Verification system based on radar signal data and verification method thereof

Abstract

The invention relates to the technical field of radar signal processing, in particular to a verification system and method based on radar signal data, and the system comprises an information collection module, a pulse network module, a signal processing module, a feature analysis module and an error distribution evaluation module. The method comprises the following steps: collecting radar signal data, and recording an interference dense area of the radar signal data under each deviation angle; forming a radar pulse time sequence by the radar signal data at each deviation angle, and identifying a pulse sequence network of the radar signal data at each deviation angle; interference accumulation clustering is carried out on each grid node in the pulse sequence network, accumulated superposed interference variables are recorded, and a label distribution result corresponding to the radar signal data is determined; and performing maximum similarity evaluation on the error distribution condition in the label distribution result to obtain an interference evaluation result of each error. The verification efficiency of the radar signal and the accuracy of interference identification are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of radar signal processing, and specifically to a verification system and a verification method based on radar signal data. Background Art

[0002] Signal processing plays a crucial role in modern radar systems, which determines the overall performance of the radar. Signal processing in modern radar systems needs to face various application requirements and complex working environments. When there are multiple interference sources in the radar working environment, if only a single means is used to verify the radar signal, it is easy to cause missing verification data, resulting in a decrease in the accuracy of tracking and identifying interference sources in the radar signal.

[0003] For example, Chinese Patent Publication No. CN115994096A discloses a verification method for radar signal processing and simulation, which relates to the technical field of radar signal processing. In the method of the present invention, on the one hand, the echo signal data is processed using the simulation algorithm in the simulation software to obtain simulation data; on the other hand, the echo signal data is processed using the signal processing algorithm of the radar to obtain processed data; by comparing the simulation data and the processed data, the correctness of the processed data is verified, so as to realize the test of the signal processing algorithm of the radar. During this process, the DDR memory in the DSP chip is partitioned in the present invention, and the echo signal data, simulation data, and processed data are stored in partitions for easy data comparison.

[0004] For example, Chinese Patent Publication No. CN116974453A discloses a signal processing method, a signal processing device, a signal processor, a device, and a medium, belonging to the technical field of computers. The method includes: respectively sampling a preset target through a first sampler and a second sampler to obtain a plurality of first sampling signals with a first sampling frequency and a plurality of second sampling signals with a second sampling frequency; for a prediction moment, performing signal prediction based on historical output signals to determine a predicted sampling signal at the prediction moment, where the historical output signals include at least one sampling output signal before the prediction moment; determining an error comparison result based on the predicted sampling signal at the prediction moment and the second sampling signal at the prediction moment; and selecting and outputting a sampling output signal from the first sampling signals and the second sampling signals based on the error comparison result.

[0005] In the prior art, it separately describes verifying the data header of the radar signal to determine whether the radar signals are consistent, and comparing the errors of the radar signals to select the sampled output data. These methods tend to compare the radar signals based on their signal strengths to determine the received part of the radar signals. These data are more targeted at stationary radar data, making it difficult to identify the timing characteristics of non-stationary radar signals. When there are a large number of interferences in the radar signals, it is necessary to comprehensively identify the relevant timing characteristics and cumulative effects of the radar signals to determine the corresponding characteristics of the radar signals under the condition of multiple interference sources, so as to improve the anti-interference ability of the radar signals. Summary of the Invention

[0006] To solve the above technical problems, the technical solution adopted by the present invention is: A verification system based on radar signal data, including: An information acquisition module, which is used to acquire radar signal data, perform signal offset processing on the radar signal data according to the relative speeds and distances of each interference source, divide the spatial sectors of the radar signal data, and record the interference-dense areas of the radar signal data at each offset angle.

[0007] A pulse network module, which is used to form a radar pulse time series from the radar signal data at each offset angle, use the data of the radar pulse time series as nodes, and identify the pulse sequence network of the radar signal data at each offset angle according to the pre-state and post-state of each node.

[0008] A signal processing module, which is used to judge the adjacency relationship of each grid node in the pulse sequence network, perform interference cumulative clustering on the interference-dense areas of the corresponding radar signal data according to the adjacency relationship of each grid node, and record the superimposed interference amount after accumulation.

[0009] A feature analysis module, which is used to determine the distribution label corresponding to the radar signal data in the pulse sequence network based on the superimposed interference amount of the radar signal data, and determine the label distribution result corresponding to the radar signal data according to the iterative results of each distribution label.

