Method for constructing radar active jamming classification model based on multi-dimensional image features

Abstract

The invention provides a method for constructing a radar active interference classification model based on multi-dimensional image features, and the method comprises the steps: constructing an active interference database under the target background of an unmanned plane group according to the obtained echo data of different interference types; obtaining a plurality of image domain data matrixes according to all the echo data in the database; performing feature analysis and extraction on each image domain data matrix to obtain a plurality of feature parameters, and constructing a feature parameter library by using each feature parameter; through the feature parameter library and the active interference database, a classification model established by a support vector machine (SVM) is trained to obtain a radar active interference classification model based on multi-dimensional image features, the effect of comprehensively extracting space-time characteristics of interference signals from an image domain is achieved, and by means of the constructed feature parameter library and an SVM classification algorithm, the radar active interference classification model is obtained. The noise type active interference signals are classified and recognized, accurate classification of the interference signals is achieved, and the recognition accuracy and robustness of the noise type active interference of the unmanned aerial vehicle are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing, and particularly relates to a method for constructing a radar active interference classification model based on multi-dimensional image features. Background Art

[0002] With the rapid development of unmanned aerial vehicle (UAV) technology, radar systems face severe challenges from noise-like active interferences when detecting UAVs. These interference signals not only increase the difficulty of target detection but also may affect the accurate identification and classification of UAV swarms by the radar system. Especially in the complex environment of UAV swarm operations, the diversity and dynamic changes of interference signals significantly increase the technical difficulty of identification.

[0003] Existing interference identification methods mainly rely on extracting characteristic parameters from the time-frequency domain and performing identification and classification by analyzing the frequency characteristics, time-domain variations, etc. of interference signals.

[0004] However, due to the complexity and non-linear characteristics of noise-like interference signals, these time-frequency domain-based methods often show problems of insufficient identification accuracy and poor robustness in the face of complex backgrounds. In addition, traditional methods usually only utilize single-pulse data and cannot fully exploit the deep characteristic information contained in multi-pulse data. Summary of the Invention

[0005] To solve the above problems existing in the prior art, the present invention provides a method for constructing a radar active interference classification model based on multi-dimensional image features, specifically including:

[0006] In a first aspect, the present invention provides a method for constructing an active interference database under the background of UAV swarm targets according to the obtained echo data of different interference types, where the interference types include one or more of radio frequency noise interference, noise product interference, noise amplitude modulation interference, and noise convolution interference;

[0007] According to all the echo data in the active interference database, multiple image domain data matrices are obtained;

[0008] Feature analysis and extraction are respectively performed on each image domain data matrix to obtain multiple characteristic parameters, and a characteristic parameter library of active interference data under the background of UAV swarm targets is constructed by using each characteristic parameter;

[0009] Through the characteristic parameter library of active interference data and the active interference database, a classification model established by a support vector machine (SVM) is trained to obtain a radar active interference classification model based on multi-dimensional image features.

[0010] In a second aspect, the present invention further provides a method for inputting an acquired signal to be classified into a radar active interference classification model based on multi-dimensional image features constructed by any one of the construction methods of the radar active interference classification model based on multi-dimensional image features provided in the first aspect, so as to obtain the interference type corresponding to the signal to be classified.

[0011] In a third aspect, the present invention further provides an electronic device, including a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete mutual communication through the communication bus;

[0012] The memory is used to store a computer program;

[0013] The processor is used to implement any one of the methods provided in the first aspect or the second aspect when executing the program stored on the memory.

[0014] In a fourth aspect, the present invention provides a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, any one of the methods provided in the first aspect or the second aspect is implemented.

