CN-121978633-A - Radar target resolution one-dimensional range profile identification method

CN121978633ACN 121978633 ACN121978633 ACN 121978633ACN-121978633-A

Abstract

The application discloses a radar target resolution one-dimensional range profile identification method, and relates to the field of radar signal processing. The method comprises the steps of S1, constructing a target high-resolution one-dimensional range profile training sample set X, a tag matrix H and a tag block diagonal matrix Q, S2, extracting features from the training sample set X and the tag block diagonal matrix Q, optimizing to obtain a feature dictionary matrix D and a feature representation matrix Z, S3, constructing a linear classifier C by using the feature representation matrix Z and the tag matrix H, S4, constructing a feature representation matrix Zt of a test sample set Xt according to the feature dictionary matrix D, and S5, performing target classification by using the feature representation matrix Zt of the test sample set Xt and the linear classifier C. The method is used for solving the problem of poor training effect of the traditional target recognition method.

Inventors

HAN CHANG
BAI JIAN
ZHANG XIAOYANG
SHI YUETING
LIU XIAOLU
Han Xuehan

Assignees

北京遥感设备研究所

Dates

Publication Date: 20260505
Application Date: 20251230

Claims (10)

1. The radar target resolution one-dimensional range profile identification method is characterized by comprising the following steps of: s1, constructing a target high-resolution one-dimensional range profile training sample set X, a tag matrix H and a tag block diagonal matrix Q; S2, extracting features from the training sample set X and the tag block diagonal matrix Q, and optimizing to obtain a feature dictionary matrix D and a feature representation matrix Z; s3, constructing a linear classifier C by utilizing the characteristic representation matrix Z and the label matrix H; s4, constructing a feature representation matrix Z t of the test sample set X t according to the feature dictionary matrix D; S5, performing target classification by using the characteristic representation matrix Z t of the test sample set X t and the linear classifier C.
2. The method according to claim 1, wherein S2 comprises: and obtaining a feature dictionary matrix D and a feature representation matrix Z through solving the following optimization: Wherein, |·| * is the dual norm of the matrix, |·| 1 is the l 1 -norm of the matrix, |·| F is the F-norm of the matrix, and makes Z and Q as similar as possible, α, β, γ, λ are weight parameters, and E is the noise matrix contained in the training sample data.
3. The method according to claim 2, wherein said S3 comprises: the linear classifier C was constructed as follows: Wherein h= [ H 1 ,h 2 ,…,h N ],h i is the label of the i-th sample, N is the number of samples, and i·i 2 is the l 2 -norm of the matrix.
4. A method according to claim 3, wherein S4 comprises: Extracting a feature representation matrix Z t by solving the following formula: Where E t is the noise matrix of the test sample set.
5. The method of claim 4, wherein S5 comprises: object classification is performed by the following formula: Wherein Z tm is the m-th column vector of the feature representation matrix Z t , namely the feature representation of the m-th test sample, and l is the prediction label of the obtained m-th test sample.
6. The method of claim 2, wherein extracting features by solving for: s21, introducing variables J and L, and performing function rewriting by using an extended Lagrangian multiplier method to obtain a rewritten objective function as follows: S22, further rewriting the rewritten objective function into an augmented Lagrangian function corresponding to the objective function, wherein the augmented Lagrangian function is as follows: Wherein < a, B > = tr (a T , B), tr () represents the trace of the matrix, the superscript T represents the transposed operation of the matrix, Y 1 、Y 2 、Y 3 is the lagrange multiplier, μ >0 is the penalty factor; S23, solving an augmented Lagrangian function corresponding to the objective function through repeated iteration updating of the variable Z, L, E, D, J to obtain a feature dictionary matrix D and a feature representation matrix Z.
7. The method according to claim 1, wherein S1 comprises: Constructing a label block diagonal matrix Q according to the actual labels of the training samples, wherein Q= [ Q 1 ,q 2 ,…,q N ],q i ] is an idealized label representation of the ith sample, and N is the number of samples.
8. A radar target-resolved one-dimensional range profile recognition device, comprising: the matrix construction module is used for constructing a target high-resolution one-dimensional range profile training sample set X, a label matrix H and a label block diagonal matrix Q; The feature extraction module is used for extracting features from the training sample set X and the tag block diagonal matrix Q to obtain a feature dictionary matrix D and a feature representation matrix Z; The classifier generating module is used for constructing a linear classifier C by utilizing the characteristic representation matrix Z and the label matrix H; The test sample extraction module is used for constructing a feature representation matrix Z t of the test sample set X t according to the feature dictionary matrix D; And the target classification module is used for performing target classification by utilizing the characteristic representation matrix Z t of the test sample set X t and the linear classifier C.
9. An electronic device, comprising: A processor and a memory for storing computer executable instructions which when executed cause the processor to perform the steps of the method of any one of claims 1 to 7.
10. A storage medium, comprising: The storage medium stores thereon a processing program for radar target-resolved one-dimensional range profile identification, which when executed by a processor implements the method steps of any of claims 1-7.

