DBscan combines distribution posterior assessment and optimized unknown radar signal clustering method

Abstract

The application provides a kind of unknown radar signal clustering method of DBscan combination distribution posterior evaluation and optimization, comprising: for the two two-dimensional sequences of signal angle of arrival DOA and signal amplitude AMP two kinds of parameters, according to the performance of reconnaissance system, measurement error and measurement value, define the weighted distance D of two two-dimensional sequences, and define the two parameters Eps and MinPts of Dbscan clustering operation;Wherein, Eps represents neighborhood radius, and MinPts represents density threshold;According to the defined weighted distance D, neighborhood radius Eps and density threshold MinPts, the two two-dimensional sequences are carried out Dbscan clustering operation, and the clustering result is obtained;The obtained clustering result is evaluated, and the parameters of clustering result are updated according to the posterior evaluation result.The application can improve the accuracy and robustness of radar signal sequence clustering.

Description

Technical Field

[0001] This invention relates to the field of radar reconnaissance technology, and more specifically, to a DBscan method for clustering unknown radar signals by combining distributed posterior evaluation and optimization. Background Technology

[0002] One of the crucial steps in signal analysis for radar reconnaissance systems is clustering pulses that intersect in the time, frequency, and spatial domains of the electromagnetic spectrum. This involves separating multidimensional parameters from the overlapping signal sequences of numerous targets to obtain information about the electromagnetic targets for further analysis and identification. The challenges of this problem are twofold: first, electronic reconnaissance systems intercept a large number of unknown signals, and the number of targets in space is uncertain, making clustering difficult to adapt to different situations; second, due to factors such as signal strength and spatial relationships, the number of sample signals and sample errors obtained by electronic reconnaissance systems for each target vary greatly, making it difficult to cluster using fixed thresholds and rules.

[0003] Commonly used methods include histogram analysis, which involves statistically analyzing the distribution of parameters in a signal sequence over a given time period using histograms to determine the number of clusters and their centers. This method is computationally efficient and simple, but its robustness is relatively poor due to the need for fine-tuning rules to adapt to different signal environments. Another common method is dynamic clustering based on expert knowledge, where cluster radii for each dimension of the multi-dimensional parameters are determined. However, this method, by fixing the cluster radii, is difficult to adapt to different measurement errors and signal strengths in real-world electromagnetic environments. Summary of the Invention

[0004] The present invention aims to provide a clustering method for unknown radar signals that combines DBscan with distributed posterior evaluation and optimization, so as to improve the accuracy and robustness of radar signal clustering.

[0005] This invention provides a clustering method for unknown radar signals that combines DBscan with distributed posterior evaluation and optimization, comprising:

[0006] For two two-dimensional sequences consisting of two parameters, namely the angle of arrival (DOA) and the amplitude of signal (AMP), the weighted distance D between the two two-dimensional sequences is defined based on the reconnaissance system performance, the reconnaissance system measurement error, and the reconnaissance system's measurement values ​​of DOA and AMP. Two parameters, Eps and MinPts, are also defined for the Dbscan clustering operation. Here, Eps represents the neighborhood radius and MinPts represents the density threshold.

[0007] Based on the defined weighted distance D, neighborhood radius Eps, and density threshold MinPts, Dbscan clustering operation is performed on the two two-dimensional sequences to obtain the clustering results;

[0008] The obtained clustering results are evaluated posteriorly, and the parameters of the clustering results are updated based on the posterior evaluation results.

[0009] Furthermore, the weighted distance D between the two two-dimensional sequences is defined as:

[0010] D=W_DOA*d(DOA) / max_DOA+ W_AMP *d(AMP) / max_AMP

[0011] in:

[0012] W_DOA is the weight of the signal angle of arrival (DOA), W_AMP is the weight of the signal amplitude (AMP), and W_DOA>0, W_AMP>0, W_DOA +W_AMP=1;

[0013] max_DOA is the maximum value of the signal angle of arrival (DOA), and max_AMP is the maximum value of the signal amplitude (AMP). max_DOA and max_AMP are determined by experts based on the performance of the reconnaissance system and the measurement error of the reconnaissance system.

