diagram can reflect the start time and duration of the signal component, but this information is greatly affected by noise, and it is easy to affect each other in the presence of multiple targets and multiple interferences, making it difficult to observe. In order to improve the performance of the model, a simple feature engineering is designed and embedded into the network as the input layer, and the feature engineering is called a horizontal accumulation module. The specific operation is: use a fixed-parameter all-1 convolution layer to perform a horizontal one-dimensional convolution on the input data, set different convolution kernel parameters to obtain horizontal accumulations of different lengths, and then splice all convolution results in the channel dimension to obtain the input data after simple feature extraction.

[0132] In this embodiment, after the matched transform enhancement network is trained using training data, a matched transform enhancement network having the ability to predict signal component matching probabilities is obtained.

[0133] In step S140, the echo signal received by the radar system is converted into a corresponding matching matrix using the above method, and then input into the trained matching transformation enhancement network to obtain a matching probability prediction matrix of the probability of a signal existing at each pixel point in the matching matrix.

[0134] In step S150, judging the target in the echo signal based on the matching probability prediction matrix includes: first using a preset threshold The probability prediction value of each pixel in the matching probability prediction matrix is ​​judged to generate an enhanced matching matrix Then sum the enhanced matching matrix to get the signal duration vector , the signal component duty cycle is calculated according to the signal duration vector and the radar transmission signal time width, and finally, the signal component duty cycle is judged according to the true and false target judgment threshold, so as to realize the recognition of true and false targets.

[0135] Specifically, the entire matching probability prediction matrix is ​​thresholded and the values ​​are converted into 01 distribution. The pixel point is considered as a no-signal point (noise point), with a value of 0, which is greater than or equal to the threshold The pixel points are regarded as signal points and the value is 1, thus obtaining the enhanced matching matrix:

[0136]

[0137] Furthermore, in the enhanced matching matrix, each row represents the signal component of the corresponding time sampling point, and the value of each pixel represents the presence or absence of a signal. Then the enhanced matching matrix is ​​summed up in rows, and the result vector The value of each element in represents the duration of the signal component at the corresponding time sampling point:

[0138]

[0139] Furthermore, the obtained signal duration vector In the example, all non-zero element index values ​​represent the existence of a target at the corresponding time sampling delay point. These element values ​​are compared with the radar transmission signal time width. Divide to get the signal component duty cycle:

[0140]

[0141] Specifically, set the threshold for determining true or false targets ,when The corresponding time sampling point is The signal component at is the target echo component. Otherwise, The signal component at is the ISRJ interference echo component.

[0142] Preferably, the true and false target judgment threshold is 50%.

[0143] In this embodiment, the radar signal system of the echo signal can be a pulse signal with arbitrary intra-pulse modulation, including but not limited to a linear frequency modulation signal, a phase coded signal, a frequency coded signal, a frequency-phase integrated coded signal, and the like.

[0144] like Figure 2 As shown, it is a block diagram of the specific steps of the radar anti-ISRJ method based on the two-dimensional complex convolutional network.

[0145] In this paper, different schematic diagrams are also given to assist in understanding the method and to demonstrate its effectiveness, among which: Figure 3 The basic information of the simulated radar receiving echo and the target and ISRJ interference contained in the echo is given. Figure 4 The matching transformation results of the simulated radar receiving echo are given. Figure 4 (a) It can be seen that when the signal component is partially or completely consistent with the reference signal through time shift, the phase will become the same in the consistent time period, which is reflected in the corresponding position of the phase diagram. Figure 4 (b) The amplitude information can reflect the starting time and duration of the signal component. Figure 4 The exact location of the signal components can be observed more clearly in (c). Figure 5 The enhanced results of the deep learning model on the matching transformation are given, combined with Figure 4It can be seen that this enhancement method is feasible and effective. The enhanced matching matrix obtained by global threshold judgment can clearly reflect the distribution of each signal component in the radar received echo. Figure 6 The signal component duration estimation and duty cycle estimation results are shown, combined with Figure 3 Given the one-dimensional image, it is easy to see that the number of time sampling points corresponding to the signal component with a duty cycle greater than or equal to the threshold is consistent with the number of time sampling points of the target reflected in the one-dimensional image of the target echo. Further comparison with the actual situation verifies the correctness of the algorithm.