[0010] An error distribution evaluation module, which is used to perform statistics on the errors between the actual data and the simulation data corresponding to the label distribution result, determine the error distribution situation of each grid node in the pulse sequence network, and perform a maximum similarity evaluation on the error distribution situation to obtain the interference evaluation result of each error.

[0011] A verification method based on radar signal data, including: S1, acquiring radar signal data, dividing the spatial sectors of the radar signal data, and recording the interference-dense areas of the radar signal data at each offset angle.

[0012] S2. Compose the radar signal data at each offset angle into a radar pulse time series. Using the data of the radar pulse time series as nodes, identify the pulse sequence network of the radar signal data at each offset angle according to the pre-state and post-state of each node.

[0013] S3. Judge the adjacency relationship of each grid node in the pulse sequence network, and perform interference cumulative clustering on the interference dense areas of the corresponding radar signal data according to the adjacency relationship of each grid node, and record the superimposed interference amount after accumulation.

[0014] S4. Based on the superimposed interference amount of the radar signal data, determine the distribution label corresponding to the radar signal data in the pulse sequence network, and determine the label distribution result corresponding to the radar signal data according to the iterative results of each distribution label.

[0015] S5. Statistically analyze the error between the actual data and the simulation data corresponding to the label distribution result, determine the error distribution of each grid node in the pulse sequence network, and perform the maximum similarity evaluation on the error distribution to obtain the interference evaluation result of each error.

[0016] The beneficial effects of the present invention are as follows: First, after dividing the collected radar signals into multiple spatial sectors, the present invention records the interference dense areas of the radar signal data at each offset angle; it can know the positions where the radar signals are offset in the scenario where the interference source is moving, and segment according to these positions, and can identify the relative dynamic characteristics of the interference source; then, according to the changes in the radar signal data in the time series, a pulse sequence network is composed, and the pre-state and post-state under multiple interference sources can be known, so as to quantify the timing characteristics of the radar signal within the pulse repetition period, and determine whether there will be relative changes in the relative distribution and distribution of each node under the time change of the received radar signal, so as to verify whether the current radar signal is interfered by multiple interference sources, resulting in the problem of reduced positioning accuracy of the received radar signal itself, and improve the accuracy of radar signal interference recognition and the accuracy of spatial division tracking.

[0017] Second, by identifying the adjacency relationship of each grid node in the pulse sequence network and performing interference cumulative clustering on the corresponding data, the present invention can identify the specific situation of the radar accumulation and the relative interference characteristics of the radar signal after accumulation in the scenario where there is an interference accumulation effect, so as to further quantify the influence of the interference source on the radar signal and improve the comprehensiveness of the description of the radar signal characteristics.

[0018] III. The present invention iteratively analyzes the grid nodes corresponding to the superimposed interference amount, and uses the label propagation method to extract data from the distributed labels to obtain the label distribution result, which can identify the part of the data mainly affected by the interference source in the current radar signal under multiple interference characteristics or radar signal changes, identify the combination of various distributed labels under mixed interference, so as to provide the probability distribution of the interference type, facilitate subsequent adjustment for the interference signal; the error statistics of the label distribution result further quantifies the error situation of the data under the probability distribution of the corresponding interference type, and finds out the parts with similar errors and the same maximum error, so as to verify the interference mode generated by the current interference source on the radar signal under the corresponding interference type; improve the accuracy of subsequent radar signal verification and the efficiency of target tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The present invention will be further described below in conjunction with the drawings and embodiments.

[0020] Figure 1 It is a system framework diagram of a verification system based on radar signal data.

[0021] Figure 2 It is a schematic flowchart of an information acquisition module of a verification system based on radar signal data.

[0022] Figure 3 It is a schematic flowchart of a pulse network module of a verification system based on radar signal data.

[0023] Figure 4 It is a schematic flowchart of a signal processing module of a verification system based on radar signal data.

[0024] Figure 5 It is a schematic flowchart of a verification method based on radar signal data. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] The embodiments of the present invention will be described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as a limitation of the present invention. For those not specified in the embodiments, the techniques or conditions described in the literature in the field or according to the product specifications are followed.

[0026] Refer to Figure 1 , a verification system based on radar signal data, includes: an information acquisition module, a pulse network module, a signal processing module, a feature analysis module, and an error distribution evaluation module; wherein, the output end of the information acquisition module is connected to the pulse network module, the output end of the pulse network module is connected to the signal processing module, the output end of the signal processing module is connected to the feature analysis module, and the output end of the feature analysis module is connected to the error distribution evaluation module.