[0015] Advantages of the present invention:

[0016] The construction method of the radar active interference classification model based on multi-dimensional image features provided by the present invention constructs an active interference database under the background of an unmanned aerial vehicle (UAV) swarm target according to the acquired echo data of different interference types; according to all the echo data in the active interference database, multiple image domain data matrices are obtained; feature analysis and extraction are respectively performed on each image domain data matrix to obtain multiple feature parameters, and a feature parameter library of active interference data under the background of the UAV swarm target is constructed by using each feature parameter; through the feature parameter library of active interference data and the active interference database, the classification model established by the support vector machine (SVM) is trained to obtain a radar active interference classification model based on multi-dimensional image features, achieving the effect of comprehensively extracting the spatio-temporal characteristics of interference signals from the image domain. By using the feature parameter library constructed under the background of the UAV swarm target and combining with the SVM classification algorithm, the noise-type active interference signals are classified and identified, which can efficiently and accurately classify the interference signals, and significantly improve the recognition accuracy and robustness of the UAV noise-type active interference.

[0017] The following will further describe the present invention in detail with reference to the drawings and embodiments. Description of the Drawings

[0018] Figure 1 It is a schematic flow chart of a construction method of a radar active interference classification model based on multi-dimensional image features provided by the present invention;

[0019] Figure 2Schematic diagram of the echo data simulation process provided by the present invention;

[0020] Figure 3 Schematic diagram of the classification result graph of active interference provided by the present invention. Specific implementation manners

[0021] The following further describes the present invention in detail with reference to specific embodiments, but the implementation manners of the present invention are not limited thereto.

[0022] Authors such as Zhang Tongbing, in their published literature "Modeling, Simulation and Feature Parameter Extraction of Radar Active Interference", aiming at the problem of insufficient recognition accuracy of traditional radar anti-jamming technology in the face of modern complex interferences (such as suppression interference, deception interference and their composite interference), proposed an interference classification method based on the extraction of time-domain and frequency-domain feature parameters. The authors carried out modeling and simulation on common interference types (such as radio frequency noise interference, noise amplitude modulation interference, dense false target interference, etc.), and extracted a variety of separable feature parameters, such as time-domain moment skewness, time-domain moment kurtosis, envelope fluctuation degree, and frequency-domain moment skewness and frequency-domain additive Gaussian white noise factor. By analyzing the separability of feature parameters under different signal-to-interference ratios, the authors further established a classification criterion based on decision trees. Experimental results show that these feature parameters can significantly improve the classification effect of interference signals within a certain signal-to-interference ratio range (10 - 20 dB). However, this interference recognition method only extracts limited feature parameters from the time-frequency domain, and the radar echo data processing also includes information such as range domain, repetition period domain, Doppler dimension, etc., and fails to make full use of the image information generated by the permutation and combination of received data.

[0023] Authors such as Wang Hong, in their published patent "Radar Interference Signal Recognition Method Based on Feature Parameter Extraction", aiming at the problems of low automation degree and low accuracy in traditional radar interference signal recognition, proposed a method for discriminating interference types by extracting time-domain and frequency-domain feature parameters. This method compares time-domain feature parameters (such as time-domain moment skewness, moment kurtosis, kurtosis coefficient) and frequency-domain feature parameters (such as frequency-domain carrier factor, normalized 3 dB bandwidth and normalized spectrum impulse part standard deviation) with the characteristics of known interference signals, and combines the D-S evidence theory to fuse the judgment results in the time-domain and frequency-domain, so as to determine the interference type. This method improves the recognition efficiency and accuracy of various interference types such as radio frequency noise interference and amplitude modulation noise interference through automatic feature extraction and discrimination. However, this traditional time-frequency domain feature parameter extraction method only aims at single-pulse data, and the information in a frame of multi-pulse data fails to be fully utilized.

[0024] Based on this, the present invention proposes a radar active interference classification method based on multi-dimensional image features, innovatively extracting features from the image domain and multi-pulse data to more comprehensively capture the spatio-temporal characteristics of interference signals. By constructing a new feature parameter set and combining efficient algorithms such as support vector machine (SVM) for classification, this method not only overcomes the limitations of time-frequency domain methods but also significantly improves the recognition accuracy of interference signals and the system robustness, providing a brand-new solution for the classification and recognition of UAV noise-like active interference.