[0014] d(DOA) is the absolute value of the difference between two elements in the two-dimensional sequence formed by the signal angle of arrival (DOA); d(AMP) is the absolute value of the difference between two elements in the two-dimensional sequence formed by the signal amplitude (AMP); d(DOA) and d(AMP) are obtained by the reconnaissance system from the measurements of the signal angle of arrival (DOA) and the signal amplitude (AMP).

[0015] Furthermore, the neighborhood radius Eps is defined as:

[0016] Eps=W_DOA*max_DOA+W_AMP*max_AMP.

[0017] Furthermore, the density threshold MinPts is determined by experts based on the electromagnetic signal environmental density and the length of the DBscan clustering operation.

[0018] Furthermore, assuming the obtained clustering results have n classes, the posterior evaluation of the obtained clustering results includes:

[0019] (1) Calculate the earliest arrival time and the latest end time of the pulse in each type of signal sequence;

[0020] (2) Sort each class in ascending order of the earliest arrival time;

[0021] (3) Determine whether the time compatibility of two adjacent classes meets the conditions. If the time compatibility of two adjacent classes meets the conditions, perform subsequent operations; otherwise, perform the judgment in step (3) on the next class.

[0022] (4) Matching judgment is performed on the pulse width PW and radio frequency RF of two adjacent classes:

[0023] (4.1) For pulse width PW:

[0024] (4.11) For the pulse width PW of the pulse in the i-th class, calculate its mean PW. mean and root mean square PW ems ;

[0025] (4.12) Determine the root mean square PW ems Is it greater than the pulse width threshold PW? thr :

[0026] If the root mean square PW ems Greater than the pulse width threshold PW thr Then proceed to step (4.13);

[0027] Otherwise, proceed to step (4.14);

[0028] (4.13) Calculate the pulse width PW center value matching degree between the i-th class and the j-th class, and determine whether the pulse width PW of the i-th class and the j-th class match based on the pulse width PW center value matching degree. If they match, proceed to step (4.15); otherwise, proceed to step (3) for the next class.

[0029] (4.14) Determine the mean PW of the pulse width PW of the i-th class and the j-th class. mean If a match is found, proceed to step (4.15); otherwise, proceed to step (3) for the next class.

[0030] (4.15) Update the parameters of the i-th class;

[0031] (4.16) Then perform the judgment in step (3) on the next class;

[0032] (4.2) For the radio frequency RF, the matching judgment is performed using the same method as in step (4.1).

[0033] Furthermore, the method for determining whether the time compatibility of two adjacent classes meets the conditions in step (3) is as follows:

[0034] (3.1) Calculate the time matching degree Enij_toa:

[0035] Enij_toa=fabs(T i_end -Tj_Start ) / Max(fabs(T i_end- T i_Start ),fabs(T j_end- T j_Start ))

[0036] in:

[0037] T i_Start This represents the earliest arrival time of the pulse in the i-th type of signal sequence;

[0038] T i_end This represents the final end time of the pulse in the i-th type of signal sequence;

[0039] T j_Start This represents the earliest arrival time of the pulse in the j-th type of signal sequence;

[0040] T j_end This represents the final end time of the pulse in the j-th type of signal sequence;

[0041] fabs() represents taking the absolute value;

[0042] Max() means to retrieve the maximum value;

[0043] (3.2) Determine whether the time compatibility of two adjacent classes meets the conditions by comparing the time matching degree Enij_toa with the time matching degree threshold:

[0044] If the time matching degree Enij_toa is less than the time matching degree threshold, it means that the time compatibility of two adjacent classes meets the conditions.

[0045] Otherwise, the time compatibility of two adjacent classes will not meet the requirements.

[0046] Furthermore, step (4.13) includes:

[0047] Calculate the pulse width PW center value matching degree En_pw between the i-th class and the j-th class:

[0048] En_pw=fabs(PW m -PW n ) / PW thr

[0049] in:

[0050] PW m This represents the center value of the m centers in the i-th type of signal sequence;

[0051] PW n Let m represent the center value of the n centers in the j-th class of signal sequence; m > n.

[0052] If the pulse width PW center value matching degree En_pw is less than the center value matching degree threshold, it means that the pulse widths of the i-th class and the j-th class match.