[0027] An information acquisition module, which is used to acquire radar signal data, perform signal offset processing on the radar signal data according to the relative speed and distance of each interference source, divide the spatial sectors of the radar signal data, and record the interference dense areas of the radar signal data at each offset angle.

[0028] A pulse network module, which is used to form a radar pulse time series from the radar signal data at each offset angle, use each data of the radar pulse time series as a node, and identify the pulse sequence network of the radar signal data at each offset angle according to the pre-state and post-state of each node.

[0029] A signal processing module, which is used to judge the adjacency relationship of each grid node in the pulse sequence network, perform interference cumulative clustering on the interference dense areas of the corresponding radar signal data according to the adjacency relationship of each grid node, and record the superimposed interference amount after accumulation.

[0030] A feature analysis module, which is used to determine the distribution label corresponding to the radar signal data in the pulse sequence network based on the superimposed interference amount of the radar signal data, and determine the label distribution result corresponding to the radar signal data according to the iteration result of each distribution label.

[0031] An error distribution evaluation module, which is used to perform statistics on the error between the actual data and the simulation data corresponding to the label distribution result, determine the error distribution situation of each grid node in the pulse sequence network, and perform a maximum similarity evaluation on the error distribution situation to obtain the interference evaluation result of each error.

[0032] In an embodiment of the present invention, the radar signal data will include the original radar signal and the interference signal data to determine the offset and error generated by the radar signal in the case of multiple interference sources.

[0033] The original radar signal usually has a stable pulse repetition interval, a fixed pulse width, a modulation method, and an arrival angle. These values are basically stable and can be compared with the standard radar signal to determine whether there is interference signal data. The included interference signal data will be modulated in advance to generate interference signal data in the way of adding multiple noise sources. The interference signal data will appear as random pulses, abnormally high power, or deceptive interference imitating the radar signal, such as false repetition frequency, etc. At this time, the corresponding frequency characteristics in the radar signal data will be extracted to represent the interference signal data.

[0034] After collecting radar signal data, the current radar signal data will be recorded according to the speed of the identified objects at the positions and times shown on the collected signals, and then the polar coordinate relative positions where interference signals exist will be calculated to determine the spatial sector processed by the radar signal data. After that, the occurrence of interference signals in the spatial sector will be identified, and the areas with dense interference will be recorded to describe the impact on the current radar.

[0035] As Figure 2 shown, the implementation method of the information acquisition module includes: processing the collected radar signal data using a queue, and identifying the interference sources that appear in the radar signal data during queue processing.

[0036] Extract the relative speed and distance of each interference source, and calculate the offset when the interference source interferes with the radar signal.

[0037] Map the offset to the coordinate system corresponding to the radar signal data to obtain the offset angle corresponding to the radar signal data.

[0038] According to the offset angle corresponding to the radar signal data, the radar signals are divided into multiple spatial sectors. Using the total signal energy of the radar signal data in each spatial sector, if the total signal energy is greater than the signal energy threshold, the corresponding sector is marked as a dense interference area. For the signal energy threshold, it can be judged according to the average value of the received radar signals in the historical data plus three times the standard deviation to determine its threshold situation relative to the statistical characteristics. If the dynamic threshold method is adopted, the larger part of the average value of the radar signal and the maximum theoretical value is selected as the signal energy threshold used at this time to divide the interference source signals that need to be identified.

[0039] For the offset processing, it is mainly based on the methods of Doppler frequency shift compensation and time delay compensation. After tracking the interference sources existing in the radar signal data, the time delay compensation is calculated based on the time difference when the interference signal data arrives at the radar, and then its relative speed is used to calculate its Doppler frequency shift. After that, the calculated offset is mapped to the coordinate system of the radar signal data. For example, based on the arrival angle, the angle offset caused by its time delay compensation is represented, and then with the interval of 0 - 360 degrees, multiple small intervals are divided, and the sum corresponding to the radar signal data in each interval is calculated to judge whether the corresponding interval is an area with dense interference signals.

[0040] The offset angle indicates the angular position of the signal received by the current radar after processing using the relative speed and distance of each interference source. At this time, with 0 degrees as the reference point, the relative positions of each interference source are determined.

[0041] In one embodiment of the present invention, the pre-state mainly obtains the values of the node in the previous period of time, and calculates whether the nodes are similar at adjacent offset angles according to these values.

[0042] For the post-state, it is to record the subsequent data change rate of the node. For example, when an interference signal appears, the change rate between the data in the previous period and the data in the subsequent period is used to describe the change of the point in the time series. Then, after classifying each node using the pre-state and the post-state, a pulse sequence network is used to represent the relative situation of the current radar signal.