[0025] Figure 1 The flow chart of a method for constructing a radar active interference classification model based on multi-dimensional image features provided by the present invention is as Figure 1 shown, and this method includes:

[0026] S101. According to the obtained echo data of different interference types, construct an active interference database under the background of UAV swarm targets.

[0027] Among them, the interference types include one or more of radio frequency noise interference, noise product interference, noise amplitude modulation interference, and noise convolution interference.

[0028] The echo data is usually a pulse signal, and the types of interference carried by it can generally be divided into traditional noise interference and smart noise interference, specifically radio frequency noise interference, noise product interference, noise amplitude modulation interference, and noise convolution interference.

[0029] Exemplarily, the echo data of different interference types can be obtained through simulation. Specifically as Figure 2 shown, it includes the following steps A1 - A8:

[0030] A1. Start: Initiate the simulation process.

[0031] A2. Import of simulation parameter set: Import the parameter set required for simulation, including radar parameters, target characteristics, environmental settings, etc.

[0032] A3. Calculation of UAV swarm target information: According to the imported parameters, calculate the target information of the UAV swarm, such as position, speed, heading, etc.

[0033] A4. Generation of target echo data: Based on the calculated target information, generate the target echo data of the UAV swarm, simulating the target reflection signal received by the radar.

[0034] A5. Generation of interference echo data: Generate interference echo data, simulating possible noise, clutter, or artificial interference signals.

[0035] A6. Fusion of echo data: Fusion the target echo data and the interference echo data to form a complete echo data set containing the target and interference.

[0036] A7. Echo data saving: Save the fused echo data for subsequent analysis and processing.

[0037] A8. End: Complete the simulation process.

[0038] S102. Obtain multiple image domain data matrices based on all echo data in the active interference database.

[0039] Optionally, the image domain data matrix includes: fast time - period diagram, fast time - Doppler diagram, frequency - period diagram, frequency - Doppler diagram, range - period diagram, and range - Doppler diagram.

[0040] Further, in a possible implementation manner, obtaining the fast time - period diagram based on all echoes in the active interference database includes: performing fast time sampling on all echo data corresponding to each interference type in the active interference database to obtain M sampling points corresponding to each echo data, where M is a positive integer greater than or equal to 1; using the pulse period as the column vector and the sampling time as the row vector, and constructing the fast time - period diagram corresponding to each interference type respectively according to the sampling points of all echo data corresponding to each interference type. The fast time - period diagram is a matrix with P rows and M columns, and P is the total number of echo data corresponding to any interference type of the echo data.

[0041] It can be understood that one radar echo data corresponds to one communication cycle. Correspondingly, the total number of echo data corresponding to any interference type is the number of cycles corresponding to that interference type.

[0042] Further, in a possible implementation manner, obtaining the fast time - Doppler diagram based on all echo data in the active interference database includes: performing fast Fourier transform (FFT) on each column of the fast time - period diagram to obtain the fast time - Doppler diagram.

[0043] Further, in a possible implementation manner, obtaining the frequency - Doppler diagram based on all radar active interference signal data in the active interference database includes: performing FFT on each row of the fast time - period diagram to obtain the frequency - Doppler diagram.

[0044] Further, in a possible implementation manner, obtaining the frequency - period diagram based on all radar active interference signal data in the active interference database includes: performing FFT on each column of the frequency - Doppler diagram to obtain the frequency - period diagram.

[0045] Further, in a possible implementation, based on all the radar active interference signal data in the active interference database, a range-period diagram and a range-Doppler diagram are obtained, including: performing pulse compression on the data in the fast time-period diagram to obtain the range-period diagram; performing FFT transformation on each column of the range-period diagram to obtain the range-Doppler diagram.

[0046] Specifically, after pulse compression of the radar received data for the entire CPI, then arranging every M points into a row (meaning ranging for single-period data), and arranging them in sequence by period, an RP matrix, that is, the range-period diagram, is obtained.

