A method, system and terminal for detecting the number of radar signal sources based on density clustering

Abstract

The present invention discloses a method, system and terminal for detecting the number of radar signal sources based on density clustering. The method includes: generating an antenna array received data representation according to the target radar antenna array information, obtaining a target covariance matrix according to the antenna array received data representation, performing a unitary transformation on the target covariance matrix to obtain a target transformation matrix, and generating a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem; obtaining a noise background, and when the noise background is a Gaussian colored noise background, compressing and normalizing the radius vector to obtain a target radius matrix; clustering the target radius matrix according to a target clustering algorithm to obtain a clustering result; obtaining a target cluster according to the clustering result, and obtaining the number of radar signal sources according to the target cluster. The present invention improves the accuracy of estimating the number of signal sources under low signal-to-noise ratio and colored noise.

Description

Technical Field

[0001] The present invention relates to the field of array signal processing, and in particular to a method, system and terminal for detecting the number of radar signal sources based on density clustering. Background Art

[0002] In the technology of array radar signal processing, signal source angle estimation, that is, the estimation of the direction of arrival (DOA) of radar echo signals, is a key issue. Many spatial angle estimation algorithms, such as Multiple Signal Classification (MUSIC), Estimation of Signal Parameters via Rotational Invariance Techniques (ESPRIT), etc., rely on the accurate number of radar signal sources (radar echo signals) as prior input. However, in actual applications, the number of signal sources is often unknown. Therefore, it is necessary to accurately estimate the number of signal sources.

[0003] Currently, various proposed methods for estimating the number of signal sources have poor performance under colored noise conditions. Although the Gerschgorin circle method can estimate the number of signal sources under spatial colored noise, its accuracy is not high under low signal-to-noise ratio conditions. Therefore, it will affect the estimation of the number of signal sources, resulting in inaccurate estimation of the direction of arrival of radar echo signals.

[0004] Therefore, the existing technology still needs to be improved and developed. Summary of the Invention

[0005] The main purpose of the present invention is to provide a method, system, terminal and computer-readable storage medium for detecting the number of radar signal sources based on density clustering, aiming to solve the problem that various existing methods for estimating the number of signal sources have poor performance under colored noise conditions. Although the Gerschgorin circle method can estimate the number of signal sources under spatial colored noise, its accuracy is not high under low signal-to-noise ratio conditions. Therefore, it will affect the estimation of the number of signal sources, resulting in inaccurate estimation of the direction of arrival of radar echo signals.

[0006] To achieve the above object, the present invention provides a method for detecting the number of radar signal sources based on density clustering. The method for detecting the number of radar signal sources based on density clustering includes the following steps:

[0007] Generate an antenna array received data representation according to the target radar antenna array information, obtain a target covariance matrix according to the antenna array received data representation, perform a unitary transformation on the target covariance matrix to obtain a target transformation matrix, and generate a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem;

[0008] Obtain the noise background. When the noise background is a Gaussian colored noise background, compress and normalize the radius vector to obtain a target radius matrix;

[0009] Cluster the target radius matrix according to a target clustering algorithm to obtain a clustering result;

[0010] Obtain a target cluster according to the clustering result, and obtain the number of radar signal sources according to the target cluster.

[0011] Optionally, the generating an antenna array received data representation according to the target radar antenna array information, obtaining a target covariance matrix according to the antenna array received data representation, performing a unitary transformation on the target covariance matrix to obtain a target transformation matrix, and generating a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem specifically includes:

[0012] Construct an array steering vector corresponding to each signal source according to the target radar antenna array information, generate a steering vector matrix according to all the array steering vectors, and generate the antenna array received data representation according to the steering vector matrix;

[0013] Obtain the target covariance matrix and the covariance matrix of the radar echo signal according to the antenna array received data representation;

[0014] Perform a unitary transformation on the target covariance matrix according to the covariance matrix of the radar echo signal to obtain the target transformation matrix;

[0015] According to the Gerschgorin circle theorem, obtain the Gerschgorin circle radius of each Gerschgorin circle from the target transformation matrix, and generate the radius vector according to all the Gerschgorin circle radii.

[0016] Optionally, the performing a unitary transformation on the target covariance matrix according to the covariance matrix of the radar echo signal to obtain the target transformation matrix specifically includes:

[0017] Block the target covariance matrix according to the covariance matrix of the radar echo signal, and obtain a principal submatrix according to the blocked result;

[0018] Perform an eigenvalue decomposition on the principal submatrix to obtain the eigenspace vectors of the principal submatrix;

[0019] Construct a unitary transformation matrix according to the eigenspace vectors, and perform a unitary transformation operation on the eigenspace vectors according to the unitary transformation matrix to obtain the target transformation matrix.

[0020] Optionally, the obtaining the noise background. When the noise background is a Gaussian colored noise background, compress and normalize the radius vector to obtain a target radius matrix specifically includes:

[0021] Obtain a noise background and determine the type of the noise background;

[0022] When the noise background is a Gaussian colored noise background, compress each Gerschgorin circle radius in the radius vector according to the unitary transformation matrix and the center of each Gerschgorin circle to obtain a plurality of compressed radii;

[0023] Perform a normalization process on each of the compressed radii to obtain the target radius of each Gerschgorin circle;

[0024] Generate the target radius matrix according to all the target radii.

[0025] Optionally, after obtaining the noise background and determining the type of the noise background, it further includes:

[0026] When the noise background is a Gaussian white noise background, perform a logarithmic transformation on the radius vector to obtain a logarithmic transformation matrix;

[0027] Cluster the logarithmic transformation matrix according to a target clustering algorithm to obtain a white noise clustering result;

[0028] Obtain a white noise target cluster according to the white noise clustering result, and obtain the corresponding number of radar signal sources according to the white noise target cluster.