Description

Radar target resolution one-dimensional range profile identification method Technical Field The application relates to the technical field of radar signal processing, in particular to a radar target high-resolution one-dimensional range profile identification method. Background Radar automatic target recognition (radar automatic target recognition, RATR) is a technique that determines the class and properties of a target by analyzing echoes scattered by the target. In RATR technology, since the target-resolved one-dimensional range profile (High resolution range profile, HRRP) contains abundant physical information such as the size and structure of the target, and compared with the synthetic aperture image of the target, the echo is easier to acquire and process, so that the method has very important research significance and practical value. The traditional target recognition method is to put the manually extracted features into a classification model such as machine learning and the like for classification recognition. In addition, representation learning and deep learning models are also used for HRRP recognition, which have the advantage of feature autonomous learning. Representation learning is a method for automatically extracting input data and can be used for classifying and identifying task feature expression. The method utilizes related methods such as sparse representation and dictionary learning combination, and is also applied to the field of HRRP identification. In addition, various deep learning models have good effects in the field of HRRP identification. The method has good effect in the field of target HRRP identification, but has certain limitation. On the one hand, they only focus on the features that a single sample has, ignoring the structural information between like samples. On the other hand, in a complex environment of practical application, the target HRRP recognition faces the problem of poor training effect due to noise interference of training sample data. Disclosure of Invention The application aims to provide a radar target resolution one-dimensional range profile recognition method, which solves the problem of poor training effect of the traditional target recognition method. In order to achieve the above purpose, the application adopts the following technical scheme: in one aspect, the application provides a radar target resolution one-dimensional range profile identification method, which comprises the following steps: s1, constructing a target high-resolution one-dimensional range profile training sample set X, a tag matrix H and a tag block diagonal matrix Q; s2, extracting features from the training sample set X and the tag block diagonal matrix Q to obtain a feature dictionary matrix D and a feature representation matrix Z; s3, constructing a linear classifier C by utilizing the characteristic representation matrix Z and the label matrix H; s4, constructing a feature representation matrix Z t of the test sample set X t according to the feature dictionary matrix D; S5, performing target classification by using the characteristic representation matrix Z t of the test sample set X t and the linear classifier C. On the other hand, the application also provides a radar target resolution one-dimensional range profile recognition device, which comprises: the matrix construction module is used for constructing a target high-resolution one-dimensional range profile training sample set X, a label matrix H and a label block diagonal matrix Q; The feature extraction module is used for extracting features from the training sample set X and the tag block diagonal matrix Q to obtain a feature dictionary matrix D and a feature representation matrix Z; The classifier generating module is used for constructing a linear classifier C by utilizing the characteristic representation matrix Z and the label matrix H; The test sample extraction module is used for constructing a feature representation matrix Z t of the test sample set X t according to the feature dictionary matrix D; And the target classification module is used for performing target classification by utilizing the characteristic representation matrix Z t of the test sample set X t and the linear classifier C. Based on the technical scheme, the application can obtain the following technical effects: And the global structural characteristics of the high-resolution one-dimensional range profile are characterized through low-rank constraint, so that the influence of noise is reduced. On the basis of retaining the low-rank model effect, the similar samples have more similar representation by adding the structural constraint, meanwhile, the difference between different categories is enlarged, and compared with the traditional machine learning, the method is more beneficial to classifying and identifying the target under the noise condition, and has practical value for target resolution one-dimensional range profile identification rese