[0047] The present invention realizes the effect of comprehensively extracting the spatio-temporal characteristics of interference signals from the image domain by arranging and combining the multi-pulse data received by the radar to construct image domain data matrices, including the fast time-period diagram, the fast time-Doppler diagram, the frequency-period diagram, the frequency-Doppler diagram, the range-period diagram, and the range-Doppler diagram. By arranging and transforming the multi-pulse data, converting it into multiple image matrices, and extracting multi-dimensional characteristic parameters in the image domain, the potential of the multi-pulse data is fully exploited.

[0048] S103. Respectively perform feature analysis and extraction on each image domain data matrix to obtain multiple characteristic parameters, and use each characteristic parameter to construct a characteristic parameter library of the active interference data under the background of the UAV swarm target.

[0049] Optionally, the characteristic parameters include: the correlation of the 90-degree texture image of the range trajectory, the correlation of the 135-degree texture image of the range trajectory, the sparsity of the range trajectory image, the correlation of the 45-degree texture image of the spectrum, the contrast moment of the 0-degree texture image of the spectrum, the variance of the spectrum image, the correlation of the 90-degree texture image of the spectrum, the correlation of the 135-degree texture image of the spectrum, the contrast moment of the 0-degree texture image of the range trajectory, the contrast moment of the 45-degree texture image of the spectrum, the correlation of the 0-degree texture image of the spectrum, and the contrast of the 0-degree texture image of the spectrum.

[0050] In a possible implementation, using each characteristic parameter to construct a characteristic parameter library of the active interference data under the background of the UAV swarm target includes: extracting the correlation of the 90-degree texture image of the range trajectory, the correlation of the 135-degree texture image of the range trajectory, and the sparsity of the range trajectory image from the range-period diagram; extracting the contrast moment of the 0-degree texture image of the range trajectory from the range-Doppler diagram; extracting the correlation of the 90-degree texture image of the spectrum, the correlation of the 0-degree texture image of the spectrum, the variance of the spectrum image, and the contrast moment of the 0-degree texture image of the spectrum from the frequency-period diagram; extracting the correlation of the 45-degree texture image of the spectrum, the correlation of the 90-degree texture image of the spectrum, the correlation of the 135-degree texture image of the spectrum, the contrast moment of the 45-degree texture image of the spectrum, and the contrast of the 0-degree texture image of the spectrum from the frequency-Doppler diagram.

[0051] The present invention comprehensively captures the spatio-temporal characteristics of interference signals by constructing image domain data matrices such as fast time-period diagrams, fast time-Doppler diagrams, frequency-period diagrams, frequency-Doppler diagrams, range-period diagrams, and range-Doppler diagrams, significantly improving the integrity of feature extraction.

[0052] S104. Train the classification model established by the support vector machine (SVM) through the feature parameter library and active interference database of active interference data to obtain a radar active interference classification model based on multi-dimensional image features.

[0053] Specifically, use the feature parameter library of active interference data as the training set and the active interference database as the test set to train the classification model established by the support vector machine (SVM) to obtain a radar active interference classification model based on multi-dimensional image features.

[0054] The SVM optimizes the hyperplane to achieve maximum margin classification of different types of interference signals in the high-dimensional feature space, making the classification results have high accuracy and robustness, thus providing guarantee for the accurate identification of UAV noise-like active interference.

[0055] The method for constructing a radar active interference classification model based on multi-dimensional image features provided by the present invention includes: constructing an active interference database under the background of UAV swarm targets according to the acquired echo data of different interference types; obtaining a variety of image domain data matrices according to all the echo data in the active interference database; respectively performing feature analysis and extraction on each image domain data matrix to obtain multiple feature parameters, and using each feature parameter to construct a feature parameter library of active interference data under the background of UAV swarm targets; training the classification model established by the support vector machine (SVM) through the feature parameter library and active interference database of active interference data to obtain a radar active interference classification model based on multi-dimensional image features, achieving the effect of comprehensively extracting the spatio-temporal characteristics of interference signals from the image domain, using the feature parameter library constructed under the background of UAV swarm targets, combining with the SVM classification algorithm, classifying and identifying noise-like active interference signals, and being able to efficiently achieve accurate classification of interference signals, significantly improving the recognition accuracy and robustness of UAV noise-like active interference.