[0029] Optionally, clustering the target radius matrix according to a target clustering algorithm to obtain a clustering result specifically includes:

[0030] Obtain a preset distance threshold and a minimum neighborhood number, and use the target radius matrix, the distance threshold, and the minimum neighborhood number as inputs for clustering;

[0031] During the clustering process, traverse each element in the target radius matrix, select core objects according to the minimum neighborhood number, add them to the core object set and generate an ordered arrangement, calculate the core distance and reachable distance of each core object, and divide all the core objects into one of the current cluster clustering, new clustering cluster, and noise points according to the core distance and reachable distance of each core object in the ordered arrangement;

[0032] Generate a clustering result according to the current cluster clustering, the new clustering cluster, and the noise points.

[0033] Optionally, obtaining a target cluster according to the clustering result and obtaining the number of radar signal sources according to the target cluster specifically includes:

[0034] Calculate the average value of the current cluster clustering and the new clustering cluster in the clustering result, and select the cluster with the larger average value as the target cluster;

[0035] Count the number of elements in the target cluster and output this number as the number of radar signal sources.

[0036] In addition, to achieve the above object, the present invention also provides a radar signal source number detection system based on density clustering, wherein the radar signal source number detection system based on density clustering includes:

[0037] A radius vector generation module, configured to generate a representation of antenna array received data according to target radar antenna array information, obtain a target covariance matrix according to the representation of antenna array received data, perform a unitary transformation on the target covariance matrix to obtain a target transformation matrix, and generate a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem;

[0038] A target radius matrix generation module, configured to obtain a noise background, and when the noise background is a Gaussian colored noise background, compress and standardize the radius vector to obtain a target radius matrix;

[0039] A clustering result generation module, configured to perform clustering on the target radius matrix according to a target clustering algorithm to obtain a clustering result;

[0040] A result output module, configured to obtain a target cluster according to the clustering result and obtain the number of radar signal sources according to the target cluster.

[0041] In addition, to achieve the above object, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a radar signal source number detection program based on density clustering stored on the memory and executable on the processor. When the radar signal source number detection program based on density clustering is executed by the processor, the steps of the above-mentioned radar signal source number detection method based on density clustering are implemented.

[0042] In addition, to achieve the above object, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a radar signal source number detection program based on density clustering. When the radar signal source number detection program based on density clustering is executed by a processor, the steps of the above-mentioned radar signal source number detection method based on density clustering are implemented.

[0043] In the present invention, an antenna array received data representation is generated according to the target radar antenna array information, a target covariance matrix is obtained based on the antenna array received data representation, a unitary transformation is performed on the target covariance matrix to obtain a target transformation matrix, and a radius vector corresponding to the target transformation matrix is generated according to the Gerschgorin circle theorem; a noise background is acquired, and when the noise background is a Gaussian colored noise background, the radius vector is compressed and normalized to obtain a target radius matrix; a target clustering algorithm is used to cluster the target radius matrix to obtain a clustering result; a target cluster is obtained according to the clustering result, and the number of radar signal sources is obtained according to the target cluster. In the present invention, the Gerschgorin circle radii obtained after performing a unitary transformation on the signal covariance matrix under a Gaussian colored noise background are compressed and normalized, thereby reducing redundant information and improving the discrimination of noise signals. Furthermore, a target clustering algorithm is used to cluster the target radius matrix, and the corresponding radii are divided into a noise cluster and a signal cluster through clustering, thereby effectively distinguishing noise and signals, and thus improving the accuracy of estimating the number of signal sources under low signal-to-noise ratio and colored noise conditions. Description of the Drawings

[0044] Figure 1 is a flowchart of a preferred embodiment of the method for detecting the number of radar signal sources based on density clustering according to the present invention;

[0045] Figure 2 is a schematic diagram of the Gerschgorin disk of the original covariance matrix in the method for detecting the number of radar signal sources based on density clustering according to the present invention;

[0046] Figure 3 is a schematic diagram of the Gerschgorin disk of the target transformation matrix after unitary transformation in the method for detecting the number of radar signal sources based on density clustering according to the present invention;

[0047] Figure 4 is a schematic diagram of the performance comparison of the method for estimating the number of signal sources under a white noise background in the method for detecting the number of radar signal sources based on density clustering according to the present invention;

[0048] Figure 5 is a schematic diagram of the performance comparison of the method for estimating the number of signal sources under a colored noise background in the method for detecting the number of radar signal sources based on density clustering according to the present invention;

[0049] Figure 6 is a structural diagram of a preferred embodiment of the system for detecting the number of radar signal sources based on density clustering according to the present invention;

[0050] Figure 7 is a structural diagram of a preferred embodiment of the terminal according to the present invention. Detailed Embodiments

[0051] To make the objectives, technical solutions and advantages of the present invention more clear and definite, the following further describes the present invention in detail with reference to the accompanying drawings and by way of examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0052] In the field of array radar signal processing technology, signal source angle estimation, that is, the estimation of the arrival angle of radar echo signals, is a key issue. Many spatial angle estimation algorithms, such as Multiple Signal Classification (MUSIC), Estimation of Signal Parameters via Rotational Invariance Techniques (ESPRIT), etc., rely on the accurate number of radar signal sources as prior input. However, in practical applications, the number of signal sources is often unknown. Therefore, it is necessary to accurately estimate the number of signal sources. For the problem of estimating the number of radar signal sources, in current various methods for estimating the number of signal sources, including information-theoretic methods such as the Akaike Information Criterion (AIC), the Minimum Description Length (MDL) criterion, and the Effective Detection Criterion (EDC), the Spatial Smoothing Rank (SSR) method, and the Gerschgorin Disk Estimator (GDE), etc. In the case of spatial white noise, the MDL criterion is a consistent estimator according to information-theoretic methods, but the AIC criterion is not. Therefore, in the case of large samples, the AIC criterion has the problem of overestimation. And because these two methods are derived based on the white noise signal model, their performance drops significantly under colored noise conditions and may even completely fail. Although the Gerschgorin disk method can estimate the number of signal sources under spatial colored noise, its accuracy is relatively low under low signal-to-noise ratio conditions. Therefore, it will affect the estimation of the number of signal sources, resulting in inaccurate estimation of the arrival angle of radar echo signals.

