A Multi-Radar Cooperative Target Recognition Method Based on Adaptive Manifold Discriminant Regression

Abstract

This invention provides a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression, relating to the field of radar signal processing technology. The method includes: determining a first data matrix based on first HRRP data from at least two viewpoints corresponding to the object to be identified; performing feature projection on the first data matrix according to a target projection matrix to obtain a first feature vector; the target projection matrix is ​​obtained by iteratively updating the initial projection matrix using second HRRP data from at least two viewpoints, training labels corresponding to the second HRRP data from each viewpoint, and a first objective function for multi-view recognition; the objective parameters of the first objective function include viewpoint weights, the projection matrix, and a nearest neighbor weight matrix for each viewpoint; and determining the object category corresponding to the first HRRP data based on the first feature vector. This invention improves the accuracy of multi-view HRRP target recognition.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing technology, and in particular to a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression. Background Technology

[0002] The fluctuations in high-resolution radar echoes reflect the distribution characteristics of targets along the radar line of sight, and are known as High Resolution Range Profiles (HRRP). Target recognition methods based on HRRP data occupy an important position in the field of radar target recognition. Existing traditional single-view HRRP recognition methods can be mainly divided into three types: some methods extract aspect-invariant features, such as scattering center features, transform domain features, feature combination and optimization, and sparse representation features; some methods implement angle-domain framing to avoid excessive amplitude fluctuations within frames; and some methods utilize deep networks, such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and hybrid networks, to extract angle-invariant features. In multi-radar cooperative detection scenarios, the key issue of multi-view HRRP cooperative recognition is modeling the data correlations and differences between multiple views.

[0003] Most current target recognition techniques for multi-view HRRP joint processing are extensions of single-view HRRP methods. Some works use decision-level fusion methods or sparse representations to integrate multi-view HRRP data. For example, target recognition of multi-view HRRP data is achieved by simply integrating data from each viewpoint and performing a weighted average, but the recognition accuracy is low. Summary of the Invention

[0004] This invention provides a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression to address the shortcomings of low recognition accuracy in existing technologies and improve recognition accuracy.

[0005] In a first aspect, the present invention provides a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression, the method comprising the following steps:

[0006] Acquire the first HRRP data of at least two viewpoints corresponding to the object to be identified;

[0007] Based on the first HRRP data from each of the aforementioned perspectives, determine the first data matrix;

[0008] Based on the target projection matrix, feature projection is performed on the first data matrix to obtain the first feature vector corresponding to the first data matrix; the target projection matrix is ​​obtained by iteratively updating the initial projection matrix using second HRRP data from at least two perspectives, training labels corresponding to the second HRRP data from each perspective, and a first objective function for multi-view recognition; the training labels are associated with the second feature vectors corresponding to the second HRRP data from each perspective, and the second feature vectors corresponding to the second HRRP data from each perspective are used to determine the object category corresponding to the second HRRP data; the objective parameters of the first objective function include the viewpoint weights corresponding to each perspective, the projection matrix, and the nearest neighbor weight matrix, and the nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in the preset category under each perspective.

[0009] Based on the first feature vector, the object category corresponding to the first HRRP data is determined.

[0010] According to the present invention, a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression is provided, wherein the target projection matrix is ​​determined through the following steps, including:

[0011] Construct the first objective function for the multi-view recognition;

[0012] Initialize the target parameters in the first objective function;

[0013] If the number of algorithm iterations has not reached the preset number of iterations, the target parameters are iteratively updated using the second HRRP data from at least two perspectives and the training labels corresponding to the second HRRP data from each perspective, so as to obtain the updated target parameters.

[0014] Based on the updated target parameters, determine the updated projection matrix;

[0015] When the algorithm iteration count reaches the preset iteration count, the updated projection matrix corresponding to the termination of the iteration is determined as the target projection matrix; the number of rows of the target projection matrix is ​​determined based on the number of feature dimensions contained in the second HRRP data, and the number of columns of the target projection matrix is ​​determined based on the preset number of object categories.

[0016] According to the present invention, a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression is provided, wherein the target parameters are iteratively updated using second HRRP data from at least two perspectives and training labels corresponding to the second HRRP data from each perspective to obtain updated target parameters, including:

[0017] Construct a second data matrix based on the second HRRP data from at least two perspectives;

[0018] For any of the target parameters, other parameters besides the target parameter are treated as fixed parameters. The first objective function is updated according to the second data matrix, and each updated target parameter is obtained based on the updated second objective function.

[0019] According to the present invention, a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression is provided. The target parameters further include an adjustment matrix, which is used as an adjustment term for adaptive weighted fusion. For any target parameter, other parameters besides the target parameter are treated as fixed parameters. The first objective function is updated according to the second data matrix, and based on the updated second objective function, each updated target parameter is solved. According to the second data matrix, any parameter among the target parameters includes:

[0020] The first objective function is updated by using the nearest neighbor weight matrix, the adjustment matrix, and the view weights corresponding to each view as fixed parameters to obtain the second objective function.

[0021] Calculate the gradient of the second objective function, and determine the updated projection matrix as the projection matrix corresponding to when the gradient of the second objective function is 0;

[0022] The updated projection matrix, the adjustment matrix, and the view weights corresponding to each viewpoint are used as fixed parameters to update the first objective function, thereby obtaining the third objective function.

[0023] Calculate the gradient of the third objective function, and determine the nearest neighbor weight matrix corresponding to the gradient of the third objective function being 0 as the updated nearest neighbor weight matrix;

[0024] The updated projection matrix, the updated nearest neighbor weight matrix, and the view weights corresponding to each view are used as fixed parameters to update the first objective function, resulting in the fourth objective function.

[0025] Based on the fourth objective function, determine the updated adjustment matrix;

[0026] The updated projection matrix, the updated nearest neighbor weight matrix, and the updated adjustment matrix are used as fixed parameters to update the first objective function, resulting in the fifth objective function.

[0027] Based on the fifth objective function, the updated view weights corresponding to each view are determined;

[0028] The updated projection matrix, the updated nearest neighbor weight matrix, the updated adjustment matrix, and the updated view weights corresponding to each viewpoint are determined as the updated target parameters.

[0029] According to the present invention, a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression is provided, wherein determining the first data matrix based on the first HRRP data of each viewpoint includes:

[0030] Based on the first HRRP data from each of the aforementioned perspectives, determine the first HRRP data for each feature dimension;

[0031] The first data matrix is ​​determined based on the first HRRP data of each of the aforementioned feature dimensions.

[0032] According to the present invention, a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression is provided, wherein the first data matrix is ​​feature-projected according to the target projection matrix to obtain a first feature vector corresponding to the first data matrix.

[0033] Multiply the target projection matrix by the first data matrix to obtain the feature vector matrix;

[0034] Based on the eigenvector matrix, the first eigenvector is determined.

