A collaborative electromagnetic signal intelligent recognition method and system

Abstract

The present invention provides a collaborative electromagnetic signal intelligent identification method and system, which belongs to the technical field of electromagnetic signal analysis and processing. The collaborative sensing node and the collaborative identification controller cooperate with each other. The collaborative sensing node is responsible for air interface electromagnetic signal reception, preprocessing, time-frequency analysis and potential electromagnetic target identification and uses machine learning classification methods to identify the current scene; the collaborative identification controller is responsible for collecting the electromagnetic target list reported by the subordinate collaborative sensing nodes and comprehensively judging and weighting the status, time-frequency position consistency, and graphic morphology consistency of each collaborative sensing node, thereby improving the accuracy of the electromagnetic identification results and the scene adaptation ability.

Description

Technical Field

[0001] The present invention belongs to the technical field of electromagnetic signal analysis and processing, and specifically relates to a collaborative electromagnetic signal intelligent recognition method and system. Background Art

[0002] Electromagnetic signal recognition is a signal detection technology used to obtain the use of wireless spectrum at a specific time, place and frequency band. According to different recognition principles, electromagnetic signal recognition methods can be mainly summarized into the following types (including but not limited to): 1. Traditional statistical electromagnetic recognition methods; 2. Traditional machine learning electromagnetic recognition methods; 3. Deep learning electromagnetic recognition methods. Among them, traditional statistical electromagnetic recognition methods belong to early electromagnetic signal recognition methods, which are mainly represented by energy detection, cyclostationary feature detection, matched filter detection and covariance matrix detection methods. A big problem with this type of method is that its detection accuracy and false alarm rate and other indicators deteriorate significantly under low signal-to-interference ratio conditions. With the development of technology, data-driven processing technologies represented by machine learning have greatly improved the ability to analyze and process electromagnetic spectrum data. These methods are based on multi-dimensional features of signals in the frequency domain and time domain, such as high-order cumulants, cyclic spectra, and constellation diagrams. They use decision trees, support vector machines, etc. to perform pattern classification and correlation analysis on the received signals. They can achieve significant performance improvements by training with only a small amount of data sets. However, there are still problems, mainly reflected in the fact that the above-mentioned machine learning methods rely on experts to predefine features and pre-process feature parameters, require rich professional knowledge and experience, and have a limited scope of application. In addition, in complex electromagnetic environments, predefined feature parameters are difficult to adapt to rapid environmental changes and have poor recognition accuracy.

[0003] In recent years, with the further development of technology, deep learning electromagnetic recognition methods have provided a new way for electromagnetic signal recognition and control with their excellent feature extraction and expression capabilities, and have achieved performance and efficiency that surpass traditional algorithms. They can automatically learn and extract features from raw electromagnetic signal data without manually defining the extracted features, and their performance and efficiency have been greatly improved compared to traditional methods. However, the current deep learning electromagnetic recognition methods still have the following problems: 1. Insufficient scene adaptation capability: The electromagnetic spectrum environment is an open space where wireless terminals, communication systems, radars, jammers and other equipment alternately release high-density, high-intensity, multi-spectrum electromagnetic signals. The composition of the received signal is becoming increasingly complex and highly dynamic. 1. Once the existing deep learning perception models complete the training process, the network structure and hyperparameters will no longer change, and the algorithm performance and application are limited; 2. Performance degradation under low SNR (signal-to-noise ratio) conditions: Studies have shown that deep learning-based algorithms have excellent perception performance under high SNR conditions, but the performance will be greatly deteriorated under low SNR conditions; in particular, due to factors such as spatial selective fading and hidden terminals, the dynamic range of the SNR of a single-point received signal is large, and the detection performance is greatly reduced; 3. Problems of small samples and poor generalization: Deep learning-based methods often require massive signal samples for model training. However, in actual scenarios, due to limitations in information collection capabilities, storage capabilities, and sample labeling capabilities, it is difficult for nodes to obtain global, multi-scale, and multi-perspective signal samples, resulting in weak robustness and poor generalization.