[0043] Such as Figure 3 As shown, the implementation methods of the pre-state and the post-state of each node in the pulse network module include: based on the radar pulse time series, extracting multiple signal anchor points within the radar pulse repetition period. The pulse repetition period refers to the minimum time interval between two consecutive radar pulses; the signal anchor points are several data points randomly selected within the rising or falling period of the radar pulse time series. When these data points are selected, they will include the offset angle corresponding to the data point, that is, the relative position and signal power shown by the point in the radar signal data. When the radar signal is collected, the signal power of the radar signal can be known through analog-to-digital conversion.

[0044] Calculate the signal gain difference of each signal anchor point at adjacent offset angles as the pre-state of each node, and calculate the change rate of the signal gain difference of each signal anchor point within adjacent time periods as the post-state of each node.

[0045] Taking the pre-state and the post-state of each node as the main factors, connect each node to combine and obtain a pulse sequence network.

[0046] In this pulse sequence network, the connection is mainly carried out through the node visibility between each node. For example, using the requirements for each node in the ILPVG algorithm, or methods such as the VG model, HVG model, and SLPVG model. After judging the relative relationship and data difference between the current nodes, a pulse sequence network of the current radar pulse time series is formed. In this network, the difference value between each node is mainly used to determine the connection relationship between the nodes, so as to determine whether there will be relative changes in the relative distribution and distribution of each node under the time change of the received radar signal, so as to verify whether the current radar signal is interfered by multiple interference sources, resulting in the problem of reduced positioning accuracy of the received radar signal itself.

[0047] That is, the implementation method of the pulse sequence network includes: performing node grid mapping according to the pre-state and the post-state of each node, and dividing the pre-state and the post-state of each node into multiple grids according to the values.

[0048] Map the offset angles of each node to the grids corresponding to the pre-state and post-state, and record the number of data in each grid.

[0049] Merge the data in each grid into a grid node, identify the similar edges between the grid nodes, and combine them into a pulse sequence network according to the similar edges.

[0050] When merging the data in each grid into a grid node, merge the grids under the same offset angle and the adjacent low-count grids into a grid node. Then, calculate the similarity between the grid nodes using the signal gain difference. When the similarity is greater than the similarity threshold, the similar edges between the corresponding grid nodes are obtained. The similarity threshold can be set to 0.6, which is used to represent the data of radar signals in the case of similarity.

[0051] Preferably, the signal gain difference of each signal anchor at adjacent offset angles can be expressed as: ; where represents the signal gain at the i-th and j-th moments, represents the signal power at the i-th moment, represents the signal power at the j-th moment. Then, the signal gain difference of adjacent offset angles is to find the difference between the signal gains of adjacent offset angles; for example ; where represents the signal gain difference at adjacent offset angles at the i-th and j-th moments, and represent the signal gains at two adjacent offset angles.

[0052] Then, the change rate of the signal gain difference of each signal anchor in adjacent time periods is to find the difference between the signal gains in adjacent time periods and divide it by the length of the interval time period to represent the change rate of the signal gain difference. At the same time, the method of taking the derivative can be used. For example, the signal gain difference is expressed as, ; where represents the length of the interval of adjacent time periods, and this length is the time unit; and respectively represent the signal gains at times and respectively.

[0053] Then, the change rate of the signal gain difference is expressed as: , so as to know the situation of the signal gain described currently at adjacent offset angles. The i and j mentioned above both represent the serial numbers of the time index, and are only used to illustrate the situation of the signal gain obtained.

[0054] In one embodiment of the present invention, when determining the adjacency relationship of each grid node in the pulse sequence network, the adjacency type of each grid node is mainly judged, and after mapping to the positions of each interference-dense area according to the adjacency type, clustering is performed on these points to verify the cumulative effect of the grid nodes; for the radar signal data of interference cumulative clustering, it is mainly to judge whether there is an intersection or mutual influence in each interference-dense area when the radar signal appears with interference superposition, so as to determine the direction in which the radar working mode is mainly affected when the current radar is under the influence of multiple interference sources.

[0055] As Figure 4 shown, the implementation method of the signal processing module includes: using each network node of the pulse sequence network as the judgment basis, counting the effective time of the radar signal, and the interference occupancy ratio of the effective time of each grid node under a given adjacency relationship; the effective time of the radar signal is mainly represented by the product of the pulse width and the pulse repetition times. If the radar signal is in the form of a continuous wave signal, the product of the signal duration and the duty cycle is selected to determine the effective time of the radar signal. For a given adjacency relationship representing time adjacency and frequency similarity, that is, the time difference of the radar signal data contained in each grid node is within the radar pulse repetition interval, and the frequency difference of each grid node is less than the average instantaneous bandwidth value. When both of these two conditions are met, the interference occupancy ratio existing in these grid nodes is counted. The interference occupancy ratio is the ratio of the time with interference signals to the limited time.