[0053] To address one or more of the above problems, the present invention generates a representation of the antenna array received data based on the target radar antenna array information, obtains the target covariance matrix from the representation of the antenna array received data, performs a unitary transformation on the target covariance matrix to obtain the target transformation matrix, and generates a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem; obtains the noise background, and when the noise background is a Gaussian colored noise background, compresses and normalizes the radius vector to obtain the target radius matrix; performs clustering on the target radius matrix according to the target clustering algorithm to obtain the clustering result; obtains the target cluster according to the clustering result, and obtains the number of radar signal sources according to the target cluster.

[0054] The method for detecting the number of radar signal sources based on density clustering according to a preferred embodiment of the present invention, such as Figure 1As shown in the figure, the method for detecting the number of radar signal sources based on density clustering includes the following steps:

[0055] Step S10: Generate an antenna array received data representation according to the target radar antenna array information, obtain the target covariance matrix based on the antenna array received data representation, perform a unitary transformation on the target covariance matrix to obtain a target transformation matrix, and generate a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem.

[0056] It should be noted that in the case of low signal-to-noise ratio and spatial colored noise, the present invention solves the problem of estimating the number of array radar signal sources. Among them, in the present invention, the target radar antenna array information includes the number of array elements, radar echo signals, angles, radar echo wavelengths, and antenna element intervals. The antenna array received data representation is obtained corresponding to the target radar antenna array information, and then the corresponding processing process in the present invention is applied to obtain the number of radar signal sources.

[0057] Further, the step of generating an antenna array received data representation according to the target radar antenna array information, obtaining the target covariance matrix based on the antenna array received data representation, performing a unitary transformation on the target covariance matrix to obtain a target transformation matrix, and generating a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem specifically includes:

[0058] Construct an array steering vector corresponding to each signal source according to the target radar antenna array information, generate a steering vector matrix based on all the array steering vectors, and generate the antenna array received data representation according to the steering vector matrix;

[0059] Obtain the target covariance matrix and the radar echo signal covariance matrix based on the antenna array received data representation;

[0060] Perform a unitary transformation on the target covariance matrix according to the radar echo signal covariance matrix to obtain the target transformation matrix;

[0061] According to the Gerschgorin circle theorem, obtain the Gerschgorin circle radius of each Gerschgorin circle from the target transformation matrix, and generate the radius vector according to all the Gerschgorin circle radii.

[0062] Specifically, for a radar antenna array, it is composed of array elements arranged in a straight line. There are radar echo signals incident on the array from an angle . The radar signal echo wavelength is , and the antenna element interval is . Then the array steering vector corresponding to the (Array flow pattern) can be expressed by the following formula (1):

[0063] ; (1)

[0064] Where, represents exponentiation, is the representation of the complex number unit, , represents the transpose operation.

[0065] According to all the array steering vectors, a steering vector matrix can be generated. The steering vector matrix corresponding to the signals received by the antenna array is expressed by the following formula (2):

[0066] ; (2)

[0067] Therefore, at moment, the data received by the antenna array (containing radar echo signals and noise) can be represented by the data received by the antenna array, as shown in the following formula (3):

[0068] ; (3)

[0069] Where, is the radar echo signal incident on the array at moment, which is a dimensional column vector, is Gaussian white noise or colored noise with cross-correlation, which is a and are respectively recorded as the following formula (4) and formula (5):

[0070] ; (4)

[0071] ; (5)

[0072] Where, represents the first radar echo signal incident on the array at moment, represents the second radar echo signal incident on the array at moment, represents the th radar echo signal incident on the array at moment, The Gaussian white noise or the colored noise with cross-correlation corresponding to the second array element at a certain moment denote at a certain moment the Gaussian white noise or the colored noise with cross-correlation corresponding to the

[0073] Then for the covariance matrix in the representation of the received data of the antenna array (i.e., the target covariance matrix) and the covariance matrix of (i.e., the covariance matrix of the radar echo signal), they are respectively represented by the following formulas (6) and (7):

[0074] ; (6)

[0075] ; (7)

[0076] In the above formula, denotes conjugate transpose, denotes the number of time-domain samples, i.e., the number of snapshots.

[0077] Furthermore, the unitary transformation of the target covariance matrix according to the covariance matrix of the radar echo signal to obtain the target transformation matrix specifically includes:

[0078] Partition the target covariance matrix according to the covariance matrix of the radar echo signal, and obtain the principal submatrix according to the partition result;

[0079] Perform eigenvalue decomposition on the principal submatrix to obtain the eigen-space vectors of the principal submatrix;

[0080] Construct a unitary transformation matrix according to the eigen-space vectors, and perform a unitary transformation operation on the eigen-space vectors according to the unitary transformation matrix to obtain the target transformation matrix.