[0004] In view of this, the prior art needs to be further improved. Summary of the invention

[0005] In view of the above-mentioned problems existing in the prior art, the present invention provides a collaborative electromagnetic signal intelligent recognition method and system to solve the problems of insufficient scene adaptability, poor recognition performance under low SNR conditions, and weak generalization ability in the current electromagnetic signal recognition method based on deep learning models.

[0006] In order to solve the above technical problems, the present invention is implemented by the following technical solutions:

[0007] A collaborative electromagnetic signal intelligent recognition method is implemented by means of a multi-point collaborative intelligent electromagnetic recognition system, wherein the multi-point collaborative intelligent electromagnetic recognition system comprises at least a collaborative sensing node and a collaborative recognition controller capable of information exchange, and is characterized in that the method comprises the following steps:

[0008] Step S1, the collaborative sensing node collects and preprocesses electromagnetic signals to obtain original electromagnetic signals;

[0009] Step S2: The collaborative sensing node performs parameter detection on the original electromagnetic signal obtained in step S1 to obtain detection parameters, and obtains a corresponding scene mode based on the detection parameters;

[0010] Step S3, constructing corresponding training data sets and test data sets according to the scene mode obtained in step S2;

[0011] Step S4: Based on the training data set and the test data set constructed in step S3, the collaborative sensing node converts the original electromagnetic signal obtained in step S1 into a two-dimensional time-frequency diagram using a time-frequency analysis method;

[0012] Step S5: The collaborative sensing node searches for target detection network model parameters corresponding to the scene mode from its model pool according to the scene mode, and obtains a trained target detection network model corresponding to the model parameters based on the model parameters;

[0013] Step S6, the collaborative sensing node inputs the two-dimensional time-frequency graph obtained in step S4 into the target detection network model for detection, and obtains potential electromagnetic target information after processing;

[0014] Step S7: The collaborative sensing node transmits each potential electromagnetic target information and its corresponding SNR information to the collaborative identification controller to construct a potential electromagnetic target information matrix and a node status information table;

[0015] Step S8: Based on the constructed potential electromagnetic target information matrix and node status information table, the collaborative identification controller comprehensively considers the status of each collaborative sensing node, performs multi-information weighted matching fusion identification processing on the potential electromagnetic target list reported by each collaborative sensing node, obtains the final result of the electromagnetic identification target, and forwards it to other modules for subsequent processing.

[0016] Furthermore, the multi-information weighted matching fusion recognition process in step S8 includes the following steps:

[0017] S81, in descending order, the current node status value of the collaborative sensing node Sort by The highest collaborative sensing node is the anchor node, denoted as anchor node M;

[0018] S82, scan the potential electromagnetic target list of the anchor node M, and record the potential electromagnetic target information of the anchor node M as ,in, , N is the total number of potential electromagnetic target information scanned at the anchor node M;

[0019] S83, for the anchor node M i Potential electromagnetic target information , and its target center is recorded as , the time and frequency coordinates corresponding to the target center are marked as and , and Uniquely determined by x and y respectively, set the corresponding time range and frequency range ,in, To characterize the time difference of each sensing node receiving the same signal, it is related to the interval distance between each sensing node and is preset globally. The frequency difference related to the receiver frequency difference and Doppler shift factor is preset as needed;

[0020] S84, the first potential electromagnetic target information from the anchor node M At the beginning, the N potential electromagnetic target information of the anchor node M is processed in turn until the information processing is completed.

[0021] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0022] 1. Aiming at a series of problems in current electromagnetic signal identification, the present invention proposes a collaborative electromagnetic signal intelligent identification method and system. The system is mainly composed of two entities: collaborative sensing nodes and collaborative identification controllers. The collaborative sensing nodes are responsible for air interface electromagnetic signal reception, preprocessing, time-frequency analysis and potential electromagnetic target identification; the collaborative identification controller is responsible for collecting the electromagnetic target list reported by the subordinate collaborative sensing nodes, and then adopts multi-information weighted matching to obtain the final result of electromagnetic identification target; by fusing the potential electromagnetic target list reported by each collaborative sensing node, the method comprehensively considers the state of each collaborative sensing node, the consistency of time-frequency position, the consistency of graphic morphology, etc., and performs weighted processing, thereby improving the accuracy of the identification result;