[0056] Taking the position of each grid node in the interference-dense area, the interference occupancy ratio of the effective time of each grid node is used for accumulation, and the fitting value of each grid node during accumulation is recorded; it is explained here that the grid nodes after statistics are accumulated according to their positions, and the radar signal data received at these positions is accumulated. If a single value is too large after data fitting, the accumulated data at this time is output.

[0057] If the fitting value of each grid node is greater than the first preset threshold, the data accumulated in the current interference-dense area is output, and the superimposed interference amount is output according to the size of the accumulated value.

[0058] If the accumulated quantity of each grid node is greater than the second preset threshold, interference cumulative clustering is performed in the direction corresponding to each grid node, and the signal peak value after clustering is used as the superimposed interference amount. The maximum value among the products of the effective time and the interference weight of all nodes within the cluster is regarded as its corresponding signal peak value. If there are multiple clusters output at this time, there will be multiple signal peak values.

[0059] The first preset threshold is a critical value set for the fitting value obtained by performing time series fitting on the interference occupancy ratio of the effective time of the radar signal; it can be set according to the linear dynamic range of the radar receiver, and usually takes the interference occupancy ratio corresponding to 60%-70% of the ADC saturation power; it mainly reflects the continuous action intensity of the interference signal in the space-time dimension and characterizes the pressure level of the interference source on the radar system.

[0060] When exceeding the first preset threshold, the following interference characteristics generally exist: presenting continuous waves or high-duty-cycle pulses in the time domain, covering more than 80% of the radar operating bandwidth in the frequency domain, and the power spectral density exceeding the linear dynamic range of the receiver; then these interference characteristics will directly affect the ADC sampling to appear clipping distortion, the orthogonality of I / Q signals to be damaged, the main lobe of pulse compression to split (when SNR < 15dB), the adaptive failure of the CFAR detection threshold, the maximum detection range to decrease by 40%-60%, the target update rate to decrease to 1 / 3 of the normal value, etc., which will directly affect the current use of the radar.

[0061] The second preset threshold is a critical value of the number of grids in the interference-dense area set on the basis of three-dimensional space grid division; during accumulation, these grid nodes will continuously map to the interference-dense area, and the grids represented by these grid nodes will also continuously increase; this second preset threshold characterizes the spatial distribution density of the interference signal and reflects the number or coverage range of the interference sources; it is usually set according to the maximum track processing capacity of the radar signal processor, and usually takes 30%-50% of the target capacity number; when exceeding this threshold, it indicates that the coverage range of the interference signal is too large.

[0062] When exceeding the second preset threshold, the following interference characteristics may exist: the spatial distribution presents multi-beam direction modulation, the time-frequency characteristics are highly similar to the target echo, and the interference signals have co-varying characteristics. Then it will cause the situation awareness refresh rate to drop from 1 second to 5 seconds, and the threat level assessment accuracy to decrease by 75%, etc., which will directly affect the accuracy of identifying multiple interference sources in the interference signal. As shown in Table 1, the data related to the superimposed interference amount is finally output.

[0063] Table 1. Output schematic table

[0064]

[0065] When performing interference accumulation and clustering, it also includes recording the interference characteristics of the superimposed interference amount and sorting the output superimposed interference amount according to the occurrence probability of each interference characteristic. When sorting at this time, the interference characteristic that appears the most in total during the processing of the obtained radar signal data will be used as the first sorting index, and then the number of occurrences of the interference characteristics will be used as the sorting index in turn to complete the sorting of the data corresponding to the superimposed interference amount.

[0066] In one embodiment of the present invention, when setting distribution labels, the label propagation algorithm is mainly used to verify whether the labels of grid nodes corresponding to the superimposed interference amount change during continuous iteration, so as to identify the existing label distribution results.

[0067] That is, the implementation manner of the feature classification module includes: setting distribution labels for each grid node according to each grid node corresponding to the superimposed interference amount, constructing a similarity matrix with periodic constraints based on the similarity weights under each distribution label, determining the data of each element in the similarity matrix in continuous time approximation, updating the similarity weights of the corresponding elements of the similarity matrix, and using the updated similarity matrix as the label distribution result.