[0053] Otherwise, the pulse widths of the i-th class and the j-th class will not match.

[0054] Furthermore, when there are multiple center values, the matching degree between the center value of the nth center of the jth class and the center value of the ith class is calculated one by one. When the number of center values ​​that are judged to be matched is greater than 2 / 3 of m, it is determined that the pulse width of the ith class and the jth class are matched.

[0055] Furthermore, updating the parameters of the i-th class in step (4.15) includes:

[0056] (a) Class central value update and merging:

[0057] If two adjacent classes are both centers of a single center, then the two adjacent classes will still be centers of a single center after being merged, and the distribution centers will be recounted.

[0058] If two adjacent classes are both central branches of multiple centers, then the union of the central values ​​of the two adjacent classes is taken.

[0059] (b) Recalculate the first arrival time and the last end time of the pulse in the signal sequence of the class.

[0060] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0061] This invention enables rapid, unsupervised clustering of multidimensional parameters of unknown electromagnetic signals in situations where parameters of multiple targets overlap and interweave. It employs expert rules and combines the characteristics of electromagnetic targets and the environment, as well as the characteristics of electromagnetic signal receiving equipment, to determine the initial parameters of DBSCAN. Furthermore, it evaluates and optimizes the preliminary clustering results from the perspective of the distribution of multidimensional signal parameters, greatly improving the accuracy and robustness of signal sequence clustering for unknown electromagnetic targets. Attached Figure Description

[0062] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a flowchart of the unknown radar signal clustering method combining DBscan with distributed posterior evaluation and optimization in an embodiment of the present invention.

[0064] Figure 2 This is a flowchart of the posterior evaluation in an embodiment of the present invention.

[0065] Figure 3 This is a diagram showing the signal parameters for scenario one in an application example.

[0066] Figure 4 This is a diagram showing the signal parameters for scenario two in an application example.

[0067] Figure 5 This is a diagram showing the signal parameters for scenario three in an application example.

[0068] Figure 6 This is a diagram illustrating the signal parameters for scenario four in an application example.

[0069] Figure 7 This is a diagram showing the signal parameters for scenario five in an application example. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0071] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0072] Example

[0073] For a radar signal sequence over a period of time, each pulse is described by multidimensional parameters, typically including five dimensions: RF (radio frequency), PW (pulse width), PRI (pulse repetition interval), AMP (signal amplitude), and DOA (angle of arrival). The radar signal sequence to be clustered consists of interleaved and superimposed sequences of N targets, where N is the number of clusters. For example... Figure 1 As shown, this embodiment proposes a clustering method for unknown radar signals that combines DBscan with distributed posterior evaluation and optimization, including:

[0074] S1, for two two-dimensional sequences composed of signal angle of arrival (DOA) and signal amplitude (AMP), based on the reconnaissance system performance, measurement error, and the reconnaissance system's measurements of DOA and AMP, a weighted distance D between the two sequences is defined, and two parameters Eps and MinPts for the Dbscan clustering operation are defined; where Eps represents the neighborhood radius and MinPts represents the density threshold; specifically:

[0075] The weighted distance D between the two two-dimensional sequences is defined as:

[0076] D=W_DOA*d(DOA) / max_DOA+ W_AMP *d(AMP) / max_AMP

[0077] in:

[0078] W_DOA is the weight of the signal angle of arrival (DOA), W_AMP is the weight of the signal amplitude (AMP), and W_DOA>0, W_AMP>0, W_DOA +W_AMP=1;

[0079] max_DOA is the maximum value of the signal angle of arrival (DOA), and max_AMP is the maximum value of the signal amplitude (AMP). max_DOA and max_AMP are determined by experts based on the performance of the reconnaissance system and the measurement error of the reconnaissance system.

[0080] d(DOA) is the absolute value of the difference between two elements in the two-dimensional sequence formed by the signal angle of arrival (DOA); d(AMP) is the absolute value of the difference between two elements in the two-dimensional sequence formed by the signal amplitude (AMP); d(DOA) and d(AMP) are obtained by the reconnaissance system from the measurements of the signal angle of arrival (DOA) and the signal amplitude (AMP).