[0035] According to the present invention, a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression is provided, which determines the object category corresponding to the first HRRP data based on the first feature vector, including:

[0036] For any first feature vector, determine the Euclidean distance between the first feature vector and the second feature vector corresponding to the second HRRP data of each viewpoint;

[0037] Based on the Euclidean distance between the first feature vector and the second feature vector corresponding to the second HRRP data of each viewpoint, at least one neighbor that is closest to the first feature vector is determined.

[0038] Based on the preset object category corresponding to the nearest neighbor, determine the object category corresponding to the first feature vector;

[0039] Based on the object category corresponding to each of the first feature vectors, the object category corresponding to the first HRRP data is determined.

[0040] Secondly, the present invention also provides a multi-radar cooperative target recognition device based on adaptive manifold discriminant regression, the device comprising the following modules:

[0041] The acquisition module is used to acquire first HRRP data from at least two perspectives;

[0042] The identification module is used to determine the first data matrix based on the first HRRP data from each of the aforementioned viewpoints;

[0043] Based on the target projection matrix, feature projection is performed on the first data matrix to obtain the first feature vector corresponding to the first data matrix; the target projection matrix is ​​obtained by iteratively updating the initial projection matrix using second HRRP data from at least two perspectives, training labels corresponding to the second HRRP data from each perspective, and a first objective function for multi-view recognition; the training labels are used to characterize the second feature vector corresponding to the second HRRP data from each perspective, and the second feature vector is used to determine the object category corresponding to the second HRRP data; the objective parameters of the first objective function include the viewpoint weights, projection matrix, and nearest neighbor weight matrix corresponding to each perspective, and the nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in the preset category under each perspective.

[0044] Based on the first feature vector, the object category corresponding to the first HRRP data is determined.

[0045] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression as described above.

[0046] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression as described above.

[0047] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression as described above.

[0048] The multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by this invention acquires first HRRP data from at least two perspectives, determines a first data matrix based on the first HRRP data from each perspective, and then projects features onto the first data matrix according to the target projection matrix to obtain a first feature vector corresponding to the first data matrix. The target projection matrix is ​​obtained by iteratively updating the initial projection matrix using second HRRP data from at least two perspectives, training labels corresponding to the second HRRP data from each perspective, and a first objective function for multi-view recognition. The training labels are associated with second feature vectors corresponding to the second HRRP data from each perspective, and the second feature vectors corresponding to the second HRRP data from each perspective are used to determine the object category corresponding to the second HRRP data. Furthermore, the object category corresponding to the first HRRP data is determined based on the first feature vector.

[0049] In this invention, the target parameters of the first objective function include an adjustment matrix, viewpoint weights corresponding to each viewpoint, a projection matrix, and a nearest neighbor weight matrix. The nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in a preset category under each viewpoint. First, the initial projection matrix is ​​iteratively updated using the second HRRP data from at least two viewpoints, the training labels corresponding to the second HRRP data from each viewpoint, and the first objective function for multi-view recognition to obtain the target projection matrix. The training labels are used to characterize the object category corresponding to the second HRRP data from each viewpoint. This invention fully considers the relationship between HRRP samples under different viewpoints and constructs the first objective function by modeling the relationship between the second HRRP data from different viewpoints through adaptive weights. Then, feature projection is performed on the first HRRP data from multiple viewpoints based on the target projection matrix to obtain a low-dimensional representation of the data, ultimately improving the accuracy of multi-view HRRP target recognition. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0051] Figure 1 This is one of the flowcharts of the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by the present invention.

[0052] Figure 2 This is the second flowchart of the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by the present invention.

[0053] Figure 3This is the third flowchart of the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by the present invention.

[0054] Figure 4 This is a schematic diagram of the structure of the multi-radar cooperative target recognition device based on adaptive manifold discriminant regression provided by the present invention.

[0055] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0057] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, not limited in number; for example, a first node can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0058] The following is combined with Figures 1-5 This invention describes a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression.

[0059] Figure 1 This is one of the flowcharts illustrating the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by this invention, such as... Figure 1 As shown, the method includes the following:

[0060] Step 101: Obtain the first HRRP data of at least two viewpoints corresponding to the object to be identified;

[0061] Specifically, it should first be noted that the execution subject of this embodiment is an electronic device, which is used to realize the identification of multi-view targets based on high resolution range profile (HRRP) data from multiple perspectives, thereby improving the accuracy of target identification.

[0062] First, the first HRRP data corresponding to the object to be identified is acquired from multiple perspectives. Here, "multiple perspectives" refers to the perspectives corresponding to collaborative detection by multiple radars. HRRP is an abbreviation for High Resolution Range Profile, a radar technique used to obtain the projection vector sum of the target's scattered point echoes along the radar's line-of-sight. HRRP contains much important structural information such as target size and scattering point distribution, which is crucial for radar target identification and classification. In the field of radar target identification, HRRP-based target identification is an important research direction because it can provide geometric structural information of the target and is a promising technology in radar automatic target recognition (RATR) applications. In summary, HRRP-based target identification is a key technology for target identification and classification in the radar field. It provides structural information about the target by analyzing the target's scattering point distribution and is an important component of radar automatic target recognition technology.

[0063] The specific way to obtain the first HRRP data from multiple perspectives is to use multiple radars with multiple different perspectives to detect the target and obtain the HRRP.

[0064] Step 102: Determine the first data matrix based on the first HRRP data from each perspective;

[0065] Specifically, after obtaining the first HRRP data from each perspective, it is necessary to further perform target identification based on the first HRRP data from multiple perspectives.

[0066] First, a first data matrix is ​​determined based on the first HRRP data from each perspective. The number of rows in the first data matrix is ​​the feature dimension d after concatenating all perspectives, and the number of columns is the total number of samples n.

[0067]

[0068] Where, d k Let d represent the feature dimension of the k-th view, d represent the feature dimension after connecting all views, and v represent the v views.

[0069] The first data matrix can be obtained as follows:

[0070]

[0071] in, This represents the first HRRP data from the first perspective. This represents the first HRRP data from a second perspective. This represents the first HRRP data from the v-th viewpoint.