[0081] It should be noted that the present invention uses the Gerschgorin circle method. Without knowing the exact values of the eigenvalues, based on the Gerschgorin disk theorem, by estimating the distribution range of the eigenvalues, the estimation of the number of signal sources is realized. Specifically, it can be assumed that is a dimensional matrix, the element in the row and the column is . Sum the absolute values of the elements in the row after removing the element in the row and the column to obtain the radius of the disk, which is expressed as the following formula (8):

[0082] ; (8)

[0083] Then the corresponding disc is represented by the following formula (9) in the complex plane:

[0084] ; (9)

[0085] where is the center of the th disc, which is a Gerschgorin disc, and the symbol represents the complex number satisfying formula (8) in the complex plane.

[0086] When drawing the Gerschgorin discs of the covariance matrix as shown in Figure 2 , it can be found that both the radius of the Gerschgorin disc of the radar echo signal and the radius of the Gerschgorin disc of the noise are significantly large, and the central positions of all Gerschgorin discs almost coincide, which indicates that using the current covariance matrix to estimate the number of signal sources will lack resolution and accurate results cannot be obtained. In view of this situation, the present invention needs to separate the signal discs and the noise discs.

[0087] Specifically, in order to separate the signal discs and the noise discs, the present invention needs to perform a unitary transformation on the covariance matrix . First, is partitioned to obtain the partitioned result, which is expressed as formula (10):

[0088] : (10)

[0089] where is the -dimensional principal submatrix, is the first elements of the th column in , and is the element in the last row and last column of the matrix .

[0090] After that, the principal submatrix is subjected to eigenvalue decomposition to obtain formula (11):

[0091] : (11)

[0092] where is the diagonal matrix composed of the eigenvalues of , that is , Indicates generating a diagonal matrix with the elements of a vector as diagonal elements, Indicates the first eigenvalue of Indicates the second eigenvalue of Indicates the third eigenvalue of Indicates the th eigenvalue of is the eigenvector space of , that is Indicates the eigenvector space of the first eigenvector in Indicates the eigenvector space of the second eigenvector in Indicates the eigenvector space of the th eigenvector in. According to the properties of eigenvectors, we can obtain , where is the identity matrix of dimension

[0093] Furthermore, according to the eigenvector space of construct the unitary transformation matrix , and the corresponding unitary transformation matrix is expressed as the following formula (12):

[0094] ; (12)

[0095] According to the unitary transformation operation on the covariance matrix , obtain the matrix after unitary transformation, that is, the target transformation matrix , which is expressed as the following formula (13):

[0096] ; (13)

[0097] Among them, , is the matrix formed by taking the first rows and the first columns of the steering vector matrix , represents complex conjugate, is the first parameter, which is expressed as the following formula (14):

[0098] ; (14)

[0099] Among them, , , .

[0100] Furthermore, in formula (13), is the radius element, expressed as the following formula (15):

[0101] ; (15)

[0102] Among them, is 's th eigenvector.

[0103] According to the Gerschgorin circle theorem, in formula (13), that is, the target transformation matrix, the Gerschgorin circle radius of is expressed as the following formula (16):

[0104] ; (16)

[0105] Among them, represents the absolute value operation on elements or vectors, and elements in matrices.

[0106] Then, according to 's Gerschgorin circle radii, a corresponding radius vector can be formed, expressed as the following formula (17):

[0107] ; (17)

[0108] For each Gerschgorin circle radius, according to the inequality principle, the following formula (18) can be obtained:

[0109] ; (18)

[0110] Among them, is a row vector.

[0111] It can be obtained from formula (18) that is independent of , indicating that the Gerschgorin circle radius only depends on . When the eigenvector takes the eigenvector of noise, it is orthogonal to the array manifold. Therefore, 's elements are close to zero. When the eigenvector takes the eigenvector of the radar echo signal, 's elements are significantly greater than zero.

[0112] Based on the Gerschgorin disk theorem, correspondingly draw the Gerschgorin disk, as Figure 3 shown, it can be seen from the figure that after the unitary transformation in the generated Gerschgorin disk, the disk radius corresponding to the radar echo signal is significantly larger than the disk radius corresponding to the noise, and the separation degree between the disk centers is relatively high. This result shows that the matrix after the unitary transformation the generated Gerschgorin disk has the distribution characteristics of high contrast, low overlap, and strong separability.

[0113] Step S20: Obtain the noise background. When the noise background is a Gaussian colored noise background, compress and standardize the radius vector to obtain a target radius matrix.

[0114] Specifically, the characteristics of the Gerschgorin disk generated by the target transformation matrix meet the requirements of the Ordering Points To Identify the Clustering Structure (OPTICS) algorithm for density separability and cluster spacing. Using this data as the input of the OPTICS algorithm can significantly improve the reliability and accuracy of radar signal source number estimation. Therefore, the present invention performs OPTICS clustering on the extracted radii, where the OPTICS algorithm is the target clustering algorithm. Before clustering, in order to eliminate the influence of data scale and provide more standardized data input for the OPTICS clustering algorithm, the present invention performs corresponding processing based on the noise background.

[0115] Further, the obtaining of the noise background, when the noise background is a Gaussian colored noise background, compressing and standardizing the radius vector to obtain a target radius matrix specifically includes:

[0116] Obtain the noise background and determine the type of the noise background;

[0117] When the noise background is a Gaussian colored noise background, compress each Gerschgorin disk radius in the radius vector according to the unitary transformation matrix and the center of each Gerschgorin disk to obtain a plurality of compressed radii;

[0118] Perform standardization processing on each of the compressed radii to obtain the target radius of each Gerschgorin disk;

[0119] Generate the target radius matrix according to all the target radii.

[0120] Specifically, when the noise background is a Gaussian colored noise background, then compress each Gerschgorin disk radius in the radius vector according to the unitary transformation matrix and the center of each Gerschgorin disk to obtain a plurality of compressed radii , that is, compress through the following formula (19):

[0121] ; (19)

[0122] Among them, , is the center of the Gerschgorin circle, is the element at the end row and end column of the matrix after unitary transformation. By compressing the ratio of to , the numerical range of the radius can be effectively reduced, thereby reducing the dynamic range of the data.