[0146] In the above radar anti-ISRJ method based on two-dimensional complex convolutional network, the essential difference between the duration of target echo and ISRJ interference echo in the time domain is fully utilized, the in-phase accumulation property in matched filtering is utilized, and a signal transformation method based on matched filtering design is adopted. On this basis, a matching matrix enhancement method is designed. Aiming at the problem that the matching matrix itself is difficult to extract the duration of signal components, a two-dimensional complex convolution deep learning network is designed to integrate phase and amplitude information. Considering the situation that the phase information is not clear enough, a fixed-parameter lateral one-dimensional convolution is used in the input layer of the network to enhance the in-phase accumulation feature, and a threshold judgment is performed on the model output to obtain an enhanced matching matrix that can accurately reflect the signal components of each time sampling delay point, so that the distribution of signal components can be obtained, and the accurate identification of real targets and false targets caused by intermittent sampling and forwarding interference can be achieved by calculating the duration and duty cycle of the signal components.

[0147] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0148] In one embodiment, Figure 7 As shown, a radar anti-ISRJ device based on a two-dimensional complex convolutional network is provided, comprising: a sample data set construction module 200, a matching matrix sample generation module 210, a training data construction module 220, a neural network training module 230, a matching probability prediction matrix acquisition module 240 and an anti-ISRJ target recognition module 250, wherein:

[0149] A sample data set construction module 200 is used to obtain a sample data set of radar echo signals based on simulation means, wherein the sample data set includes echo signal samples that are not disturbed and echo signal samples that are polluted by intermittent sampling and forwarding interference;

[0150] A matching matrix sample generating module 210 is used to perform matching transformation on the echo signal samples in the sample data set by using a signal transformation method based on matched filter design and a reference signal to obtain corresponding matching matrix samples;

[0151] A training data construction module 220, configured to generate corresponding ideal enhanced matching matrix samples according to the simulation parameters of the echo signal samples, and to construct a set of training data according to the matching matrix samples and the corresponding ideal enhanced matching matrix samples;

[0152] A neural network training module 230 is used to train a matched transform enhancement network using the training data to obtain a trained matched transform enhancement network, wherein the trained matched transform enhancement network has a signal component matching probability prediction capability;

[0153] A matching probability prediction matrix obtaining module 240 is used to obtain the echo signal received by the radar system, convert it into a corresponding matching matrix, and then input it into the trained matching transformation enhancement network to obtain a matching probability prediction matrix;

[0154] The anti-ISRJ target recognition module 250 is used to judge the target in the echo signal based on the matching probability prediction matrix to achieve anti-ISRJ target recognition.

[0155] For the specific definition of the radar anti-ISRJ device based on the two-dimensional complex convolutional network, please refer to the definition of the radar anti-ISRJ method based on the two-dimensional complex convolutional network mentioned above, which will not be repeated here. Each module in the above-mentioned radar anti-ISRJ device based on the two-dimensional complex convolutional network can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0156] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 8As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a radar anti-ISRJ method based on a two-dimensional complex convolutional network is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0157] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0158] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0159] A sample data set of radar echo signals is obtained based on simulation means, wherein the sample data set includes echo signal samples that are not disturbed and echo signal samples that are polluted by intermittent sampling and forwarding interference;

[0160] Using a signal transformation method based on matched filtering design and a reference signal to perform matching transformation on the echo signal samples in the sample data set to obtain corresponding matching matrix samples;

[0161] Generate corresponding ideal enhanced matching matrix samples according to the simulation parameters of the echo signal samples, and construct a set of training data according to the matching matrix samples and the corresponding ideal enhanced matching matrix samples;

[0162] Using the training data to train a matched transform enhancement network to obtain a trained matched transform enhancement network, wherein the trained matched transform enhancement network has a signal component matching probability prediction capability;

[0163] Acquire the echo signal received by the radar system, convert it into a corresponding matching matrix, and then input it into the trained matching transformation enhancement network to obtain a matching probability prediction matrix;

[0164] The target in the echo signal is judged based on the matching probability prediction matrix to achieve anti-ISRJ target identification.