[0072] Step 103: Based on the target projection matrix, perform feature projection on the first data matrix to obtain the first feature vector corresponding to the first data matrix; the target projection matrix is ​​obtained by iteratively updating the initial projection matrix using the second HRRP data from at least two perspectives, the training labels corresponding to the second HRRP data from each perspective, and the first objective function for multi-view recognition; the training labels are associated with the second feature vectors corresponding to the second HRRP data from each perspective, and the second feature vectors corresponding to the second HRRP data from each perspective are used to determine the object category corresponding to the second HRRP data; the objective parameters of the first objective function include the perspective weights corresponding to each perspective, the projection matrix, and the nearest neighbor weight matrix, and the nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in the preset category under each perspective;

[0073] Specifically, the training labels are used to represent the object categories corresponding to the second HRRP data from each viewpoint. Multiple object categories can be pre-defined, and different labels can correspond to different second feature vectors. The training labels are associated with the second feature vectors corresponding to the second HRRP data from each viewpoint, and the second feature vectors corresponding to the second HRRP data from each viewpoint are used to determine the object category corresponding to the second HRRP data. The pre-constructed first objective function is, for example:

[0074]

[0075]

[0076] Where λ1 and λ2 are the trade-off parameters, Let S denote the weighted projection matrix, and S denote the nearest neighbor weight matrix. Indicates the first From a perspective The weights of the nearest neighbors between the m-th and n-th samples in a class, where α represents the viewpoint weights for each viewpoint, and M = [M1, M2, ..., Mn]. n ]∈R n×c This represents the adjustment matrix, such as a nonnegative adjustment vector matrix, where k represents the k-th viewpoint, and X... k This represents the data sample matrix of HRRP from the k-th viewpoint. k Let Y represent the weighted projection transformation matrix for the k-th viewpoint, Y represent the sample label matrix, and ⊙ represent the Hadamard product operation.

[0077] Manifold learning regularization in the objective function can reveal the low-dimensional geometric structure of the data, capturing its intrinsic nature and exploring the data distribution structure of multi-view HRRP. We use adaptive weights to measure the similarity of different nearest neighbor pairs. Furthermore, HRRP from different perspectives can be considered to have different data distributions and underlying patterns. Therefore, we introduce manifold regularization terms modeled separately for different perspectives to explore the geometric information of the data. Through manifold learning, features can be preserved and the projection of the geometric structure can be extracted, but this projection contains limited class information. To better serve the final target recognition task, a multi-view adaptive weighted fusion framework based on improved least squares regression is proposed. After multi-view fusion, the resulting data features often have redundant features. These redundant features are useless or even harmful for model training. To address this issue, our method further imposes sparsity constraints on the projection. By... Norm regularization is applied to the projection matrix, where we identify key feature groups and set the weights of other groups to zero.

[0078] The target projection matrix is ​​obtained by iteratively updating the initial projection matrix using second HRRP data from at least two perspectives, the training labels corresponding to the second HRRP data from each perspective, and the first objective function for multi-view recognition. For example, the target parameters are updated within a preset number of iterations, and the projection matrix corresponding to the termination of the iteration is determined as the target projection matrix.

[0079] Furthermore, based on the target projection matrix, the first data matrix is ​​subjected to feature projection to obtain the first eigenvector corresponding to the first data matrix. For example, after obtaining the target projection matrix, the first eigenvector is obtained by multiplying the target projection matrix with the first data matrix.

[0080] Step 104: Determine the object category corresponding to the first HRRP data based on the first feature vector.

[0081] Specifically, after obtaining the first feature vector corresponding to the first data matrix, the predicted label corresponding to the first feature vector can be determined based on the first feature vector, that is, the predicted object category corresponding to the target of the first HRRP data, so as to achieve high-precision target recognition.

[0082] For example, the KNN (K-Nearest Neighbors) algorithm is used on the first feature vector to finally obtain the predicted label (predicted object category) of the HRRP sample.

[0083] The method provided in this embodiment acquires first HRRP data from at least two perspectives, determines a first data matrix based on the first HRRP data from each perspective, and then performs feature projection on the first data matrix according to the target projection matrix to obtain a first feature vector corresponding to the first data matrix. The target projection matrix is ​​obtained by iteratively updating the initial projection matrix using second HRRP data from at least two perspectives, training labels corresponding to the second HRRP data from each perspective, and a first objective function for multi-view recognition. The training labels are associated with the second feature vectors corresponding to the second HRRP data from each perspective, and the second feature vectors corresponding to the second HRRP data from each perspective are used to determine the object category corresponding to the second HRRP data. Furthermore, the object category corresponding to the first HRRP data is determined based on the first feature vector.

[0084] In this invention, the target parameters of the first objective function include an adjustment matrix, viewpoint weights corresponding to each viewpoint, a projection matrix, and a nearest neighbor weight matrix. The nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in a preset category under each viewpoint. First, the initial projection matrix is ​​iteratively updated using the second HRRP data from at least two viewpoints, the training labels corresponding to the second HRRP data from each viewpoint, and the first objective function for multi-view recognition to obtain the target projection matrix. The training labels are used to characterize the object category corresponding to the second HRRP data from each viewpoint. This invention fully considers the relationship between HRRP samples under different viewpoints and constructs the first objective function by modeling the relationship between the second HRRP data from different viewpoints through adaptive weights. Then, feature projection is performed on the first HRRP data from multiple viewpoints based on the target projection matrix to obtain a low-dimensional representation of the data, ultimately improving the accuracy of multi-view HRRP target recognition.

[0085] According to the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by the present invention, the target projection matrix is ​​determined by the following steps, including:

[0086] Construct the first objective function for multi-view recognition;

[0087] Initialize the objective parameters in the first objective function;

[0088] If the number of algorithm iterations has not reached the preset number of iterations, the target parameters are iteratively updated using the second HRRP data from at least two perspectives and the training labels corresponding to the second HRRP data from each perspective, so as to obtain the updated target parameters.

[0089] Based on the updated target parameters, determine the updated projection matrix;

[0090] If the algorithm iteration count reaches the preset iteration count, the updated projection matrix corresponding to the termination of the iteration is determined as the target projection matrix; the number of rows of the target projection matrix is ​​determined based on the number of viewpoints contained in the second HRRP data, and the number of columns of the target projection matrix is ​​determined based on the preset number of object categories.

[0091] Specifically, in some embodiments, the target projection matrix is ​​determined through the following steps:

[0092] First, we construct the first objective function for multi-view recognition. An example of the first objective function is shown below:

[0093]

[0094]

[0095] Where λ1 and λ2 are the trade-off parameters, Let S denote the weighted projection matrix, and S denote the nearest neighbor weight matrix. Indicates the first From a perspective The weights of the nearest neighbors between the m-th and n-th samples in a class, where α represents the viewpoint weights for each viewpoint, and M = [M1, M2, ..., Mn]. n ]∈R n×c This represents the adjustment matrix, such as a nonnegative adjustment vector matrix, where k represents the k-th viewpoint, and X... k This represents the data sample matrix of HRRP from the k-th viewpoint. k Let Y represent the weighted projection transformation matrix for the k-th viewpoint, and let Y represent the sample label matrix.

[0096] The first objective function comprises a manifold regularization cost function based on adaptive nearest neighbor weights and a multi-view adaptive weighted fusion cost function. The manifold regularization cost function evaluates the similarity of different nearest neighbor pairs by introducing adaptive weights, and models different perspectives separately to learn the distribution patterns of HRRP from different angles. The multi-view adaptive weighted fusion cost function, based on an improved least squares regression method, further fuses the obtained manifold projections and introduces supervision information to better serve the recognition task. By modeling the relationships between different perspectives through adaptive weights, it more fully models the consistency, complementarity, and differences among different perspectives, effectively improving recognition performance.