[0123] After that, each of the compressed radii is normalized to obtain the target radius of each Gerschgorin circle, that is, the normalization is performed through the following formula (20):

[0124] ; (20)

[0125] Among them, represents returning the maximum value of the elements in a set, represents selecting the maximum value from all the compressed radii, normalizing the compressed radii, and normalizing the radius values to the range of , which helps to eliminate the influence of the data scale.

[0126] After that, the obtained target radii are summarized to obtain a target radius matrix, as shown in formula (21):

[0127] ; (21)

[0128] Among them, is the target radius matrix.

[0129] Furthermore, the obtaining of the noise background and the judgment of the type of the noise background further include:[[]]

[0130] When the noise background is a Gaussian white noise background, perform a logarithmic transformation on the radius vector to obtain a logarithmic transformation matrix;

[0131] Cluster the logarithmic transformation matrix according to the target clustering algorithm to obtain a white noise clustering result;

[0132] Obtain a white noise target cluster according to the white noise clustering result, and obtain the corresponding number of radar signal sources according to the white noise target cluster.

[0133] Specifically, when the noise background is a Gaussian white noise background, perform a logarithmic transformation on the radius vector to obtain a logarithmic transformation matrix , that is, perform a logarithmic transformation through formula (22):

[0134] ; (22)

[0135] After that, the obtained logarithmic transformation matrix is clustered by the target clustering algorithm, and the white noise clustering result is obtained, so as to obtain the number of radar signal sources at this time. The clustering process and the process of obtaining the number of radar signal sources according to the corresponding clusters are the same as the processing process of the target radius matrix.

[0136] Step S30: Cluster the target radius matrix according to the target clustering algorithm to obtain a clustering result.

[0137] Specifically, after eliminating the influence of the data scale, the target radius matrix is clustered by the target clustering algorithm to distinguish the signal component and the noise component.

[0138] Further, the clustering the target radius matrix according to the target clustering algorithm to obtain a clustering result specifically includes:

[0139] Obtain a preset distance threshold and a minimum neighborhood number, and use the target radius matrix, the distance threshold, and the minimum neighborhood number as inputs for clustering;

[0140] During the clustering process, traverse each element in the target radius matrix, select core objects according to the minimum neighborhood number, add them to the core object set and generate an ordered arrangement, calculate the core distance and reachable distance of each core object, and according to the core distance and the reachable distance of each core object in the ordered arrangement, divide all the core objects into one of the current cluster clustering, new clustering cluster, and noise points respectively;

[0141] Generate a clustering result according to the current cluster clustering, the new clustering cluster, and the noise points.

[0142] Specifically, in the present invention, the target clustering algorithm is the OPTICS clustering algorithm. For the OPTICS clustering algorithm, given a data set , sample point , and a distance threshold and the minimum neighborhood number MinPts, the relevant definitions of OPTICS are as follows:

[0143] -neighborhood, for the sample point , its -neighborhood is defined as the set of sample points whose distance from is within , that is , the number of the sample set is denoted as , represents from the sample point to the sample point distance

[0144] The core object, for , The condition for being a core object is that its -neighborhood corresponding sample set contains no less than MinPts sample points;

[0145] Density direct reach, if is a core object, and the sample point is located in 's -neighborhood, then it is said that is density directly reached by ;

[0146] Density reachable, for and , there exists a sample sequence , if it satisfies , , and the in the sample sequence is density directly reached by , then it is said that is density reachable by ;

[0147] Density connected, for two points and , if there exists a core object , and are both density reachable by , then it is said that and are density connected;

[0148] In the neighborhood of the current core point , the minimum neighborhood radius that makes become a core point is called the core distance of , defined by the following formula (23):

[0149] ; (23)

[0150] where represents the node that is the MinPts-th nearest neighbor to the node in the set . For the sample point , when the number of sample points in its -neighborhood is less than MinPts, then does not meet the condition of being a core point, and the core distance Undefined, so it is marked as , denotes to the distance.

[0151] For , Regarding the reachable distance is defined by the following formula (24):

[0152] ; (24)

[0153] For two sample points and , if the number of sample points within the -neighborhood of is less than MinPts, then Regarding the reachable distance is undefined, so it is marked as , denotes the core distance of denotes to the distance.

[0154] Specifically, in the present invention, the radii in the target radius matrix are classified by the OPTICS clustering algorithm. Taking the target radius matrix obtained under the colored noise background as an example, the goal of the OPTICS algorithm is to output an ordered arrangement and two attributes of each Gerschgorin circle target radius , namely the core distance set and the reachable distance set.

[0155] During the clustering process, the specific steps are as follows:

[0156] Step 1, Input: The present invention inputs and the neighborhood parameter into the algorithm as a given sample set, where the neighborhood parameter includes the distance threshold and the minimum neighborhood number (MinPts). is infinite, so each sample point is a core object. If MinPts is 2, then the core distance of each core object is the distance from the nearest sample point to itself, and the sample point itself is also one of the neighborhood elements.

[0157] Step 2, Initialization: Traverse all elements in It is empty, and start traversing each element in the sample set.

[0158] Step 3, processing: First, traverse all elements in the sample set, calculate the core distance of each element. If its core distance exists, then use it as a core object and put it into the core object set If the core distance does not exist, it is not a core object.

[0159] After all elements in the sample set have been processed, randomly select an unprocessed core object from the core object set Push it into the ordered list, and at the same time check the neighborhood of this core object, calculate the reachable distance from the unvisited points in this neighborhood to the core object, and put these points into the seed set in turn according to the size of the reachable distance.