[0081] The neighborhood radius Eps is defined as:

[0082] Eps= W_DOA*max_DOA+W_AMP*max_AMP.

[0083] The density threshold MinPts is determined by experts based on the electromagnetic signal environmental density and the length of the DBscan clustering operation. It typically ranges from 4 to 10.

[0084] S2, based on the defined weighted distance D, neighborhood radius Eps and density threshold MinPts, performs Dbscan clustering operation on the two two-dimensional sequences to obtain the clustering results; Dbscan clustering operation is an existing technology and will not be described in detail here.

[0085] S3 performs a posterior evaluation of the obtained clustering results and updates the parameters of the clustering results based on the posterior evaluation results. For example... Figure 2As shown, assuming the obtained clustering results have n classes, the posterior evaluation of the obtained clustering results includes:

[0086] (1) Calculate the earliest arrival time and the latest end time of the pulse in each type of signal sequence;

[0087] (2) Sort each class in ascending order of the earliest arrival time;

[0088] (3) Determine whether the time compatibility of two adjacent classes meets the condition. If the time compatibility of two adjacent classes meets the condition, perform subsequent operations; otherwise, perform the judgment in step (3) on the next class.

[0089] (3.1) Calculate the time matching degree Enij_toa:

[0090] Enij_toa=fabs(T i_end -T j_Start ) / Max(fabs(T i_end- T i_Start ),fabs(T j_end- T j_Start ))

[0091] in:

[0092] T i_Start This represents the earliest arrival time of the pulse in the i-th type of signal sequence;

[0093] T i_end This represents the final end time of the pulse in the i-th type of signal sequence;

[0094] T j_Start This represents the earliest arrival time of the pulse in the j-th type of signal sequence;

[0095] T j_end This represents the final end time of the pulse in the j-th type of signal sequence;

[0096] fabs() represents taking the absolute value;

[0097] Max() means to retrieve the maximum value;

[0098] (3.2) Determine whether the time compatibility of two adjacent classes meets the conditions by comparing the time matching degree Enij_toa with the time matching degree threshold:

[0099] If the time matching degree Enij_toa is less than the time matching degree threshold, it means that the time compatibility of two adjacent classes meets the conditions.

[0100] Otherwise, the time compatibility of two adjacent classes will not meet the requirements.

[0101] (4) Matching judgment is performed on the pulse width PW and radio frequency RF of two adjacent classes:

[0102] (4.1) For pulse width PW:

[0103] (4.11) For the pulse width PW of the pulse in the i-th class, calculate its mean PW. mean and root mean square PW ems ;

[0104] PW ems =sqrt(∑(PW i -PW mean ) 2 )

[0105] Where sqrt() represents taking the root mean square; PW i Let be the pulse width of the pulse in the i-th class;

[0106] (4.12) Determine the root mean square PW ems Is it greater than the pulse width threshold PW? thr :

[0107] If the root mean square PW ems Greater than the pulse width threshold PW thr Then proceed to step (4.13);

[0108] Otherwise, proceed to step (4.14);

[0109] (4.13) Calculate the pulse width PW center value matching degree En_pw between the i-th class and the j-th class:

[0110] En_pw=fabs(PW m -PW n ) / PW thr

[0111] in:

[0112] PW m This represents the center value of the m centers in the i-th type of signal sequence;

[0113] PW n Let m represent the center value of the n centers in the j-th class of signal sequence; m > n.

[0114] If the pulse width PW center value matching degree En_pw is less than the center value matching degree threshold, it means that the pulse widths of the i-th class and the j-th class match, and step (4.15) is executed;

[0115] Otherwise, if the pulse widths of the i-th class and the j-th class do not match, then the next class will be judged in step (3).

[0116] Specifically, when there are multiple center values, the matching degree between the center value of the nth center of the jth class and the center value of the ith class is calculated one by one. When the number of center values ​​that are judged to be matched is greater than 2 / 3 of m, the pulse width of the ith class and the jth class are judged to be matched.