[0097] Furthermore, the multi-view high-resolution range profile (HRRP) is input to initialize the target parameters in the first objective function. The parameters to be initialized are: the adjustment matrix M, and the view weights α. k and projection matrix .

[0098] First, the adjustment matrix is ​​initialized to M=0, and the projection matrix is ​​initialized to... 0=0, initialize the view weights to 0. Furthermore, for the manifold regularization term, we first initialize the nearest neighbor weight matrix using the Laplacian matrix of the data samples, constructing it as a k-nearest neighbor graph with binary values, i.e. .

[0099] Furthermore, if the algorithm iteration count has not reached the preset iteration count, the target parameters are iteratively updated using second HRRP data from at least two perspectives and the corresponding training labels for each perspective's second HRRP data, resulting in updated target parameters. The preset iteration count can be set based on the data sample size, such as 100, 500, or 1000 iterations, until the function converges. Based on the updated target parameters, the updated projection matrix is ​​determined.

[0100] Furthermore, when the algorithm iteration count reaches a preset number of iterations, the updated projection matrix corresponding to the termination of the iteration is determined as the target projection matrix. The number of rows in the target projection matrix is ​​determined based on the number of feature dimensions contained in the second HRRP data, and the number of columns in the target projection matrix is ​​determined based on the preset number of object categories.

[0101] For example, Figure 2 This is the second flowchart of the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by this invention, illustrating the process of obtaining the target projection matrix through the algorithm, as follows: Figure 2 As shown, the method includes:

[0102] First, input the multi-view HRRP;

[0103] Initialize the adjustment matrix, weight matrix, and projection matrix;

[0104] Determine if the number of iterations has exceeded the limit;

[0105] If the number of iterations has not exceeded the limit: update the parameters; check if the difference meets the standard; if it does, stop the iteration and output the projection matrix; if it does not meet the standard, increment the number of iterations by 1 and check again if the number of iterations has exceeded the limit.

[0106] If the number of iterations exceeds the limit, the iteration will stop.

[0107] The method provided in this embodiment first constructs a first objective function for multi-view recognition. The first objective function includes a manifold regularization cost function based on adaptive nearest neighbor weights and a multi-view adaptive weighted fusion cost function. The objective parameters in the first objective function are initialized. The algorithm is iteratively updated using second HRRP data from at least two views and the training labels corresponding to the second HRRP data from each view to obtain the updated objective parameters. When the algorithm iteration count reaches a preset number of iterations, the updated projection matrix corresponding to the termination of iteration is determined as the target projection matrix. This invention learns a target projection matrix through adaptive manifold learning, adaptive weighted fusion, and L2,1 norm regularization. This target projection matrix can be used to obtain a low-dimensional representation of multi-view HRRP data, ultimately realizing multi-view HRRP target recognition and improving the accuracy of target recognition.

[0108] According to the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by the present invention, the target parameters are iteratively updated using second HRRP data from at least two perspectives and the training labels corresponding to the second HRRP data from each perspective, to obtain the updated target parameters, including:

[0109] Construct a second data matrix based on second HRRP data from at least two perspectives;

[0110] For any target parameter, other parameters are treated as fixed parameters. The first objective function is updated based on the second data matrix, and the updated target parameters are obtained based on the updated second objective function.

[0111] Specifically, in some embodiments, the process of obtaining the second HRRP data from at least two perspectives in the target projection matrix and the training labels corresponding to the second HRRP data from each perspective, and iteratively updating the target parameters to obtain the updated target parameters includes the following steps:

[0112] First, a second data matrix is ​​constructed based on second HRRP data from at least two perspectives;

[0113] Furthermore, for any target parameter, other parameters besides the target parameter are treated as fixed parameters. The first objective function is updated based on the second data matrix, and the updated target parameters are obtained based on the updated second objective function. The target parameters include the adjustment matrix M and the viewpoint weights corresponding to each viewpoint. Projection matrix And the nearest neighbor weight matrix S, where the nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in the preset category under each viewpoint.

[0114] For example, the neighbor relationship weight matrix S and the view weights corresponding to each view. Then, by adjusting matrix M as a fixed parameter to update the first objective function, a second objective function is obtained. Furthermore, by calculating the gradient of the second objective function and making the gradient zero, the projection matrix corresponding to the second objective function having a gradient of zero is obtained. This is determined as the target parameter for updating. The update process for other target parameters is similar.

[0115] The method provided in this embodiment first constructs a second data matrix based on second HRRP data from at least two perspectives. Then, for any target parameter, other parameters besides the target parameter are treated as fixed parameters. The first objective function is updated based on the second data matrix, and each updated target parameter is obtained based on the updated second objective function, realizing iterative updating of parameters. When the algorithm reaches the updated target parameter corresponding to the preset number of iterations, it is determined as the final parameter value, such as the target projection matrix. Finally, feature vector extraction is performed based on the target projection matrix, and target recognition is finally achieved.

[0116] According to the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by the present invention, the target parameters further include an adjustment matrix, which is used as an adjustment term for adaptive weighted fusion; for any target parameter, other parameters besides the target parameter are treated as fixed parameters, the first objective function is updated according to the second data matrix, and each updated target parameter is obtained based on the updated second objective function; according to the second data matrix, any parameter in the target parameters, including:

[0117] The nearest neighbor weight matrix, adjustment matrix, and view weights corresponding to each viewpoint in the first objective function are used as fixed parameters to update the first objective function, thus obtaining the second objective function.

[0118] Calculate the gradient of the second objective function, and determine the projection matrix corresponding to the gradient of the second objective function being 0 as the updated projection matrix;

[0119] The updated projection matrix, adjustment matrix, and view weights corresponding to each view are used as fixed parameters to update the first objective function, resulting in the third objective function.

[0120] Calculate the gradient of the third objective function, and determine the nearest neighbor weight matrix corresponding to the gradient of the third objective function being 0 as the updated nearest neighbor weight matrix;

[0121] The updated projection matrix, the updated nearest neighbor weight matrix, and the view weights corresponding to each view are used as fixed parameters to update the first objective function, resulting in the fourth objective function.

[0122] Based on the fourth objective function, determine the updated adjustment matrix;

[0123] The updated projection matrix, the updated nearest neighbor weight matrix, and the updated adjustment matrix are used as fixed parameters to update the first objective function, resulting in the fifth objective function.

[0124] Based on the fifth objective function, determine the view weights corresponding to each updated view.

[0125] The updated projection matrix, the updated nearest neighbor weight matrix, the updated adjustment matrix, and the updated view weights for each view are determined as the updated target parameters.