[0160] If the seed set is empty, randomly select the next unprocessed core object from the core object set for processing. If the seed object is not empty, obtain the point with the smallest reachable distance and unvisited status from the seed set, mark its status as visited, and push it into the ordered rehearsal column If the obtained neighborhood point is a core object at this time, then correspondingly add the unvisited neighborhood points of this point to and recalculate the reachable distance between the remaining points in the seed set and this point, and rearrange them in ascending order of reachable distance, continue to judge whether the seed set is empty and repeat the above steps; until the status of all points in the seed set is visited, randomly select the next unprocessed core object from the core object set for processing. When all elements in the core object set have been processed, then end this step and output the corresponding ordered arrangement , the core distance set and the reachable distance set.

[0161] Step 4, output: After step 3 ends, obtain the ordered arrangement , the core distance set and the reachable distance set, and then input a threshold of the reachable distance , in the ordered arrangement obtained after the algorithm execution, take out the sample points in order: If the reachable distance of this point , then it belongs to the current clustering cluster; if the reachable distance of this point is greater than , the core distance of this point , it is a new clustering cluster; if the reachable distance of this point is greater than , and the core distance of this point , it is a noise point.

[0162] Finally, the current cluster clustering, new cluster clustering, and noise points can be obtained through this clustering, that is, the corresponding clustering results.

[0163] Step S40: Obtain the target cluster according to the clustering result, and obtain the number of radar signal sources according to the target cluster.

[0164] After obtaining the corresponding clustering result, the corresponding number of radar signal sources is obtained through the clustering result.

[0165] Further, the obtaining the target cluster according to the clustering result and obtaining the number of radar signal sources according to the target cluster specifically include:

[0166] Calculate the average value of the current cluster clustering and the new cluster clustering in the clustering result, and select the cluster with the larger average value as the target cluster;

[0167] Count the number of elements in the target cluster, and output this number as the number of radar signal sources.

[0168] Specifically, calculate the average value of each cluster in the previous cluster clustering and the new cluster clustering, compare the sizes of the average radius values, the cluster with the largest average value is the cluster corresponding to the signal source, count the number of points in this cluster, which is the number of signal sources, and finally output the number of signal sources.

[0169] Further, the present invention correspondingly conducts a simulation experiment to compare the estimation performance of the technical method of the present invention with AIC, MDL, and the Gerschgorin circle radius method.

[0170] As Figure 4 shown, design a radar antenna array model with the number of array elements and the antenna spacing of . The radar echo signal is incident on the above array, and the incident angles are 0 degrees, 10 degrees, and 30 degrees respectively. The noise background is Gaussian white noise, the signal-to-noise ratio ranges from -20 dB to 20 dB with a step of 2 dB. Calculate the signal power and adjust the noise power according to the changing (signal-to-noise ratio) value to ensure that the noise power matches the signal-to-noise ratio of the signal . Set the number of snapshots to 500, and the OPTIC reachable distance threshold to 0.26 (heuristically selected based on actual experience). For each value, repeat 1000 Monte Carlo trials, and record the estimated number of signal sources in each experiment. If the estimated value is the same as the true number of signal sources, it is counted as a successful estimation. The final accuracy acc is calculated by dividing the number of successful times by the total number of experiments, specifically as the following formula (25):

[0171] ; (25)

[0172] where To estimate the number of successes, is the total number of experiments, and finally the accuracy curve is calculated and plotted. The abscissa represents value, and the ordinate represents the accuracy. In Figure 4 , in the background of white noise signal, when the signal-to-noise ratio is greater than -5dB, the signal source estimation accuracy of MDL always remains at 1 and shows high stability; the technical method of the present invention has significantly better accuracy in estimating the number of signal sources in the signal-to-noise ratio range of -5dB to 5dB than the Gerschgorin circle radius method, and when the signal-to-noise ratio is greater than 5dB, its accuracy is not only higher than that of AIC, but also the stability is significantly better than that of AIC.

[0173] Such as Figure 5 shown, also design a radar antenna array model with the number of array elements and the antenna spacing of . The radar echo signal is incident on the above array, and the incident angles are 0 degrees, 10 degrees, and 30 degrees respectively. The noise background is colored noise. Repeat 1000 Monte Carlo experiments with a step size of 2dB in the signal-to-noise ratio range of -20dB to 20dB, the number of snapshots is 500, and the OPTIC reachable distance threshold is 0.26 (heuristically selected based on actual experience). In order to simulate the actual noise environment, in this experiment, white noise is passed through a 4th-order Butterworth low-pass filter with a cut-off frequency of 0.6, and the generated noise after filtering becomes colored noise. In the experiment, the performance under different will also be compared according to the method of Experiment 1, and the accuracy curve is calculated and plotted. As Figure 5 shown, in the background of colored noise, the performance of MDL and AIC drops significantly, and the accuracy is close to 0, indicating that these two methods are almost completely ineffective under colored noise conditions; in contrast, the Gerschgorin circle radius method performs better than MDL and AIC in the face of colored noise. However, in the signal-to-noise ratio range of 0dB to 12dB, the technical method of the present invention has significantly higher accuracy in estimating the number of signal sources than the Gerschgorin circle radius method, and when the signal-to-noise ratio is greater than 5dB, its accuracy is close to 1 and maintains extremely high stability.