[0117] (4.14) Determine the mean PW of the pulse width PW of the i-th class and the j-th class. mean Whether it matches (the matching judgment condition is described in (4.13), the matching threshold can be manually adjusted, and the threshold values ​​of En_pw and En_rf in this embodiment are 1.5), if it matches, then execute step (4.15), otherwise perform the judgment of step (3) on the next class;

[0118] (4.15) Update the parameters of the i-th class:

[0119] (a) Class central value update and merging:

[0120] If two adjacent classes are both centers of a single center, then the two adjacent classes will still be centers of a single center after being merged, and the distribution centers will be recounted.

[0121] If two adjacent classes are both central branches of multiple centers, then the union of the central values ​​of the two adjacent classes is taken.

[0122] (b) Recalculate the first arrival time and the last end time of the pulse in the signal sequence of the class.

[0123] (4.16) Then perform the judgment in step (3) on the next class;

[0124] (4.2) For the radio frequency RF, the matching judgment is performed using the same method as in step (4.1), which will not be repeated here.

[0125] Application example:

[0126] Multiple tests were conducted to verify the dense radar targets of fishing vessels in the frequency band of fishing vessels under actual conditions, such as... Figure 3 , Figure 4 , Figure 5 , Figure 6 , Figure 7 As shown, each image depicts a dense signal scenario. In actual combat environments, a small portion of signals cannot be correctly classified by manual analysis due to false alarms in signal detection or excessive measurement errors; these signals are not included in the statistics. Below is a comparison of the effects in some scenarios:

[0127] Table 1 records the details of five scenarios, including the total length of each scenario sequence and the number of targets contained therein; as well as the results of the method of this invention and the commonly used histogram method, and their comparison.

[0128] Table 1:

[0129]

[0130] As can be seen from the above, the clustering method proposed in this invention can be used to cluster densely intersecting unknown radar targets. Through statistical analysis and comparison of the accuracy and recall rates of multiple dense fishing boat scenes, the clustering performance is greatly improved compared with the histogram-based clustering method, thus verifying the effectiveness and advantages of this method.

[0131] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A clustering method for unknown radar signals combining DBscan with distributed posterior evaluation and optimization, characterized in that, include: For two two-dimensional sequences consisting of two parameters, namely the angle of arrival (DOA) and the amplitude of signal (AMP), the weighted distance D between the two two-dimensional sequences is defined based on the reconnaissance system performance, the reconnaissance system measurement error, and the reconnaissance system's measurement values ​​of DOA and AMP. Two parameters, Eps and MinPts, are also defined for the Dbscan clustering operation. Here, Eps represents the neighborhood radius and MinPts represents the density threshold. Based on the defined weighted distance D, neighborhood radius Eps, and density threshold MinPts, Dbscan clustering operation is performed on the two two-dimensional sequences to obtain the clustering results; The obtained clustering results are subjected to posterior evaluation, and the parameters of the clustering results are updated based on the posterior evaluation results. Assuming the obtained clustering results have n classes, the posterior evaluation of the obtained clustering results includes: (1) Calculate the earliest arrival time and the latest end time of the pulse in each type of signal sequence; (2) Sort each class in ascending order of the earliest arrival time; (3) Determine whether the time compatibility of two adjacent classes meets the conditions. If the time compatibility of two adjacent classes meets the conditions, perform subsequent operations; otherwise, perform the judgment in step (3) on the next class. (4) Matching judgment is performed on the pulse width PW and radio frequency RF of two adjacent classes: (4.1) For pulse width PW: (4.11) For the pulse width PW of the pulse in the i-th class, calculate its mean PW. mean and root mean square PW ems ; (4.12) Determine the root mean square PW ems Is it greater than the pulse width threshold PW? thr : If the root mean square PW ems Greater than the pulse width threshold PW thr Then proceed to step (4.13); Otherwise, proceed to step (4.14); (4.13) Calculate the pulse width PW center value matching degree between the i-th class and the j-th class, and determine whether the pulse width PW of the i-th class and the j-th class match based on the pulse width PW center value matching degree. If they match, proceed to step (4.15); otherwise, proceed to step (3) for the next class. Calculate the pulse width PW center value matching degree En_pw between the i-th class and the j-th class: En_pw=fabs(PW m -PW n ) / PW thr in: PW m This represents the center value of the m centers in the i-th type of signal sequence; PW n This represents the center value of the n centers in the j-th type of signal sequence; m > n; If the pulse width PW center value matching degree En_pw is less than the center value matching degree threshold, it means that the pulse widths of the i-th class and the j-th class match. Otherwise, the pulse widths of the i-th class and the j-th class do not match; (4.14) Determine the mean PW of the pulse width PW of the i-th class and the j-th class. mean If a match is found, proceed to step (4.15); otherwise, proceed to step (3) for the next class. (4.15) Update the parameters of the i-th class: (a) Class central value update and merging: If two adjacent classes are both centers of a single center, then the two adjacent classes will still be centers of a single center after being merged, and the distribution centers will be recounted. If two adjacent classes are both central branches of multiple centers, then the union of the central values ​​of the two adjacent classes is taken. (b) Recalculate the first arrival time and last end time of the pulses in the signal sequence of the class; (4.16) Then perform the judgment in step (3) on the next class; (4.2) For the radio frequency RF, the matching judgment is performed using the same method as in step (4.1).