[0056] The present invention also provides a radar active interference classification method based on multi-dimensional image features, including: inputting the acquired signal to be classified into the constructed radar active interference classification model based on multi-dimensional image features to obtain the interference type corresponding to the echo data to be classified, where the radar active interference classification model based on multi-dimensional image features is constructed by any one of the methods for constructing a radar active interference classification model based on multi-dimensional image features provided in the above embodiments.

[0057] Figure 3Schematic diagram of a classification result graph for active interference provided by the present invention, as Figure 3 shown, there are 1500 groups of each of the four types of noise - type active interferences. After training the classification model established by the support vector machine (SVM) with the constructed characteristic parameters, the obtained classification results can accurately distinguish the four types of noise - type active interference types. Figure 3 What is shown is a confusion matrix, which is usually used to evaluate the performance of a classification model. Among them, the abscissa "Predicted Class" represents the predicted class, corresponding to the class predicted by the model, and it shows the result predicted by the model, that is, the number of samples judged by the model to be a certain class; the ordinate "True Class" represents the true class, corresponding to the actual class label, that is, the true class of the sample. In each square in the figure, the number represents the number of samples at the intersection of the corresponding true class and predicted class. For example, the number "1495" in the upper - left corner indicates that 1495 samples with a true class of 1 are correctly predicted as 1. And the "5" in the upper - right corner indicates that 5 samples with a true class of 1 are wrongly predicted as 4.

[0058] Specifically, the radar active - interference classification model based on multi - dimensional image features can be applied to the subsequent processing of multi - dimensional radar signal feature extraction and classification. For example: in a radar interference signal processing system based on the image domain, this model is directly placed after the image - domain data matrix generation module. Feature analysis and extraction are performed on the six generated image - domain data matrices (fast - time - periodogram, fast - time - Dopplergram, frequency - periodogram, frequency - Dopplergram, range - periodogram, and range - Dopplergram) to generate a feature parameter library containing multi - dimensional feature parameters, and then the extracted features are input into the support vector machine (SVM) to complete the interference classification operation.

[0059] For another example, in a specific interference signal analysis system, this model can be directly inserted into the main part of the feature analysis to further optimize the feature parameters generated at different stages. The specific implementation method is as follows: Feature parameter library optimization: The multi - dimensional feature parameters extracted from the six image matrices are input into the module, and the module optimizes and screens the feature parameters to improve the training effect of the classifier; Classification result enhancement: The intermediate result of the classifier is input into this module for feature fusion to generate a more reliable classification probability output, which is fed back to the classifier to complete the accurate discrimination of the interference type.

[0060] For another example, this model can also be used as a standardized feature extraction and classification module and embedded into other interference classification networks based on radar signal processing. For example, by combining with existing time - frequency domain feature extraction methods, cross - domain fusion processing is performed on the extracted features to improve the classification and recognition effect of UAV noise - type active interferences.

[0061] The present invention also provides a structure of an electronic device, including a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete communication with each other through the communication bus.

[0062] The memory is used to store a computer program.

[0063] The processor is used to implement the steps provided in the above method embodiments when executing the program stored on the memory.

[0064] The communication interface is used for communication between the above electronic device and other devices.

[0065] The method provided by the embodiments of the present invention can be applied to an electronic device. Specifically, the electronic device can be: a desktop computer, a portable computer, a smart mobile terminal, a server, etc. There is no limitation here. Any electronic device that can implement the present invention belongs to the protection scope of the present invention.

[0066] The present invention also provides a computer-readable storage medium. A computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the steps provided in the above method embodiments are implemented.

[0067] For the embodiments of the electronic device / storage medium / program product, since they are basically similar to the method embodiments, the description is relatively simple. For the specific content, beneficial effects, and other related aspects, please refer to the partial description of the method embodiments.