[0068] Because in the case of a large amount of interference in radar signals, there will be serious overlapping parts between the radar signals, and this part will affect the overall radar signal's identifiability and comprehensibility. Outputting the part with significantly too high similarity during label propagation is to represent the part with little change under label propagation, and to identify whether the part with significantly high similarity is the information of obvious interference in the current radar signal. Then, this part of the result is used as the data for subsequent verification of the error distribution to determine the data form of the current radar signal data affected by multiple interference sources, which is convenient for subsequent processing in the case of multiple interferences.

[0069] Preferably, when processing the grid nodes corresponding to the superimposed interference amount, the data of the corresponding grid nodes is converted into a time-frequency feature vector , ; where represents the number of the grid node, , , respectively represent the signal energy, instantaneous frequency offset, and dispersion coefficient of grid node a, represents the vector transpose operation. Its signal energy reflects the intensity of the signal, the instantaneous frequency offset is equivalent to the reference frequency, and the dispersion coefficient characterizes the degree of signal diffusion.

[0070] After that, the superimposed interference amount existing on these grid nodes is described, such as using the interference amount modulation index to represent: ; where represents the Fourier coefficient of the kth category of the signal of grid node a; the obtained interference amount modulation index is mainly used to reflect the signal frequency domain complexity to represent the relative situation of the data of each grid node under the current processing. The value range of k is from 1 to K, which is related to the number of set distribution labels, and the value range of a is from 1 to A, representing the number of grid nodes described.

[0071] The grid nodes at this time are divided into K initial distribution label classes , , and then combine these data into an initial label matrix, ; among them, represents the indicator function, which takes 1 when the grid node a belongs to the category and takes 0 otherwise; represents an element of the initial label matrix, indicating the confidence that the grid node a belongs to the category k; represents the data corresponding to the grid node a.

[0072] When constructing the similarity matrix, first determine the basic similarity weight of the initial label matrix, then calculate the periodic modulation factor under the periodic constraint, and then obtain the similarity matrix.

[0073] The basic similarity weight of the initial label matrix is expressed as: ; among them, represents the basic similarity weight between the grid node a and the grid node b. The value range of b is the same as that of a, which is also from 1 to A; represents the Euclidean distance between the grid node a and the grid node b in the feature space, , , respectively represent the time-frequency feature vectors of the grid node a and the grid node b; represents the bandwidth parameter of the Gaussian kernel function, which controls the similarity decay rate; represents the exponential function; then we know the similarity weight represented by the grid node after setting the distribution label.

[0074] The periodic modulation factor under the periodic constraint can be expressed as: ; among them, represents the center frequency difference between the grid node a and the grid node b, that is, the instantaneous frequency offset is used to calculate the center frequency difference between the grid node a and the grid node b; represents the frequency difference attenuation coefficient, represents the time variable, which is used as the evolution parameter in the continuous time approximation, represents the periodic modulation factor between the grid node a and the grid node b, reflecting the periodic modulation of the frequency difference on the similarity; represents the exponential constant, represents the pi; through this part of the calculation, we can know the form represented by the relative factor under its periodic constraint and limitation. Then, the sum of these two values is the weight used in the similarity matrix.

[0075] After that, under the continuous time approximation processing of the elements in the similarity matrix, it can also be expressed in the following way.

[0076] ; among them, is the propagation coefficient, whose range can be 0.7 - 0.95, and is used to control the intensity of label propagation; represents the confidence that grid node b belongs to category k at the (t + 1)-th iteration; represents the weight of grid node a and grid node b in the similarity matrix, ; represents the confidence that grid node a belongs to category k at the t-th iteration, represents the anchoring term of grid node b in the initial label matrix, which is used to maintain the influence of the original label information. At this time, in the case of continuous time, it is used to verify whether the confidence of each data in the current similarity matrix has changed, so as to identify whether the relative form in the radar signal data will affect the change of the label weight, thereby improving the accuracy and robustness of the label distribution.

[0077] After that, under the continuous time approximation processing, whether its relative output data can be stable, and the stable part is output to extract and identify the obvious interference source part in the current radar signal.

[0078] ; where represents the maximum change amount between two iterations, represents the convergence threshold, usually taking values from 1e - 4 to 1e - 6, and is used to judge whether the algorithm reaches a steady state; represents the confidence that grid node b belongs to category k at the (t + 1)-th iteration.

[0079] Finally, after the maximum change amount between iterations is less than the set convergence threshold, the output label distribution results can represent the relevant values of the obvious interference sources identified currently.