2. The unknown radar signal clustering method combining DBscan with distributed posterior evaluation and optimization according to claim 1, characterized in that, The weighted distance D between the two two-dimensional sequences is defined as: D=W_DOA*d(DOA) / max_DOA+ W_AMP *d(AMP) / max_AMP in: W_DOA is the weight of the signal angle of arrival (DOA), W_AMP is the weight of the signal amplitude (AMP), and W_DOA>0, W_AMP>0, W_DOA +W_AMP=1; max_DOA is the maximum value of the signal angle of arrival (DOA), and max_AMP is the maximum value of the signal amplitude (AMP). max_DOA and max_AMP are determined by experts based on the performance of the reconnaissance system and the measurement error of the reconnaissance system. d(DOA) is the absolute value of the difference between two elements in the two-dimensional sequence formed by the signal angle of arrival (DOA); d(AMP) is the absolute value of the difference between two elements in the two-dimensional sequence formed by the signal amplitude (AMP); d(DOA) and d(AMP) are obtained by the reconnaissance system from the measurements of the signal angle of arrival (DOA) and the signal amplitude (AMP).

3. The unknown radar signal clustering method combining DBscan with distributed posterior evaluation and optimization according to claim 2, characterized in that, The neighborhood radius Eps is defined as: Eps=W_DOA*max_DOA+W_AMP*max_AMP.

4. The unknown radar signal clustering method combining DBscan with distributed posterior evaluation and optimization according to claim 1, characterized in that, The density threshold MinPts was determined by experts based on the electromagnetic signal environment density and the length of the DBscan clustering operation.

5. The unknown radar signal clustering method combining DBscan with distributed posterior evaluation and optimization according to claim 1, characterized in that, The method for determining whether the time compatibility of two adjacent classes meets the conditions in step (3) is as follows: (3.1) Calculate the time matching degree Enij_toa: Enij_toa=fabs(T i_end -T j_Start ) / Max(fabs(T i_end- T i_Start ),fabs(T j_end- T j_Start )) in: T i_Start This represents the earliest arrival time of the pulse in the i-th type of signal sequence; T i_end This represents the final end time of the pulse in the i-th type of signal sequence; T j_Start This represents the earliest arrival time of the pulse in the j-th type of signal sequence; T j_end This represents the final end time of the pulse in the j-th type of signal sequence; fabs() represents taking the absolute value; Max() means to retrieve the maximum value; (3.2) Determine whether the time compatibility of two adjacent classes meets the conditions by comparing the time matching degree Enij_toa with the time matching degree threshold: If the time matching degree Enij_toa is less than the time matching degree threshold, it means that the time compatibility of two adjacent classes meets the conditions. Otherwise, the time compatibility of two adjacent classes will not meet the requirements.

6. The unknown radar signal clustering method combining DBscan with distributed posterior evaluation and optimization according to claim 5, characterized in that, When there are multiple center values, calculate the matching degree between the center value of the nth center of the jth class and the center value of the ith class one by one. When the number of center values ​​that are judged to be matched is greater than 2 / 3 of m, it is determined that the pulse width of the ith class and the jth class are matched.

Images