[0068] The above content is a further detailed description of the present invention in combination with specific preferred implementation manners. It cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention belongs, without departing from the concept of the present invention, several simple deductions or substitutions can still be made, and all should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for constructing a radar active jammer classification model based on multi-dimensional image features, characterized in that: include: According to the acquired echo data of different interference types, an active interference database under the target background of the drone group is constructed, wherein the interference types include one or more of radio frequency noise interference, noise product interference, noise amplitude modulation interference and noise convolution interference; Obtaining multiple image domain data matrices according to all echo data in the active interference database; Performing feature analysis and extraction on each of the image domain data matrices respectively to obtain a plurality of feature parameters, and using each of the feature parameters to construct a feature parameter library of active interference data under the target background of the drone group; The classification model established by the support vector machine (SVM) is trained through the characteristic parameter library of the active interference data and the active interference database to obtain a radar active interference classification model based on multi-dimensional image features.

2. The method according to claim 1, characterized in that The image domain data matrix includes: a fast time-period diagram, a fast time-Doppler diagram, a frequency-period diagram, a frequency-Doppler diagram, a distance-period diagram and a distance-Doppler diagram.

3. The method according to claim 2, characterized in that According to all the echoes in the active interference database, a fast time-period diagram is obtained, including: Perform fast time sampling on all echo data corresponding to each interference type in the active interference database to obtain M sampling points corresponding to each echo data, where M is a positive integer greater than or equal to 1; Taking the pulse period as the column vector and the sampling time as the row vector, a fast time-period diagram corresponding to each interference type is constructed according to the sampling points of all echo data corresponding to each interference type. The fast time-period diagram is a matrix of P rows and M columns, where P is the total number of echo data corresponding to any interference type of the echo data.

4. The method according to claim 3, characterized in that According to all echo data in the active interference database, the fast time-Doppler map, the frequency-Doppler map and the frequency-period map are obtained, including: performing a fast Fourier transform (FFT) on each column of the fast time-period map to obtain the fast time-Doppler map; Performing FFT transformation on each row of the fast time-period diagram to obtain the frequency-Doppler diagram; Performing FFT transformation on each column of the frequency-Doppler diagram to obtain the frequency-period diagram.

5. The method according to claim 4, characterized in that According to all radar active jamming signal data in the active jamming database, the range-period graph and the range-Doppler graph are obtained, including: Performing pulse compression on the data in the fast time-period diagram to obtain the distance-period diagram; Performing FFT transformation on each column of the range-period diagram to obtain the range-Doppler diagram.

6. The method according to any one of claims 2 to 5, characterized in that: The characteristic parameters include: distance track 90 degree texture image correlation, distance track 135 degree texture image correlation, distance track image sparsity, spectrum 45 degree texture image correlation, spectrum 0 degree texture image contrast moment, spectrum image variance, spectrum 90 degree texture image correlation, spectrum 135 degree texture image correlation, distance track 0 degree texture image contrast moment, spectrum 45 degree texture image contrast moment, spectrum 0 degree texture image correlation and spectrum 0 degree texture image contrast.

7. The method according to claim 6, characterized in that The method of using the characteristic parameters to construct a characteristic parameter library of active interference data under the target background of the drone group includes: Extracting the distance track 90 degree texture image correlation, the distance track 135 degree texture image correlation and the distance track image sparsity from the distance-period graph; Extracting the range track 0 degree texture image contrast moment from the range-Doppler image; Extracting the 90-degree spectrum texture image correlation, the 0-degree spectrum texture image correlation, the spectrum image variance and the 0-degree spectrum texture image contrast moment from the frequency-periodogram; The spectrum 45 degree texture image correlation, the spectrum 90 degree texture image correlation, the spectrum 135 degree texture image correlation, the spectrum 45 degree texture image contrast moment and the spectrum 0 degree texture image contrast are extracted from the frequency-Doppler image.

8. A radar active jammer classification method based on multi-dimensional image features, characterized in that: include: The acquired signal to be classified is input into the radar active interference classification model based on multidimensional image features constructed by the method for constructing a radar active interference classification model based on multidimensional image features as described in any one of claims 1 to 7 to obtain the interference type corresponding to the signal to be classified.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, for implementing any of the methods described in claims 1-8 when executing a program stored in a memory.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images