[0080] In an embodiment of the present invention, the error distribution evaluation module mainly processes the identification of the error distribution of the corresponding data after superposition under the superposition of information of multiple interference sources; to determine the error interference direction and interference intensity corresponding to each interference source under the maximum similarity evaluation, and finally to determine the interference direction and the superposition intensity value when the current radar signal is received and affected by interference sources or other signals; finally, the identified data is used as the evaluation result of the subsequent output.

[0081] That is, the implementation manner of the error distribution evaluation module further includes: taking the radar signal data corresponding to the label distribution result as the main, calculating the error, determining that the target with the maximum similarity between errors and the target corresponding to the maximum error are the same target, and outputting the interference evaluation results corresponding to the current errors according to the data proportion of the target in the label distribution result.

[0082] The similarity between the errors is calculated using cosine similarity with the generated error values, that is, the difference is calculated between the data predicted by the label distribution and the part of the true distribution to obtain the corresponding error values. Then, a set of data with the maximum similarity is extracted, and the target corresponding to this data in the label distribution result is compared with the target corresponding to the calculation of the error to determine the relative situation during the current error evaluation.

[0083] If the target with the maximum similarity is consistent with the target with the maximum error, it indicates that the model's prediction of high-error targets is consistent; if not, it may indicate data contamination or model overfitting; at this time, the final output part will include the data under the label distribution result, as well as the target with the maximum error, etc., to display the content verified by the current radar signal data, facilitating the staff to adjust the radar operation and received data situation.

[0084] For example, the predicted label distribution corresponding to the label distribution result and the true label distribution has an error value of , ; then, based on this error value, its cosine similarity is obtained, and the maximum value of the cosine similarity and the maximum value of the error value are found. After normalizing the data proportion of the corresponding target in the label distribution result and the error value, the product of the weighted sum of the error value and the data proportion of the corresponding target in the label distribution result is calculated to obtain the corresponding evaluation index. After using this index as the identifier for the corresponding data, the corresponding data is output as the interference evaluation result.

[0085] As Figure 5 shown, the present invention also provides a verification method based on radar signal data, including: S1, collecting radar signal data, after dividing the radar signal data into spatial sectors, recording the interference dense areas of the radar signal data at each offset angle.

[0086] S2, forming a radar pulse time series with the radar signal data at each offset angle, using each data of the radar pulse time series as a node, and identifying the pulse sequence network of the radar signal data at each offset angle according to the pre-state and post-state of each node.

[0087] S3, judging the adjacency relationship of each grid node in the pulse sequence network, and performing interference cumulative clustering on the interference dense areas of the corresponding radar signal data according to the adjacency relationship of each grid node, and recording the cumulative superimposed interference amount.

[0088] S4, based on the superimposed interference amount of the radar signal data, determining the distribution label corresponding to the radar signal data in the pulse sequence network, and determining the label distribution result corresponding to the radar signal data according to the iterative results of each distribution label.

[0089] S5. According to the error between the actual data corresponding to the tag distribution result and the simulation data, statistics are carried out to determine the error distribution of each grid node in the pulse sequence network, and a maximum similarity evaluation is carried out on the error distribution to obtain the interference evaluation result of each error.

[0090] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention, and still be covered by the protection scope of the present invention.

Claims

1. A verification system based on radar signal data, characterized in that: include: An information collection module is used to collect radar signal data, perform signal offset processing on the radar signal data according to the relative speed and distance of each interference source, divide the radar signal data into spatial sectors, and record the interference-intensive areas of the radar signal data at each offset angle; A pulse network module is used to form a radar pulse time sequence from radar signal data at each offset angle, and to identify a pulse sequence network of radar signal data at each offset angle according to the pre-state and post-state of each node, using each data of the radar pulse time sequence as a node; The signal processing module is used to determine the adjacency relationship of each grid node in the pulse sequence network, and to perform interference accumulation clustering on the interference-intensive areas corresponding to the radar signal data according to the adjacency relationship of each grid node, and record the accumulated superposition interference amount; A feature analysis module is used to determine the distribution labels corresponding to the radar signal data in the pulse sequence network based on the superimposed interference amount of the radar signal data, and determine the label distribution results corresponding to the radar signal data according to the iteration results of each distribution label; The error distribution evaluation module is used to perform statistics on the errors between the actual data and the simulation data corresponding to the label distribution results, determine the error distribution of each grid node in the pulse sequence network, and perform maximum similarity evaluation on the error distribution to obtain the interference evaluation results of each error.

2. A verification system based on radar signal data according to claim 1, characterized in that: The radar signal data includes the original radar signal and the interference signal data.