[0023] 2. The present invention uses collaborative sensing nodes to perform real-time signal acquisition and energy detection, signal-to-noise ratio detection and cyclic power spectrum detection calculations, uses machine learning classification methods to identify the current scene, searches for the target detection network model parameters corresponding to the mode from the model pool, and completes the model construction based on the model parameters, thereby improving the scene adaptation ability of the model;

[0024] 3. Multiple collaborative sensing nodes identify potential electromagnetic targets through two-dimensional time-frequency analysis and based on the target detection network model, and obtain independent recognition performance under different channel conditions, thus obtaining spatial diversity benefits;

[0025] 4. In addition, according to the present invention, the collaborative sensing node transmits each potential electromagnetic target information to the collaborative identification controller, and the collaborative identification controller makes a comprehensive judgment, avoiding the existing technology of directly aggregating source data to the fusion center for training and reasoning, which leads to high communication overhead, poor privacy and other problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0027] Figure 1 It is a schematic diagram of the overall architecture of the multi-point cooperative intelligent electromagnetic identification system of the present invention;

[0028] Figure 2 It is a flow chart of the present invention for performing multi-point cooperative intelligent electromagnetic identification;

[0029] Figure 3 It is a schematic diagram of the principle of constructing a data set in the present invention;

[0030] Figure 4 It is a schematic diagram of the principle of information content reporting by the collaborative sensing node of the present invention;

[0031] Figure 5 It is a flow chart of electromagnetic target fusion recognition based on multi-information weighted matching of the present invention;

[0032] Figure 6 It is a schematic diagram of the principle of similarity calculation for potential electromagnetic targets in the present invention. DETAILED DESCRIPTION

[0033] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0034] like Figures 1 to 6 As shown, the present invention provides a collaborative electromagnetic signal intelligent recognition method, which is implemented by means of a multi-point collaborative intelligent electromagnetic recognition system, wherein the multi-point collaborative intelligent electromagnetic recognition system at least includes a collaborative sensing node and a collaborative recognition controller capable of information exchange, and specifically includes the following steps:

[0035] Step S1: The collaborative sensing node collects and preprocesses electromagnetic signals to obtain original electromagnetic signals, wherein the preprocessing step mainly includes filtering, down-conversion, and A / D conversion. Preferably, in this step S1, a digital sequence is formed by preprocessing. , where n represents the number of sampling points, which is a positive integer;

[0036] Step S2: The collaborative sensing node performs parameter detection on the original electromagnetic signal obtained in step S1 to obtain energy detection parameters. , signal-to-noise ratio detection parameters and cyclic power spectrum detection parameters Then, the scene mode corresponding to the above parameters is obtained by using machine learning classification methods such as SVM ,Right now

[0037] ,

[0038] In the formula, is a classification function based on machine learning classification methods such as SVM;

[0039] Step S3: scene mode obtained according to step S2 , construct the corresponding training data set and test data set; the test data set is stored locally in the collaborative sensing node for real-time electromagnetic signal recognition; according to different learning methods, the training set data can be uploaded to the collaborative controller for unified training, or it can be retained locally in the collaborative sensing node and the federated learning mode can be used for model training;

[0040] Step S4: Based on the training data set and the test data set constructed in step S3, the collaborative sensing node converts the original electromagnetic signal obtained in step S1 into a two-dimensional time-frequency diagram using time-frequency analysis methods such as short-time Fourier transform and wavelet transform; in this step S4, the time-frequency analysis method of short-time Fourier transform is preferably used, and its expression is:

[0041] ,

[0042] In the formula, is the DFT (discrete Fourier transform) of the windowed data at time pR; g(n) is the window function of length P; p is the length of each segment in the window, f is the frequency, j is the imaginary unit, R is the size of each DFT jump, which is equal to the window size minus the overlap length L, that is, R = PL;

[0043] It should be noted that in order to ensure the convenience of subsequent processing, each collaborative sensing node and collaborative identification controller needs to be clock synchronized. Therefore, the time and frequency values ​​corresponding to the pixels at the same position in the obtained time-frequency graph are consistent;