[0174] The present invention generates a representation of antenna array received data based on target radar antenna array information, obtains a target covariance matrix from the representation of antenna array received data, performs a unitary transformation on the target covariance matrix to obtain a target transformation matrix, and generates a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem; obtains a noise background, and when the noise background is a Gaussian colored noise background, compresses and normalizes the radius vector to obtain a target radius matrix; clusters the target radius matrix according to a target clustering algorithm to obtain a clustering result; obtains a target cluster according to the clustering result, and obtains the number of radar signal sources according to the target cluster. The present invention compresses and normalizes the Gerschgorin circle radius obtained after performing a unitary transformation on the signal covariance matrix in a Gaussian colored noise background, thereby reducing redundant information and improving the discrimination of noise signals. Furthermore, a target clustering algorithm is used to cluster the target radius matrix, and the corresponding radii are divided into a noise cluster and a signal cluster through clustering, thereby effectively distinguishing noise and signals, and thus improving the accuracy of estimating the number of signal sources under low signal-to-noise ratio and colored noise.

[0175] Further, as Figure 6 shown, based on the above radar signal source number detection method based on density clustering, the present invention also correspondingly provides a radar signal source number detection system based on density clustering, wherein the radar signal source number detection system based on density clustering includes:

[0176] A radius vector generation module 61, configured to generate a representation of antenna array received data based on target radar antenna array information, obtain a target covariance matrix from the representation of antenna array received data, perform a unitary transformation on the target covariance matrix to obtain a target transformation matrix, and generate a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem;

[0177] A target radius matrix generation module 62, configured to obtain a noise background, and when the noise background is a Gaussian colored noise background, compress and normalize the radius vector to obtain a target radius matrix;

[0178] A clustering result generation module 63, configured to cluster the target radius matrix according to a target clustering algorithm to obtain a clustering result;

[0179] A result output module 64, configured to obtain a target cluster according to the clustering result, and obtain the number of radar signal sources according to the target cluster.

[0180] Further, as Figure 7 shown, based on the above radar signal source number detection method and system based on density clustering, the present invention also correspondingly provides a terminal, and the terminal includes a processor 10, a memory 20, and a display 30. Figure 7Only some components of the terminal are shown, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively.

[0181] The memory 20 may be an internal storage unit of the terminal in some embodiments, such as a hard disk or memory of the terminal. The memory 20 may also be an external storage device of the terminal in some other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the terminal. Further, the memory 20 may also include both the internal storage unit and the external storage device of the terminal. The memory 20 is used to store application software installed on the terminal and various types of data, such as program codes for installing the terminal. The memory 20 may also be used to temporarily store data that has been output or will be output. In one embodiment, a density-based clustering radar source number detection program 40 is stored on the memory 20, and the density-based clustering radar source number detection program 40 can be executed by the processor 10, thereby implementing the density-based clustering radar source number detection method in the present invention.

[0182] The processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chips in some embodiments, and is used to run the program codes stored in the memory 20 or process data, such as executing the density-based clustering radar source number detection method, etc.

[0183] The display 30 may be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light-Emitting Diode) toucher, etc. in some embodiments. The display 30 is used to display information on the terminal and to display a visual user interface.

[0184] In one embodiment, when the processor 10 executes the density-based clustering radar source number detection program 40 in the memory 20, the steps of the above density-based clustering radar source number detection method are implemented.

[0185] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a density-based clustering radar source number detection program, and when the density-based clustering radar source number detection program is executed by a processor, the steps of the density-based clustering radar source number detection method as described above are implemented.

[0186] It should be noted that, in this text, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or terminal. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or terminal including that element.

[0187] Of course, those of ordinary skill in the art can understand that all or part of the processes of implementing the above-described embodiment methods can be completed by instructing relevant hardware (such as a processor, a controller, etc.) through a computer program. The program can be stored in a computer-readable storage medium readable by a computer. When the program is executed, it can include the processes of the above-described method embodiments. The computer-readable storage medium can be a memory, a magnetic disk, an optical disk, etc.

[0188] It should be understood that the application of the present invention is not limited to the above examples. For those of ordinary skill in the art, improvements or transformations can be made according to the above description. All such improvements and transformations should fall within the protection scope of the appended claims of the present invention.

Claims

1. A method for detecting the number of radar signal sources based on density clustering, characterized in that The method for detecting the number of radar signal sources based on density clustering includes: Generating a representation of the received data of the antenna array according to the target radar antenna array information, obtaining a target covariance matrix based on the representation of the received data of the antenna array, performing a unitary transformation on the target covariance matrix to obtain a target transformation matrix, and generating a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem; Obtaining a noise background. When the noise background is a Gaussian colored noise background, compressing and normalizing the radius vector to obtain a target radius matrix; Performing clustering on the target radius matrix according to a target clustering algorithm to obtain a clustering result; Obtaining a target cluster according to the clustering result, and obtaining the number of radar signal sources according to the target cluster; The step of generating a representation of the received data of the antenna array according to the target radar antenna array information, obtaining a target covariance matrix based on the representation of the received data of the antenna array, performing a unitary transformation on the target covariance matrix to obtain a target transformation matrix, and generating a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem specifically includes: Constructing an array steering vector corresponding to each signal source according to the target radar antenna array information, generating a steering vector matrix based on all the array steering vectors, and generating the representation of the received data of the antenna array according to the steering vector matrix; Obtaining the target covariance matrix and the covariance matrix of the radar echo signal according to the representation of the received data of the antenna array; Performing a unitary transformation on the target covariance matrix according to the covariance matrix of the radar echo signal to obtain the target transformation matrix; According to the Gerschgorin circle theorem, obtaining the Gerschgorin circle radius of each Gerschgorin circle from the target transformation matrix, and generating the radius vector according to all the Gerschgorin circle radii; The step of performing a unitary transformation on the target covariance matrix according to the covariance matrix of the radar echo signal to obtain the target transformation matrix specifically includes: Partitioning the target covariance matrix according to the covariance matrix of the radar echo signal, and obtaining a principal submatrix according to the partitioning result; Performing an eigenvalue decomposition on the principal submatrix to obtain the eigenspace vectors of the principal submatrix; Constructing a unitary transformation matrix according to the eigenspace vectors, and performing a unitary transformation operation on the eigenspace vectors according to the unitary transformation matrix to obtain the target transformation matrix; The step of obtaining a noise background, and when the noise background is a Gaussian colored noise background, compressing and normalizing the radius vector to obtain a target radius matrix specifically includes: Obtaining a noise background and judging the type of the noise background; When the noise background is a Gaussian colored noise background, compressing each Gerschgorin circle radius in the radius vector according to the unitary transformation matrix and the center of each Gerschgorin circle to obtain a plurality of compressed radii; Performing a normalization process on each compressed radius to obtain the target radius of each Gerschgorin circle; Generating the target radius matrix according to all the target radii.