[0126] Specifically, in some embodiments, for any target parameter, other parameters besides the target parameter are treated as fixed parameters. The first objective function is updated according to the second data matrix, and the updated target parameters are obtained based on the updated second objective function. The specific implementation process of any parameter in the target parameters according to the second data matrix includes the following steps:

[0127] First, the nearest neighbor weight matrix, adjustment matrix, and view weights corresponding to each view in the first objective function are used as fixed parameters to update the first objective function, thereby obtaining the second objective function. The gradient of the second objective function is calculated, and the projection matrix corresponding to the gradient of the second objective function being 0 is determined as the updated projection matrix.

[0128] The second objective function is, for example:

[0129]

[0130] in, Given a vector, by calculating the gradient of the second objective function and making its gradient zero, we can obtain the value that minimizes the second objective function. Its expression is as follows:

[0131]

[0132] in, It is a diagonal matrix, where each diagonal element is , Where T is shown below:

[0133]

[0134] Furthermore, the updated projection matrix, adjustment matrix, and viewpoint weights corresponding to each viewpoint are used as fixed parameters to update the first objective function, resulting in the third objective function. The gradient of the third objective function is calculated, and the nearest neighbor weight matrix corresponding to the gradient of the third objective function being 0 is determined as the updated nearest neighbor weight matrix. The third objective function is, for example, as follows:

[0135]

[0136]

[0137] The solution to the optimization problem can be obtained as follows:

[0138]

[0139] Furthermore, the updated projection matrix M and the updated nearest neighbor weight matrix are... Viewpoint weights corresponding to each viewpoint As a fixed parameter, the first objective function is updated to obtain the fourth objective function. Then, based on the fourth objective function, the adjustment matrix for updating is determined. For example, the fourth objective function is:

[0140]

[0141]

[0142] Let the residual be Based on Frobenius norm theory, the problem can be decoupled into n×c subproblems. For matrices... , and the elements in the i-th row and j-th column of M , , The problem can be rephrased as:

[0143]

[0144] st

[0145] based on Only by " From the elemental composition, we can obtain:

[0146]

[0147] st

[0148] It is easy to see that the solution for the adjustment matrix is:

[0149]

[0150] Furthermore, the updated projection matrix, the updated nearest neighbor weight matrix, and the updated adjustment matrix are used as fixed parameters to update the first objective function, resulting in the fifth objective function; based on the fifth objective function, the updated view weights corresponding to each view are determined.

[0151] For example, fixed parameters The fifth objective function to be optimized is:

[0152]

[0153] st ,

[0154] Based on Cauchy's inequality theory, the solution can be obtained as:

[0155]

[0156] Furthermore, the updated projection matrix, the updated nearest neighbor weight matrix, the updated adjustment matrix, and the updated viewpoint weights for each viewpoint are determined as the updated target parameters, thus achieving the update of the target parameters. Finally, the target projection matrix is ​​determined based on the updated target parameters that satisfy the iteration termination condition. The first data matrix is ​​then projected using the target projection matrix, and the projected feature vectors are used for target recognition.

[0157] The method provided in this embodiment updates the first objective function using the nearest neighbor weight matrix, adjustment matrix, and viewpoint weights corresponding to each viewpoint as fixed parameters to obtain a second objective function; calculates the gradient of the second objective function, and determines the updated projection matrix as the projection matrix corresponding to the gradient of the second objective function being 0; updates the first objective function using the updated projection matrix, adjustment matrix, and viewpoint weights corresponding to each viewpoint as fixed parameters to obtain a third objective function; calculates the gradient of the third objective function, and determines the updated nearest neighbor weight matrix as the nearest neighbor weight matrix corresponding to the gradient of the third objective function being 0; and updates the first objective function using the updated projection matrix, updated nearest neighbor weight matrix, and viewpoint weights corresponding to each viewpoint as fixed parameters to obtain a second objective function. The function is updated to obtain the fourth objective function; based on the fourth objective function, the updated adjustment matrix is ​​determined; the updated projection matrix, the updated nearest neighbor weight matrix, and the updated adjustment matrix are used as fixed parameters to update the first objective function to obtain the fifth objective function; based on the fifth objective function, the updated view weights corresponding to each view are determined; the updated projection matrix, the updated nearest neighbor weight matrix, the updated adjustment matrix, and the updated view weights corresponding to each view are determined as the updated target parameters; finally, the target projection matrix is ​​determined based on the updated target parameters corresponding to the iteration termination; the first data matrix is ​​feature-projected based on the target projection matrix; and the projected feature vectors are used for target recognition, improving the accuracy of multi-view HRRP target recognition.

[0158] According to the present invention, a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression is provided, which determines a first data matrix based on first HRRP data from each viewpoint, including:

[0159] Based on the first HRRP data from each perspective, determine the first HRRP data for each feature dimension;

[0160] The first data matrix is ​​determined based on the first HRRP data of each feature dimension.

[0161] Specifically, in some embodiments, the specific implementation process of determining the first data matrix based on the first HRRP data from each viewpoint in step 102 includes the following steps:

[0162] First, the first HRRP data for each feature dimension is determined based on the first HRRP data from each viewpoint. Here, the feature dimension is the feature dimension d obtained by concatenating all viewpoints, where:

[0163]

[0164] Where, d kLet d represent the feature dimension of the k-th view, d represent the feature dimension after connecting all views, and v represent the v views.

[0165] Furthermore, based on the first HRRP data for each feature dimension, a first data matrix is ​​determined. The first HRRP data for each feature dimension are placed in a matrix, where the rows of the matrix represent the feature dimensions and the columns represent the number of samples.

[0166] The method provided in this embodiment determines a first data matrix based on the first HRRP data from each viewpoint, and then performs feature projection on the first data matrix through a target projection matrix. Finally, multi-view HRRP target recognition is achieved based on the projected feature vectors, thereby improving the accuracy of target recognition.

[0167] According to the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by the present invention, a first feature vector is obtained by performing feature projection on a first data matrix according to the target projection matrix.

[0168] Multiply the target projection matrix by the first data matrix to obtain the eigenvector matrix;

[0169] The first eigenvector is determined based on the eigenvector matrix.

[0170] Specifically, in some embodiments, step 103 can be implemented through the following steps:

[0171] The target projection matrix is, for example, as follows:

[0172]

[0173] The first data matrix is, for example:

[0174]

[0175] Multiplying the target projection matrix by the first data matrix yields an eigenvector matrix, which is an n×c matrix. Furthermore, based on this eigenvector matrix, the first eigenvector can be extracted.

[0176] The method provided in this embodiment performs feature projection on the first data matrix according to the target projection matrix to obtain the first feature vector corresponding to the first data matrix. Then, the object category corresponding to the first HRRP data can be predicted based on the first feature vector, thereby improving the accuracy of object recognition.

[0177] According to the present invention, a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression determines the object category corresponding to the first HRRP data based on a first feature vector, including:

[0178] For any first feature vector, determine the Euclidean distance between the first feature vector and the second feature vector corresponding to the second HRRP data from each viewpoint;

[0179] Based on the Euclidean distance between the first feature vector and the second feature vector corresponding to the second HRRP data of each viewpoint, at least one neighbor that is closest to the first feature vector is determined.