[0165] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0166] A sample data set of radar echo signals is obtained based on simulation means, wherein the sample data set includes echo signal samples that are not disturbed and echo signal samples that are polluted by intermittent sampling and forwarding interference;

[0167] Using a signal transformation method based on matched filtering design and a reference signal to perform matching transformation on the echo signal samples in the sample data set to obtain corresponding matching matrix samples;

[0168] Generate corresponding ideal enhanced matching matrix samples according to the simulation parameters of the echo signal samples, and construct a set of training data according to the matching matrix samples and the corresponding ideal enhanced matching matrix samples;

[0169] Using the training data to train a matched transform enhancement network to obtain a trained matched transform enhancement network, wherein the trained matched transform enhancement network has a signal component matching probability prediction capability;

[0170] Acquire the echo signal received by the radar system, convert it into a corresponding matching matrix, and then input it into the trained matching transformation enhancement network to obtain a matching probability prediction matrix;

[0171] The target in the echo signal is judged based on the matching probability prediction matrix to achieve anti-ISRJ target identification.

[0172] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0173] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0174] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A radar anti-ISRJ method based on a two-dimensional complex convolutional network, characterized in that: The method comprises: A sample data set of radar echo signals is obtained based on simulation means, wherein the sample data set includes echo signal samples that are not disturbed and echo signal samples that are polluted by intermittent sampling and forwarding interference; Using a signal transformation method based on matched filtering design and a reference signal to perform matching transformation on the echo signal samples in the sample data set to obtain corresponding matching matrix samples; Generating a corresponding ideal enhanced matching matrix sample according to the simulation parameters of the echo signal sample, comprising: generating a corresponding all-zero matrix according to the simulation parameters of the echo signal sample, Calculate the time sampling point according to the real target distance during simulation, set all the elements in the row corresponding to the time sampling point in the all-zero matrix to "1", and obtain the intermediate ideal enhanced matching matrix sample; calculate the time sampling point where the signal component forwarded by the ISRJ jammer appears at the radar receiving wave gate each time according to the jammer parameters during simulation, and set the corresponding elements in the intermediate ideal enhanced matching matrix sample of all the slice durations forwarded each time to "1", and obtain the ideal enhanced matching matrix sample; Constructing a set of training data according to the matching matrix samples and the corresponding ideal enhanced matching matrix samples; Using the training data to train a matched transform enhancement network to obtain a trained matched transform enhancement network, wherein the trained matched transform enhancement network has a signal component matching probability prediction capability; Acquire the echo signal received by the radar system, convert it into a corresponding matching matrix, and then input it into the trained matching transformation enhancement network to obtain a matching probability prediction matrix; The target in the echo signal is judged based on the matching probability prediction matrix to achieve anti-ISRJ target identification.

2. The radar anti-ISRJ method according to claim 1, characterized in that: When the signal transformation method based on matched filtering design and the reference signal are used to perform matching transformation on the echo signal samples in the sample data set: Sampling the reference signal and the echo signal samples to obtain discrete reference signal and discrete echo signal samples; According to the signal matching transformation method based on matched filtering design, the discrete reference signal is multiplied point by point by the discrete echo signal samples from left to right to obtain the corresponding matching matrix samples; the signal matching transformation method is expressed as: In the above formula, Represented as echo signal samples, where is the number of echo sampling points within the receiving gate, is represented as the reference signal, where is the number of reference signal points.