[0044] Step S5: The collaborative sensing node detects the scene mode , search from its model pool to obtain the scene model Corresponding target detection network model (such as YOLO) parameters, and obtaining a trained target detection network model corresponding to the model parameters based on the model parameters;

[0045] Step S6: The collaborative sensing node inputs the two-dimensional time-frequency graph obtained in step S4 into the target detection network model for detection to obtain a potential electromagnetic target parameter unit, wherein each potential electromagnetic target parameter unit can be expressed as [c, x, y, w, h], c is the confidence, x and y are the center coordinates of the corresponding potential electromagnetic target, and w and h are the width and height of the corresponding potential electromagnetic target; based on the area circled by the potential electromagnetic target parameter unit [c, x, y, w, h], the collaborative sensing node uses the two-dimensional time-frequency graph to calculate the average RSSI (received signal strength indication) corresponding to the potential electromagnetic target, which is calculated as , whose expression is:

[0046] ,

[0047] Where t is time, f is the frequency (normalized accordingly, dimensionless, and will not be described here);

[0048] Thus, the potential electromagnetic target information [c, x, y, w, h, p (mean)] associated with each potential electromagnetic target parameter unit is obtained;

[0049] Step S7: The collaborative sensing node sends each potential electromagnetic target information [c, x, y, w, h, p (mean)] and its corresponding SNR information (i.e., signal-to-noise ratio detection parameter ) are transmitted to the collaborative identification controller, which collects the potential electromagnetic target information [c, x, y, w, h, p (mean)] and SNR information reported by each collaborative sensing node in real time, and constructs the potential electromagnetic target information matrix and node status information table;

[0050] Preferably, see Figure 4In step S7, the information content reported by each collaborative sensing node is reflected in the form of a potential electromagnetic target list, which includes, in addition to the potential electromagnetic target information [c, x, y, w, h, p (mean)] and SNR information, node id, target matching verification mode, number of potential electromagnetic targets and screenshots of electromagnetic target information corresponding to the potential electromagnetic target information [c, x, y, w, h, p (mean)], etc.; wherein, the target matching verification mode can use the screenshots of electromagnetic target information to accurately verify whether the electromagnetic targets detected by two collaborative sensing nodes are the same target. The target matching verification mode has multiple states, which are represented in binary form, for example, 001 is the first state, 010 is the second state, 011 is the third state, and so on. In addition, see Figure 4 , to coordinate sensing nodes H For example ( H is a positive integer greater than or equal to 1), and the reported K The potential electromagnetic target information is expressed as ;

[0051] Preferably, the node status information table is composed of historical SNR information and the node status value calculated comprehensively. It can be expressed as , where the subscript H Indicates that the H The node status information table of the collaborative sensing nodes, Indicates a point in time, for example, Indicates time point The obtained SNR information, Indicates time point The obtained SNR information, and so on. Node status value The calculation is based on this time point The weighted calculation of the first A SNR values ​​is obtained. , the value of A can be preset as needed.

[0052] Step S8: Based on the constructed potential electromagnetic target information matrix and node status information table, the collaborative identification controller comprehensively considers the status of each collaborative sensing node, performs multi-information weighted matching fusion identification processing on the potential electromagnetic target list reported by each collaborative sensing node, obtains the final result of the electromagnetic identification target, and forwards it to other modules for subsequent processing.

[0053] In order to better achieve the purpose of the present invention, the multi-information weighted matching fusion recognition process in step S8 includes the following steps:

[0054] S81, in descending order, the current node status value of the collaborative sensing node Sort by The highest collaborative sensing node is the anchor node, denoted as anchor node M;

[0055] S82, scan the potential electromagnetic target list of the anchor node M, and record the potential electromagnetic target information of the anchor node M as ,in, , N is the total number of potential electromagnetic target information scanned at the anchor node M;