[0180] Determine the object category corresponding to the first feature vector based on the preset object category corresponding to at least one nearest neighbor.

[0181] Based on the object category corresponding to each first feature vector, the object category corresponding to the first HRRP data is determined.

[0182] Specifically, the specific implementation process of determining the object category corresponding to the first HRRP data based on the first feature vector in step 104 may include the following steps:

[0183] For example, the KNN algorithm can be used to predict the labels, or object categories, in HRRP data. The K-Nearest Neighbors algorithm is a basic and widely used classification and regression method. It makes predictions by measuring the distances between different feature values.

[0184] First, it should be noted that there can be multiple first feature vectors. For any feature vector, the Euclidean distance between the first feature vector and the second feature vector corresponding to the second HRRP data of each viewpoint is determined.

[0185] Furthermore, based on the Euclidean distance between the feature vector and the second feature vector corresponding to the second HRRP data of each viewpoint, at least one nearest neighbor to the feature vector can be determined. The nearest neighbor is at least one, but can be three or five, and generally an odd number of nearest neighbors are used.

[0186] Furthermore, based on the preset object category corresponding to at least one nearest neighbor, the object category corresponding to the first feature vector is determined. For example, the preset object category corresponding to the neighbor with the smallest distance value is determined as the object category corresponding to the first feature vector.

[0187] Furthermore, based on the object category corresponding to each first feature vector, the object category corresponding to the first HRRP data is determined. The KNN algorithm classifies the test data point into the category that is most frequent among these K nearest neighbors.

[0188] The method provided in this embodiment uses the K-nearest neighbor algorithm on the feature vector to obtain the predicted label of the multi-view HRRP data, that is, to perform target recognition and improve the accuracy of target recognition of multi-view data.

[0189] Figure 3 This is the third flowchart of the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by the present invention, as shown below. Figure 3 As shown, the method includes:

[0190] Enter HRRP;

[0191] Locally preserved manifold learning;

[0192] L2,1 norm (row sparsity) constraint;

[0193] Feature projection;

[0194] The nearest neighbor algorithm outputs the predicted label.

[0195] This invention addresses the current limitations of multi-view HRRP recognition methods in effectively exploring the relationships and unique distribution patterns of data samples from different perspectives. Simple multi-view fusion cannot effectively eliminate pose sensitivity and may lead to reduced accuracy. Therefore, this invention proposes a manifold regularization cost function based on adaptive nearest neighbor weights. By introducing adaptive weights to evaluate the similarity of different nearest neighbor pairs, and simultaneously modeling different perspectives separately, it achieves the learning of HRRP distribution patterns from different angles.

[0196] Furthermore, existing multi-view HRRP recognition methods focus on fusing HRRP signals from multiple perspectives, neglecting the relationships between HRRP samples from different perspectives and failing to accurately and effectively model these relationships. This invention proposes a multi-view adaptive weighted fusion cost function. Based on an improved least squares regression method, supervised information is introduced to further fuse the obtained manifold projections. Adaptive weighting models the relationships between different perspectives, thus more fully modeling the consistency, complementarity, and differences among them. Simultaneously, norm regularization constraints are applied to the projection matrix to remove redundant information, further improving recognition performance.

[0197] The following describes the multi-radar cooperative target recognition device based on adaptive manifold discriminant regression provided by the present invention. The multi-radar cooperative target recognition device based on adaptive manifold discriminant regression described below can be referred to in correspondence with the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression described above.

[0198] Figure 4 This is a schematic diagram of the structure of the multi-radar cooperative target recognition device based on adaptive manifold discriminant regression provided by the present invention, as shown below. Figure 4 As shown, the multi-radar cooperative target recognition device 400 based on adaptive manifold discriminant regression includes:

[0199] The acquisition module 410 is used to acquire first HRRP data of at least two viewpoints corresponding to the object to be identified;

[0200] The identification module 420 is used to determine the first data matrix based on the first HRRP data from each of the aforementioned viewpoints;

[0201] Based on the target projection matrix, feature projection is performed on the first data matrix to obtain the first feature vector corresponding to the first data matrix; the target projection matrix is ​​obtained by iteratively updating the initial projection matrix using second HRRP data from at least two perspectives, training labels corresponding to the second HRRP data from each perspective, and a first objective function for multi-view recognition; the training labels are used to characterize the second feature vector corresponding to the second HRRP data from each perspective, and the second feature vector is used to determine the object category corresponding to the second HRRP data; the objective parameters of the first objective function include the viewpoint weights, projection matrix, and nearest neighbor weight matrix corresponding to each perspective, and the nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in the preset category under each perspective.

[0202] Based on the first feature vector, the object category corresponding to the first HRRP data is determined.

[0203] The apparatus provided in this embodiment includes an acquisition module 410 and an identification module 420. The acquisition module 410 is used to acquire first HRRP data from at least two perspectives. The identification module 420 is used to determine a first data matrix based on the first HRRP data from each perspective. Then, based on a target projection matrix, feature projection is performed on the first data matrix to obtain a first feature vector corresponding to the first data matrix. The target projection matrix is ​​obtained by iteratively updating the initial projection matrix using second HRRP data from at least two perspectives, training labels corresponding to the second HRRP data from each perspective, and a first objective function for multi-view recognition. The training labels are used to characterize the object category corresponding to the second HRRP data from each perspective. Furthermore, the object category corresponding to the first HRRP data is determined based on the first feature vector.

[0204] In this invention, the target parameters of the first objective function include an adjustment matrix, viewpoint weights corresponding to each viewpoint, a projection matrix, and a nearest neighbor weight matrix. The nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in a preset category under each viewpoint. First, the initial projection matrix is ​​iteratively updated using the second HRRP data from at least two viewpoints, the training labels corresponding to the second HRRP data from each viewpoint, and the first objective function for multi-view recognition to obtain the target projection matrix. The training labels are used to characterize the object category corresponding to the second HRRP data from each viewpoint. This invention fully considers the relationship between HRRP samples under different viewpoints and constructs the first objective function by modeling the relationship between the second HRRP data from different viewpoints through adaptive weights. Then, feature projection is performed on the first HRRP data from multiple viewpoints based on the target projection matrix to obtain a low-dimensional representation of the data, ultimately improving the accuracy of multi-view HRRP target recognition.

[0205] According to the present invention, a multi-radar cooperative target recognition device 400 based on adaptive manifold discriminant regression is provided, the device further comprising a projection module; the projection module is specifically used for:

[0206] The target projection matrix is ​​determined through the following steps, including:

[0207] Construct the first objective function for the multi-view recognition;

[0208] Initialize the target parameters in the first objective function;

[0209] If the number of algorithm iterations has not reached the preset number of iterations, the target parameters are iteratively updated using the second HRRP data from at least two perspectives and the training labels corresponding to the second HRRP data from each perspective, so as to obtain the updated target parameters.