3. The radar anti-ISRJ method according to any one of claims 1 or 2, characterized in that: When the matching transform enhancement network is trained using the training data, the matching matrix of the echo signal samples in the training data is used as the input data of the matching transform enhancement network, and the corresponding ideal enhanced matching matrix samples are used as label data. Before the matching matrix of the echo signal samples is input into the matching transform enhancement network, the real part and the imaginary part of the echo signal samples are split into two data channels.

4. The radar anti-ISRJ method according to claim 3, characterized in that: The matching transformation enhancement network is constructed based on a channelized two-dimensional complex convolutional layer, and the two-dimensional complex convolutional network is composed of a stack of multiple layers of two-dimensional complex convolutional network layers; The data processing process of the two-dimensional complex convolutional layer in the matching transformation enhancement network is expressed as: In the above formula, ,in, , , is the function for extracting the real part of a complex number, is the function for extracting the imaginary part of a complex number, Indicates that the data is added in the new dimension To stack, It represents the matching matrix obtained by multiplying the discrete reference signal by the discrete echo signal samples from left to right point by point. Indicates the number of channels, and They represent the real convolution layer representing the real part parameters and the real convolution layer representing the imaginary part, respectively. Indicates that the data is in the existing dimension Make stitching.

5. The radar anti-ISRJ method according to claim 4, characterized in that: The judging of the target in the echo signal based on the matching probability prediction matrix includes: Using a preset threshold to judge the probability prediction value of each pixel in the matching probability prediction matrix to generate an enhanced matching matrix; Summing the enhanced matching matrix row by row to obtain a signal duration vector; Calculate the signal component duty cycle based on the signal duration vector and the radar transmission signal duration; The duty cycle of the signal component is judged according to the true and false target judgment threshold, thereby achieving the recognition of true and false targets.

6. The radar anti-ISRJ method according to claim 5, characterized in that: The radar signal system of the echo signal is a pulse signal modulated arbitrarily within the pulse, including a linear frequency modulation signal, a phase coding signal, a frequency coding signal, and a frequency-phase integrated coding signal.

7. A radar anti-ISRJ device based on a two-dimensional complex convolutional network, characterized in that: The device comprises: A sample data set construction module, used to obtain a sample data set of radar echo signals based on simulation means, wherein the sample data set includes echo signal samples that are not disturbed and echo signal samples that are polluted by intermittent sampling and forwarding interference; A matching matrix sample generation module, used to perform matching transformation on the echo signal samples in the sample data set using a signal transformation method based on matched filter design and a reference signal to obtain corresponding matching matrix samples; The ideal enhanced matching matrix sample generation module is used to generate the corresponding ideal enhanced matching matrix samples according to the simulation parameters of the echo signal samples, including: generating the corresponding all-zero matrix according to the simulation parameters of the echo signal samples, calculating the time sampling point according to the real target distance during simulation, setting all the elements in the row corresponding to the time sampling point in the all-zero matrix to "1", and obtaining the intermediate ideal enhanced matching matrix samples; calculating the time sampling point at which the signal component forwarded by the ISRJ jammer appears at the radar receiving wave gate each time according to the jammer parameters during simulation, and setting the corresponding elements in the intermediate ideal enhanced matching matrix samples of the duration of all slices forwarded each time to "1", and obtaining the ideal enhanced matching matrix samples; A training data construction module, used to construct a set of training numbers according to the matching matrix samples and the corresponding ideal enhanced matching matrix samples; A neural network training module, used to train a matched transform enhancement network using the training data to obtain a trained matched transform enhancement network, wherein the trained matched transform enhancement network has a signal component matching probability prediction capability; A matching probability prediction matrix obtaining module is used to obtain the echo signal received by the radar system, convert it into a corresponding matching matrix, and then input it into the trained matching transformation enhancement network to obtain a matching probability prediction matrix; The anti-ISRJ target recognition module is used to judge the target in the echo signal based on the matching probability prediction matrix to achieve anti-ISRJ target recognition.

Images