[0056] S83, for the anchor node M i Potential electromagnetic target information , and its target center is recorded as (Coordinates are ), the time and frequency coordinates corresponding to the target center are marked as and , and Uniquely determined by x and y respectively, set the corresponding time range and frequency range ,in, To characterize the time difference of each sensing node receiving the same signal, which is related to the interval distance between each sensing node and can be globally preset; The frequency difference is related to factors such as receiver frequency difference and Doppler frequency shift, which can be preset as needed; the first potential electromagnetic target information of anchor node M is For example, the corresponding time range is set as , the frequency range is ;

[0057] S84, the first potential electromagnetic target information from the anchor node M At the beginning, the N potential electromagnetic target information of the anchor node M is processed in turn until the information processing is completed, specifically:

[0058] S841, for i Potential electromagnetic target information , with its target center To anchor the center point, in the time range and frequency range Within, traverse the potential electromagnetic target list of other cooperative sensing nodes, if it is found that the center of the potential electromagnetic target of the other cooperative sensing node falls within the time range and frequency range Inside, and its Greater than a predetermined threshold , it is considered that the current potential electromagnetic target identified by the other collaborative sensing nodes is consistent with the corresponding first iPotential electromagnetic targets may be the same target, so they are recorded as similar potential electromagnetic targets;

[0059] S842: All detected nodes that are connected to the anchor node M i All similar potential electromagnetic targets similar to the potential electromagnetic targets are constructed into a new information set X, and the electromagnetic target is selectively identified by an agile method or a precise method according to the target matching verification mode to obtain the final result of the electromagnetic identification target;

[0060] The information set X can be expressed as:

[0061] ;

[0062] Specifically, the agile method does not consider the time-frequency diagram corresponding to the electromagnetic target. It is directly based on the average power weighting of the target area and obtains the final result of electromagnetic identification target through the following calculation formula:

[0063] ;

[0064] The precise method considers the time-frequency graph corresponding to the electromagnetic target, which requires first i The time-frequency screenshots corresponding to the potential electromagnetic targets and the time-frequency screenshots corresponding to each similar potential electromagnetic target in the set X are input into a CNN network to calculate the graphic similarity and obtain the corresponding similarity factor , ,…… etc., see Figure 6 ; Then, based on the average power and similarity factor of the target area, the final confidence of the electromagnetic identification target is obtained by comprehensive weighting through the following calculation formula: , target final center coordinates and , target final width , target final height The results are as follows:

[0065] ;

[0066] The final result of electromagnetic identification target is expressed as .

[0067] It should be noted that the present invention mainly realizes two functions through the specific steps of multi-information weighted matching fusion recognition processing in step S8 above, namely: 1. Confirm whether the potential electromagnetic targets reported by each collaborative sensing node are the same target, so as to ensure the accuracy of the recognition result; 2. For a certain target, the optimal judgment result is obtained by integrating the reported information of each collaborative sensing node. The selection of the precise method and the agile method is pre-set by the state of the target matching verification mode.

[0068] Further, in step S84, i After the information of potential electromagnetic targets is processed, if there are still potential electromagnetic targets left, i=i+1 , continue to complete the identification and processing of the next potential electromagnetic target.

[0069] In a further preferred embodiment, the multi-point cooperative intelligent electromagnetic identification system of the present invention can adopt a centralized architecture as the entire system architecture. Figure 1 As shown; a distributed architecture can also be used, in which case the collaborative identification controller can be in the same physical entity as a collaborative sensing node, and is generated by default or election mode when the system starts. Among them, the collaborative sensing node is responsible for receiving and preprocessing air interface electromagnetic signals; performing calculations and state analysis such as RSSI and SNR, and calling a suitable deep learning model for reasoning; and reporting the inferred potential electromagnetic target results (composed of multiple potential electromagnetic target distribution maps with a two-dimensional time-frequency structure) and status information to the collaborative identification controller. The collaborative identification controller is responsible for collecting the electromagnetic target list reported by the subordinate collaborative sensing nodes, using a confidence-constrained electromagnetic target fusion recognition algorithm to determine the electromagnetic target at each time-frequency point, and forwarding it to modules such as illegal target warning for processing. In an example, see Figure 1 Each collaborative sensing node includes at least a signal acquisition unit, a state analysis unit, a model configuration unit, a model pool, an inference model unit and a first information interaction unit. Each unit is used to implement the corresponding function of the above method. For example, the model pool is used to Search to get the scene mode The corresponding target detection network model (such as YOLO) parameters are obtained and the corresponding trained target detection network model is obtained; the collaborative recognition controller at least includes a potential electromagnetic target list unit, a node status information table unit, a second information interaction unit and a fusion perception decision unit. The first information interaction unit communicates with the second information interaction unit, and the fusion perception decision unit is used to perform multi-information weighted matching fusion recognition processing on the potential electromagnetic target list reported by each collaborative perception node based on the constructed potential electromagnetic target information matrix and the node status information table, comprehensively consider the status of each collaborative perception node, and output the final result of the electromagnetic recognition target.