[0210] Based on the updated target parameters, determine the updated projection matrix;

[0211] When the algorithm iteration count reaches the preset iteration count, the updated projection matrix corresponding to the iteration termination is determined as the target projection matrix; the number of rows of the target projection matrix is ​​determined based on the number of viewpoints contained in the second HRRP data, and the number of columns of the target projection matrix is ​​determined based on the preset number of object categories.

[0212] According to the present invention, a multi-radar cooperative target recognition device 400 based on adaptive manifold discriminant regression is provided, wherein the projection module is further configured to:

[0213] Construct a second data matrix based on the second HRRP data from at least two perspectives;

[0214] For any of the target parameters, other parameters besides the target parameter are treated as fixed parameters. The first objective function is updated according to the second data matrix, and each updated target parameter is obtained based on the updated second objective function.

[0215] According to the present invention, a multi-radar cooperative target recognition device 400 based on adaptive manifold discriminant regression is provided, wherein the target parameters further include an adjustment matrix, the adjustment matrix being used as an adjustment term for adaptive weighted fusion; the projection module is further used for:

[0216] The first objective function is updated by using the nearest neighbor weight matrix, adjustment matrix, and view weights corresponding to each view as fixed parameters to obtain the second objective function.

[0217] Calculate the gradient of the second objective function, and determine the updated projection matrix as the projection matrix corresponding to when the gradient of the second objective function is 0;

[0218] The updated projection matrix, the adjustment matrix, and the view weights corresponding to each viewpoint are used as fixed parameters to update the first objective function, thereby obtaining the third objective function.

[0219] Calculate the gradient of the third objective function, and determine the nearest neighbor weight matrix corresponding to the gradient of the third objective function being 0 as the updated nearest neighbor weight matrix;

[0220] The updated projection matrix, the updated nearest neighbor weight matrix, and the view weights corresponding to each view are used as fixed parameters to update the first objective function, resulting in the fourth objective function.

[0221] Based on the fourth objective function, determine the updated adjustment matrix;

[0222] The updated projection matrix, the updated nearest neighbor weight matrix, and the updated adjustment matrix are used as fixed parameters to update the first objective function, resulting in the fifth objective function.

[0223] Based on the fifth objective function, the updated view weights corresponding to each view are determined;

[0224] The updated projection matrix, the updated nearest neighbor weight matrix, the updated adjustment matrix, and the updated view weights corresponding to each viewpoint are determined as the updated target parameters.

[0225] According to the present invention, a multi-radar cooperative target identification device 400 based on adaptive manifold discriminant regression is provided, wherein the identification module 420 is specifically used for:

[0226] Based on the first HRRP data from each of the aforementioned perspectives, determine the first HRRP data for each feature dimension;

[0227] The first data matrix is ​​determined based on the first HRRP data of each of the aforementioned feature dimensions.

[0228] According to the present invention, a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression is provided, wherein the first data matrix is ​​feature-projected according to the target projection matrix to obtain a first feature vector corresponding to the first data matrix.

[0229] Multiply the target projection matrix by the first data matrix to obtain the feature vector matrix;

[0230] Based on the eigenvector matrix, the first eigenvector is determined.

[0231] According to the multi-radar cooperative target identification device 400 based on adaptive manifold discriminant regression provided by the present invention, the identification module 420 is further configured to:

[0232] For any first feature vector, determine the Euclidean distance between the first feature vector and the second feature vector corresponding to the second HRRP data of each viewpoint;

[0233] Based on the Euclidean distance between the first feature vector and the second feature vector corresponding to the second HRRP data of each viewpoint, at least one neighbor that is closest to the first feature vector is determined.

[0234] Based on the preset object category corresponding to the nearest neighbor, determine the object category corresponding to the first feature vector;

[0235] Based on the object category corresponding to each of the first feature vectors, the object category corresponding to the first HRRP data is determined.

[0236] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include: a processor 510, a communications interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communications interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a multi-radar cooperative target recognition method based on adaptive manifold discriminant regression, which includes:

[0237] Acquire the first HRRP data of at least two viewpoints corresponding to the object to be identified;

[0238] Based on the first HRRP data from each of the aforementioned perspectives, determine the first data matrix;

[0239] Based on the target projection matrix, feature projection is performed on the first data matrix to obtain the first feature vector corresponding to the first data matrix; the target projection matrix is ​​obtained by iteratively updating the initial projection matrix using second HRRP data from at least two perspectives, training labels corresponding to the second HRRP data from each perspective, and a first objective function for multi-view recognition; the training labels are associated with the second feature vectors corresponding to the second HRRP data from each perspective, and the second feature vectors corresponding to the second HRRP data from each perspective are used to determine the object category corresponding to the second HRRP data; the objective parameters of the first objective function include the viewpoint weights corresponding to each perspective, the projection matrix, and the nearest neighbor weight matrix, and the nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in the preset category under each perspective.

[0240] Based on the first feature vector, the object category corresponding to the first HRRP data is determined.

[0241] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0242] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by the above methods, the method comprising:

[0243] Acquire the first HRRP data of at least two viewpoints corresponding to the object to be identified;

[0244] Based on the first HRRP data from each of the aforementioned perspectives, determine the first data matrix;

[0245] Based on the target projection matrix, feature projection is performed on the first data matrix to obtain the first feature vector corresponding to the first data matrix; the target projection matrix is ​​obtained by iteratively updating the initial projection matrix using second HRRP data from at least two perspectives, training labels corresponding to the second HRRP data from each perspective, and a first objective function for multi-view recognition; the training labels are associated with the second feature vectors corresponding to the second HRRP data from each perspective, and the second feature vectors corresponding to the second HRRP data from each perspective are used to determine the object category corresponding to the second HRRP data; the objective parameters of the first objective function include the viewpoint weights corresponding to each perspective, the projection matrix, and the nearest neighbor weight matrix, and the nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in the preset category under each perspective.

[0246] Based on the first feature vector, the object category corresponding to the first HRRP data is determined.

[0247] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression provided by the methods described above, the method comprising:

[0248] Acquire the first HRRP data of at least two viewpoints corresponding to the object to be identified;

[0249] Based on the first HRRP data from each of the aforementioned perspectives, determine the first data matrix;

[0250] Based on the target projection matrix, feature projection is performed on the first data matrix to obtain the first feature vector corresponding to the first data matrix; the target projection matrix is ​​obtained by iteratively updating the initial projection matrix using second HRRP data from at least two perspectives, training labels corresponding to the second HRRP data from each perspective, and a first objective function for multi-view recognition; the training labels are associated with the second feature vectors corresponding to the second HRRP data from each perspective, and the second feature vectors corresponding to the second HRRP data from each perspective are used to determine the object category corresponding to the second HRRP data; the objective parameters of the first objective function include the viewpoint weights corresponding to each perspective, the projection matrix, and the nearest neighbor weight matrix, and the nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in the preset category under each perspective.