3. A verification system based on radar signal data according to claim 1, characterized in that: The implementation methods of the information collection module include: Processing the collected radar signal data using a queue, and identifying interference sources appearing in the radar signal data during the queue processing; Extract the relative speed and distance of each interference source and calculate the offset when the interference source interferes with the radar signal; Mapping the offset to the coordinate system corresponding to the radar signal data to obtain the offset angle corresponding to the radar signal data; According to the offset angle corresponding to the radar signal data, the radar signal is divided into multiple spatial sectors. The total signal energy of the radar signal data in each spatial sector is used. If the total signal energy is greater than the signal energy threshold, the corresponding sector is marked as an interference-intensive area.

4. A verification system based on radar signal data according to claim 1, characterized in that: The implementation methods of the pre-state and post-state of each node in the pulse network module include: Based on the radar pulse time series, multiple signal anchor points within the radar pulse repetition period are extracted; Calculate the signal gain difference under each signal anchor point as the pre-state of each node, and calculate the change rate of the signal gain difference inside each signal anchor point as the post-state of each node; Based on the pre-state and post-state of each node, the nodes are connected and combined to obtain a pulse sequence network.

5. A verification system based on radar signal data according to claim 4, characterized in that: The implementation of the pulse train network includes: According to the pre-state and post-state of each node, node grid mapping is performed, and the pre-state and post-state of each node are divided into multiple grids according to the values; Map the offset angle of each node to the grid corresponding to the previous state and the next state, and record the number of data in each grid; The data in each grid is merged into a grid node, similar edges between grid nodes are identified, and they are combined into a pulse train network according to similar edges.

6. A verification system based on radar signal data according to claim 1, characterized in that: The implementation of the signal processing module includes: Each network node of the pulse sequence network is used as the basis for judgment, and the effective time of the radar signal and the interference ratio of the effective time of each grid node under a given adjacency relationship are counted; Based on the location of each grid node in the interference-intensive area, the interference ratio of each grid node's effective time is accumulated, and the fitting value of each grid node during the accumulation is recorded; If the fitting value of each grid node is greater than the first preset threshold, the accumulated data of the current interference-intensive area is output, and the superimposed interference amount is output according to the accumulated value; If the accumulated number of each grid node is greater than the second preset threshold, interference accumulation clustering is performed in the direction corresponding to each grid node, and the peak value of the clustered signal is used as the superimposed interference amount.

7. A verification system based on radar signal data according to claim 6, characterized in that: Interference accumulation clustering also includes: The interference features of the superimposed interference amount are recorded, and the output superimposed interference amount is sorted according to the occurrence probability of each interference feature.

8. The verification system based on radar signal data according to claim 1, characterized in that: The implementation methods of the feature classification module include: According to each grid node corresponding to the superimposed interference amount, a distribution label is set for each grid node, and a similarity matrix with periodic constraints is constructed with the similarity weights under each distribution label. The data approximated by each element in the similarity matrix in continuous time is determined, and the similarity weights of the corresponding elements of the similarity matrix are updated. The updated similarity matrix is ​​used as the label distribution result.

9. The verification system based on radar signal data according to claim 1, characterized in that: The implementation of the error distribution evaluation module also includes: The error calculation is mainly based on the radar signal data corresponding to the label distribution results. It is determined that the target with the largest similarity between the errors is the same target as the target corresponding to the maximum error. According to the data proportion of the target in the label distribution results, the interference assessment results corresponding to the current errors are output.

10. A verification method based on radar signal data, characterized in that: include: S1, collecting radar signal data, dividing the radar signal data into spatial sectors, and recording the interference-intensive areas of the radar signal data at each offset angle; S2, composing radar signal data at each offset angle into a radar pulse time sequence, using each data of the radar pulse time sequence as a node, and identifying a pulse sequence network of the radar signal data at each offset angle according to the pre-state and post-state of each node; S3, determining the adjacency relationship of each grid node in the pulse sequence network, and performing interference accumulation clustering on the interference-intensive area corresponding to the radar signal data according to the adjacency relationship of each grid node, and recording the accumulated superimposed interference amount; S4, determining the distribution label corresponding to the radar signal data in the pulse sequence network based on the superimposed interference amount of the radar signal data, and determining the label distribution result corresponding to the radar signal data according to the iteration result of each distribution label; S5, performing statistics on the errors between the actual data and the simulation data corresponding to the label distribution results, determining the error distribution of each grid node in the pulse sequence network, and performing maximum similarity evaluation on the error distribution to obtain interference evaluation results of each error.

Images