[0070] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A collaborative electromagnetic signal intelligent recognition method, which is implemented by means of a multi-point collaborative intelligent electromagnetic recognition system, wherein the multi-point collaborative intelligent electromagnetic recognition system at least includes a collaborative sensing node and a collaborative recognition controller capable of information exchange, and is characterized in that: The method comprises the following steps: Step S1, the collaborative sensing node collects and preprocesses electromagnetic signals to obtain original electromagnetic signals; Step S2: The collaborative sensing node performs parameter detection on the original electromagnetic signal obtained in step S1 to obtain detection parameters, and obtains a corresponding scene mode based on the detection parameters; Step S3, constructing corresponding training data sets and test data sets according to the scene mode obtained in step S2; Step S4: Based on the training data set and the test data set constructed in step S3, the collaborative sensing node converts the original electromagnetic signal obtained in step S1 into a two-dimensional time-frequency diagram using a time-frequency analysis method; Step S5: The collaborative sensing node searches for target detection network model parameters corresponding to the scene mode from its model pool according to the scene mode, and obtains a trained target detection network model corresponding to the model parameters based on the model parameters; Step S6, the collaborative sensing node inputs the two-dimensional time-frequency graph obtained in step S4 into the target detection network model for detection, and obtains potential electromagnetic target information after processing; Step S7: The collaborative sensing node transmits each potential electromagnetic target information and its corresponding SNR information to the collaborative identification controller to construct a potential electromagnetic target information matrix and a node status information table; Step S8: Based on the constructed potential electromagnetic target information matrix and node status information table, the collaborative identification controller comprehensively considers the status of each collaborative sensing node, performs multi-information weighted matching fusion identification processing on the potential electromagnetic target list reported by each collaborative sensing node, obtains the final result of the electromagnetic identification target, and forwards it to other modules for subsequent processing.

2. A collaborative electromagnetic signal intelligent recognition method as claimed in claim 1, characterized in that: In step S2, the detection parameters obtained include energy detection parameters , signal-to-noise ratio detection parameters and cyclic power spectrum detection parameters .

3. A collaborative electromagnetic signal intelligent recognition method as claimed in claim 1, characterized in that: In step S3, the data set to be tested is stored locally in the collaborative sensing node for real-time electromagnetic signal recognition; the training set data is uploaded to the collaborative controller for unified training or retained locally in the collaborative sensing node and model training is performed using a federated learning model.

4. A collaborative electromagnetic signal intelligent recognition method as claimed in claim 1, characterized in that: In step S6, the processing includes obtaining a potential electromagnetic target parameter unit after detection using the target detection network model, each potential electromagnetic target parameter unit can be expressed as [c, x, y, w, h], c is the confidence, x and y are the center coordinates of the corresponding potential electromagnetic target, and w and h are the width and height of the corresponding potential electromagnetic target; based on the area circled by the potential electromagnetic target parameter unit [c, x, y, w, h], the collaborative sensing node uses a two-dimensional time-frequency diagram to calculate the average RSSI corresponding to the corresponding potential electromagnetic target, which is calculated as , whose expression is: ， Where t is time, f is the frequency; In this way, the potential electromagnetic target information [c, x, y, w, h, p (mean)] associated with each potential electromagnetic target parameter unit is obtained.