2. The method for detecting the number of radar signal sources based on density clustering according to claim 1, wherein After the step of obtaining a noise background and judging the type of the noise background, it further includes: When the noise background is a Gaussian white noise background, performing a logarithmic transformation on the radius vector to obtain a logarithmic transformation matrix; Cluster the logarithmic transformation matrix according to the target clustering algorithm to obtain the white noise clustering result; Obtain the white noise target cluster according to the white noise clustering result, and obtain the corresponding radar signal source number according to the white noise target cluster.

3. The method for detecting the number of radar signal sources based on density clustering according to claim 1, wherein The clustering of the target radius matrix according to the target clustering algorithm to obtain the clustering result specifically includes: Obtain the preset distance threshold and the minimum number of neighborhoods, and use the target radius matrix, the distance threshold, and the minimum number of neighborhoods as inputs for clustering; During the clustering process, traverse each element in the target radius matrix, select the core objects according to the minimum number of neighborhoods, add them to the core object set and generate an ordered arrangement, calculate the core distance and reachable distance of each core object, and according to the core distance and reachable distance of each core object in the ordered arrangement, divide all the core objects into one of the current cluster clustering, new cluster clustering, and noise points respectively; Generate the clustering result according to the current cluster clustering, the new cluster clustering, and the noise points.

4. The method for detecting the number of radar signal sources based on density clustering according to claim 3, wherein The obtaining of the target cluster according to the clustering result and the obtaining of the radar signal source number according to the target cluster specifically include: Calculate the average value of the current cluster clustering and the new cluster clustering in the clustering result, and select the cluster with the larger average value as the target cluster; Count the number of elements in the target cluster and output this number as the radar signal source number.

5. A radar signal source number detection system based on density clustering, characterized in that The radar signal source number detection system based on density clustering includes: A radius vector generation module, configured to generate an antenna array received data representation according to the target radar antenna array information, obtain a target covariance matrix according to the antenna array received data representation, perform a unitary transformation on the target covariance matrix to obtain a target transformation matrix, and generate a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem; A target radius matrix generation module, configured to obtain the noise background, and when the noise background is a Gaussian colored noise background, compress and standardize the radius vector to obtain the target radius matrix; A clustering result generation module, configured to cluster the target radius matrix according to the target clustering algorithm to obtain the clustering result; A result output module, configured to obtain the target cluster according to the clustering result and obtain the radar signal source number according to the target cluster; The generating of the antenna array received data representation according to the target radar antenna array information, obtaining the target covariance matrix according to the antenna array received data representation, performing a unitary transformation on the target covariance matrix to obtain the target transformation matrix, and generating a radius vector corresponding to the target transformation matrix according to the Gerschgorin circle theorem specifically includes: Construct an array steering vector corresponding to each signal source according to the target radar antenna array information, generate a steering vector matrix according to all the array steering vectors, and generate the antenna array received data representation according to the steering vector matrix; Obtain the target covariance matrix and the radar echo signal covariance matrix according to the antenna array received data representation; Perform a unitary transformation on the target covariance matrix according to the radar echo signal covariance matrix to obtain the target transformation matrix; According to the Gerschgorin circle theorem, obtain the Gerschgorin circle radius of each Gerschgorin circle from the target transformation matrix, and generate the radius vector according to all the Gerschgorin circle radii; The unitary transformation of the target covariance matrix according to the radar echo signal covariance matrix to obtain the target transformation matrix specifically includes: Block the target covariance matrix according to the radar echo signal covariance matrix, and obtain the principal submatrix according to the result of the blocking; Perform eigenvalue decomposition on the principal submatrix to obtain the eigenspace vector of the principal submatrix; Construct a unitary transformation matrix according to the eigenspace vector, and perform a unitary transformation operation on the eigenspace vector according to the unitary transformation matrix to obtain the target transformation matrix; When obtaining the noise background, when the noise background is a Gaussian colored noise background, compress and standardize the radius vector to obtain the target radius matrix, specifically including: Obtain the noise background and determine the type of the noise background; When the noise background is a Gaussian colored noise background, compress each Gerschgorin circle radius in the radius vector according to the unitary transformation matrix and the center of each Gerschgorin circle to obtain a plurality of compressed radii; Perform a standardization process on each of the compressed radii to obtain the target radius of each Gerschgorin circle; Generate the target radius matrix according to all the target radii.

6. A terminal, characterized in that, The terminal includes: a memory, a processor, and a density-based clustering radar source number detection program stored on the memory and executable on the processor. When the density-based clustering radar source number detection program is executed by the processor, the steps of the density-based clustering radar source number detection method according to any one of claims 1-4 are implemented.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a density-based clustering radar source number detection program. When the density-based clustering radar source number detection program is executed by a processor, the steps of the density-based clustering radar source number detection method according to any one of claims 1-4 are implemented.

Images