[0251] Based on the first feature vector, the object category corresponding to the first HRRP data is determined.

[0252] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0253] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0254] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-radar cooperative target recognition method based on adaptive manifold discriminant regression, characterized in that, include: Acquire the first HRRP data of at least two viewpoints corresponding to the object to be identified; Based on the first HRRP data from each of the aforementioned perspectives, determine the first data matrix; Based on the target projection matrix, feature projection is performed on the first data matrix to obtain the first feature vector corresponding to the first data matrix; The target projection matrix is ​​obtained by iteratively updating an initial projection matrix using second HRRP data from at least two perspectives, training labels corresponding to the second HRRP data from each perspective, and a first objective function for multi-view recognition. The training labels are associated with second feature vectors corresponding to the second HRRP data from each perspective, and the second feature vectors corresponding to the second HRRP data from each perspective are used to determine the object category corresponding to the second HRRP data. The objective parameters of the first objective function include the viewpoint weights, projection matrix, and nearest neighbor weight matrix corresponding to each perspective. The nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in a preset category under each perspective. Based on the first feature vector, the object category corresponding to the first HRRP data is determined.

2. The multi-radar cooperative target recognition method based on adaptive manifold discriminant regression according to claim 1, characterized in that, The target projection matrix is ​​determined through the following steps, including: Construct the first objective function for the multi-view recognition; Initialize the target parameters in the first objective function; If the number of algorithm iterations has not reached the preset number of iterations, the target parameters are iteratively updated using the second HRRP data from at least two perspectives and the training labels corresponding to the second HRRP data from each perspective, so as to obtain the updated target parameters. Based on the updated target parameters, determine the updated projection matrix; When the algorithm iteration count reaches the preset iteration count, the updated projection matrix corresponding to the termination of the iteration is determined as the target projection matrix; the number of rows of the target projection matrix is ​​determined based on the number of feature dimensions contained in the second HRRP data, and the number of columns of the target projection matrix is ​​determined based on the preset number of object categories.

3. The multi-radar cooperative target recognition method based on adaptive manifold discriminant regression according to claim 1, characterized in that, The step of iteratively updating the target parameters using second HRRP data from at least two perspectives and the corresponding training labels for the second HRRP data from each perspective to obtain updated target parameters includes: Construct a second data matrix based on the second HRRP data from at least two perspectives; For any of the target parameters, other parameters besides the target parameter are treated as fixed parameters. The first objective function is updated according to the second data matrix, and each updated target parameter is obtained based on the updated second objective function.

4. The multi-radar cooperative target recognition method based on adaptive manifold discriminant regression according to claim 3, characterized in that, The target parameters also include an adjustment matrix, which is used for the adjustment term of adaptive weighted fusion; for any target parameter, other parameters besides the target parameter are treated as fixed parameters, the first objective function is updated according to the second data matrix, and each updated target parameter is obtained based on the updated second objective function. According to the second data matrix, any parameter among the target parameters includes: The first objective function is updated by using the nearest neighbor weight matrix, the adjustment matrix, and the view weights corresponding to each view as fixed parameters to obtain the second objective function. Calculate the gradient of the second objective function, and determine the updated projection matrix as the projection matrix corresponding to when the gradient of the second objective function is 0; The updated projection matrix, the adjustment matrix, and the view weights corresponding to each viewpoint are used as fixed parameters to update the first objective function, thereby obtaining the third objective function. Calculate the gradient of the third objective function, and determine the nearest neighbor weight matrix corresponding to the gradient of the third objective function being 0 as the updated nearest neighbor weight matrix; The updated projection matrix, the updated nearest neighbor weight matrix, and the view weights corresponding to each view are used as fixed parameters to update the first objective function, resulting in the fourth objective function. Based on the fourth objective function, determine the updated adjustment matrix; The updated projection matrix, the updated nearest neighbor weight matrix, and the updated adjustment matrix are used as fixed parameters to update the first objective function, resulting in the fifth objective function. Based on the fifth objective function, the updated view weights corresponding to each view are determined; The updated projection matrix, the updated nearest neighbor weight matrix, the updated adjustment matrix, and the updated view weights corresponding to each viewpoint are determined as the updated target parameters.

5. The multi-radar cooperative target recognition method based on adaptive manifold discriminant regression according to claim 1, characterized in that, The step of determining the first data matrix based on the first HRRP data from each of the aforementioned perspectives includes: Based on the first HRRP data from each of the aforementioned perspectives, determine the first HRRP data for each feature dimension; The first data matrix is ​​determined based on the first HRRP data of each of the aforementioned feature dimensions.

6. The multi-radar cooperative target recognition method based on adaptive manifold discriminant regression according to claim 1, characterized in that, The first feature vector corresponding to the first data matrix is ​​obtained by performing feature projection on the first data matrix according to the target projection matrix. Multiply the target projection matrix by the first data matrix to obtain the feature vector matrix; Based on the eigenvector matrix, the first eigenvector is determined.

7. The multi-radar cooperative target recognition method based on adaptive manifold discriminant regression according to claim 1, characterized in that, Based on the first feature vector, the object category corresponding to the first HRRP data is determined, including: For any first feature vector, determine the Euclidean distance between the first feature vector and the second feature vector corresponding to the second HRRP data of each viewpoint; Based on the Euclidean distance between the first feature vector and the second feature vector corresponding to the second HRRP data of each viewpoint, at least one neighbor that is closest to the first feature vector is determined. Based on the preset object category corresponding to the nearest neighbor, determine the object category corresponding to the first feature vector; Based on the object category corresponding to each of the first feature vectors, the object category corresponding to the first HRRP data is determined.

8. A multi-radar cooperative target recognition device based on adaptive manifold discriminant regression, characterized in that, include: The acquisition module is used to acquire the first HRRP data of at least two viewpoints corresponding to the object to be identified; The identification module is used to determine the first data matrix based on the first HRRP data from each of the aforementioned viewpoints; Based on the target projection matrix, feature projection is performed on the first data matrix to obtain the first feature vector corresponding to the first data matrix; The target projection matrix is ​​obtained by iteratively updating an initial projection matrix using second HRRP data from at least two perspectives, training labels corresponding to the second HRRP data from each perspective, and a first objective function for multi-view recognition. The training labels are associated with second feature vectors corresponding to the second HRRP data from each perspective, and the second feature vectors corresponding to the second HRRP data from each perspective are used to determine the object category corresponding to the second HRRP data. The objective parameters of the first objective function include the viewpoint weights, projection matrix, and nearest neighbor weight matrix corresponding to each perspective. The nearest neighbor weight matrix is ​​used to characterize the weights between nearest neighbor samples in a preset category under each perspective. Based on the first feature vector, the object category corresponding to the first HRRP data is determined.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the multi-radar cooperative target recognition method based on adaptive manifold discriminant regression as described in any one of claims 1 to 7.

Images