5. A collaborative electromagnetic signal intelligent recognition method as claimed in claim 4, characterized in that: In step S7, the information content reported by each collaborative sensing node is reflected in the form of a potential electromagnetic target list, which includes not only the potential electromagnetic target information [c, x, y, w, h, p (mean)] and SNR information, but also the node id, target matching verification mode, the number of potential electromagnetic targets and the electromagnetic target information screenshot corresponding to the potential electromagnetic target information [c, x, y, w, h, p (mean)].

6. A collaborative electromagnetic signal intelligent recognition method according to any one of claims 4-5, characterized in that: The multi-information weighted matching fusion recognition process in step S8 includes the following steps: S81, in descending order, the current node status value of the collaborative sensing node Sort by The highest collaborative sensing node is the anchor node, denoted as anchor node M; S82, scan the potential electromagnetic target list of the anchor node M, and record the potential electromagnetic target information of the anchor node M as ,in, , N is the total number of potential electromagnetic target information scanned at the anchor node M; S83, for the anchor node M i Potential electromagnetic target information , and its target center is recorded as , the time and frequency coordinates corresponding to the target center are marked as and , and Uniquely determined by x and y respectively, set the corresponding time range and frequency range ,in, To characterize the time difference of each sensing node receiving the same signal, it is related to the interval distance between each sensing node and is preset globally. The frequency difference related to the receiver frequency difference and Doppler shift factor is preset as needed; S84, the first potential electromagnetic target information from the anchor node M At the beginning, the N potential electromagnetic target information of the anchor node M is processed in turn until the information processing is completed.

7. A collaborative electromagnetic signal intelligent recognition method as claimed in claim 6, characterized in that: The sequentially processing of the N potential electromagnetic target information of the anchor node M comprises the following steps: S841, for i Potential electromagnetic target information , with its target center To anchor the center point, in the time range and frequency range Within, traverse the potential electromagnetic target list of other cooperative sensing nodes, if it is found that the center of the potential electromagnetic target of the other cooperative sensing node falls within the time range and frequency range Inside, and its Greater than a predetermined threshold , it is considered that the current potential electromagnetic target identified by the other collaborative sensing nodes is consistent with the corresponding first i Potential electromagnetic targets may be the same target, so they are recorded as similar potential electromagnetic targets; S842, the detected i All similar potential electromagnetic targets similar to the potential electromagnetic targets are constructed into a new information set X, and the electromagnetic target is selectively identified by an agile method or a precise method according to the target matching verification mode to obtain the final result of the electromagnetic identification target.

8. A collaborative electromagnetic signal intelligent recognition method as claimed in claim 7, characterized in that: The information set X is expressed as: ； The agile method does not consider the time-frequency diagram corresponding to the electromagnetic target. It is directly based on the average power weighting of the target area and obtains the final result of electromagnetic identification target through the following calculation formula: 。 9. A collaborative electromagnetic signal intelligent recognition method as claimed in claim 8, characterized in that: The precise method considers the time-frequency graph corresponding to the electromagnetic target, which requires first i The time-frequency screenshots corresponding to the potential electromagnetic targets and the time-frequency screenshots corresponding to each similar potential electromagnetic target in the set X are input into a CNN network to calculate the graphic similarity and obtain the corresponding similarity factor , ,…… ; Then, based on the average power and similarity factor of the target area, the final result of electromagnetic identification target is obtained by comprehensive weighting through the following calculation formula: 。 10. A multi-point cooperative intelligent electromagnetic identification system, which is used to implement a cooperative electromagnetic signal intelligent identification method according to any one of claims 1 to 9, characterized in that: Each collaborative sensing node includes at least a signal acquisition unit, a state analysis unit, a model configuration unit, a model pool, an inference model unit and a first information interaction unit; the collaborative identification controller includes at least a potential electromagnetic target list unit, a node state information table unit, a second information interaction unit and a fusion perception decision unit; the first information interaction unit communicates with the second information interaction unit, and the fusion perception decision unit is used to perform multi-information weighted matching fusion identification processing on the potential electromagnetic target list reported by each collaborative sensing node based on the constructed potential electromagnetic target information matrix and the node state information table, comprehensively consider the state of each collaborative sensing node, and output the final result of the electromagnetic identification target.

Images