Target intelligent fusion detection method and system only supported by to-be-detected data, terminal and medium
By mining the echo characteristics of sparse scattering point targets and utilizing the generalized inner product and iterative estimation method, an intelligent target fusion detection method that only requires the support of the data to be tested is constructed. This solves the performance degradation problem of radar detectors under zero-sample conditions and achieves effective target detection and efficient algorithm performance in extreme cases.
Patent Information
- Application Number
- CN202510809792.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-17
- Publication Date
- 2025-09-16
AI Technical Summary
Existing radar range-extended target detection methods suffer from performance degradation under zero-sample conditions and are unable to effectively adapt to unknown target characteristics and environmental changes, resulting in detector failure. Especially in dynamically changing environments and application scenarios with high real-time requirements, the detection effectiveness and reliability of existing methods are insufficient.
By mining and utilizing the echo characteristics of sparse scattering point targets, the generalized inner product method is used to sort the main data and estimate the clutter covariance matrix. Combined with the estimation of the number of unknown scattering points, an intelligent target fusion detection method is constructed that only requires the support of the data to be tested. The clutter covariance matrix estimation is optimized through an iterative process to improve the applicability and accuracy of the detection algorithm.
It can still effectively detect targets in extreme sample-free situations, which improves the applicability and robustness of the detection algorithm, reduces computational complexity, improves detection performance and reliability, ensures the validity and stability of the estimation results, and avoids instability problems in numerical calculations.
Smart Images
Figure CN120652465A_ABST
Abstract
Claims
1. A target intelligent fusion detection method that only requires the support of the data to be tested, characterized in that: The following steps are involved: S1, obtain the detection data of K distance units to be detected as the main data Z = [z1, z2, ..., z K ], use the main data Z to calculate the initial value of the clutter covariance matrix Estimate the initial value based on the clutter covariance matrix Sort the master data Z by the generalized inner product; S2, perform scattering point estimation on the sorted master data Z to obtain the estimated number of unknown scattering points in the distance unit to be detected S3, using the unknown number of scattering points to estimate Differentiate the master data Z to obtain the master data estimate without target scattering points and estimation of master data containing target scatter points Then use the master data to estimate Recalculate the clutter covariance matrix estimate And according to Reorder the master data Z by generalized inner product and execute step S2 again; then detect the number of unknown scattering points in the current and previous time. Are they the same? If yes, go to step S4; otherwise, go to step S3. S4, using and Obtain clutter covariance matrix estimate According to the generalized likelihood ratio detection criterion, the clutter covariance matrix is estimated Obtain range-extended target detection statistics; S5, determining whether the range-extended target detection statistic is greater than or equal to a predetermined threshold; if so, determining that a target exists within the range unit to be detected; otherwise, determining that no target exists within the range unit to be detected.
2. The target intelligent fusion detection method according to claim 1, which only requires the support of the data to be tested, is characterized in that: Estimate the initial value based on the clutter covariance matrix Sort the master data Z by the generalized inner product, specifically including: Estimate initial values based on auxiliary data covariance The generalized inner product statistic is established for the master data Z, and the generalized inner product statistic μ of the i-th master data is i It can be expressed as, Where i = 1, 2, ..., K; For the main data, according to the generalized inner product statistic μ i Sort from large to small, assuming that the i-th statistic after sorting is μ (i) , then the main data Z is sorted and expressed as Z=[z (1) ,z (2) ,…,z (K) ].
3. The target intelligent fusion detection method requiring only the support of the data to be tested according to claim 1 is characterized in that: Step S2 specifically includes: S21, assuming that there are an unknown number of target scattering points p in the master data Z, that is, a multiple hypothesis H with an unknown number of scattering points is formed. p , in H p Assume that Z p =[z (1) ,z (2) ,…,z (p) ] represents the main data containing the target scattering point, ZZ p =[z (p+1) ,z (p+2) ,…,z (K) ] represents the main data without target scattering points; S22, assuming that the target scattering point master data Z p =[z (1) ,z (2) ,…,z (p) ] and master data ZZ without target scattering points p =[z (p+1) ,z (p+2) ,…,z (K) ] is known, and is represented by M y and M n Under this assumption, the maximum value of K likelihood functions is optimized when the number of scattering points is unknown, and the M y and M n Estimation expression of the number of scattering points; S23, using the master data Z containing the target scattering points p Calculate the covariance matrix estimate of the main data containing the target scatter points Using the master data ZZ that does not contain target scatter points p Obtain the covariance matrix estimate of the main data without target scattering points S24, use and Replace M in the scattering point number estimation expression y and M n , obtain the estimated number of unknown scattering points within the distance unit to be detected 4. The target intelligent fusion detection method according to claim 3, which only requires the support of the data to be tested, is characterized in that: Under this assumption, the maximum value of K likelihood functions is optimized when the number of scattering points is unknown, and the M y and M n The estimated expression of the number of scattering points includes: The joint conditional probability density function of Z of the master data is expressed as, Where det(·) represents the determinant of the matrix, and exp(·) represents the exponential function with the natural coefficient e as the base; In order to make f(Z|M n ,M y ; p) takes the largest value p, Expressed as That is, the estimated expression for the number of scattering points; Furthermore, the number of unknown scattering points is estimated Expressed as 5. The target intelligent fusion detection method requiring only the support of the data to be tested according to claim 3 is characterized in that: Step S23 specifically includes: Assume M n and M y Satisfies the noise additive structure, and the noise power is 1, that is, M x =I N +A x ,x=n,y,where I N is the N×N dimensional identity matrix, A x is an N×N dimensional unknown matrix, where N is the number of channels of the radar system; therefore, M n and M y The fast maximum likelihood estimation of is expressed as: Among them, Φ n and Λ n Respectively The eigenvector matrix and eigenvalue diagonal matrix after eigenvalue decomposition, Φ y and Λ y for The eigenvector matrix and eigenvalue diagonal matrix after eigenvalue decomposition; n =Diag(λ n,1 ,λ n,2 ,…,λ n,N ) and Λ y =Diag(λ y,1 ,λ y,2 ,…,λ y,N ) is loaded to a size equal to the noise power, and the reconstruction matrix is obtained and Right now:
6. The target intelligent fusion detection method requiring only the support of the to-be-tested data according to claim 1 is characterized in that: Clutter covariance matrix estimation Expressed as, Where s is the known target normalized steering vector.
7. The target intelligent fusion detection method according to any one of claims 1 to 6, which only requires the support of the data to be tested, is characterized in that: The distance extended target detection statistic is expressed as, 8. An intelligent fusion detection system for targets that only requires the support of the data to be tested, characterized in that: include: The first master data sorting module is used to obtain the detection data of K distance units to be detected as the master data Z = [z1, z2, ..., z K ], use the main data Z to calculate the initial value of the clutter covariance matrix Estimate the initial value based on the clutter covariance matrix Sort the master data Z by the generalized inner product; The scattering point number estimation module is used to estimate the scattering points of the sorted master data Z and obtain the estimated number of unknown scattering points in the distance unit to be detected. The second main data sorting module is used to estimate the number of unknown scattering points Differentiate the master data Z to obtain the master data estimate without target scattering points and estimation of master data containing target scatter points Then use the master data to estimate Recalculate the clutter covariance matrix estimate And according to Reorder the master data Z by generalized inner product and execute the scattering point number estimation module again; then, detect the unknown scattering point number estimation between this time and the previous time. Are they the same? If yes, execute the detection statistic calculation module; otherwise, execute the second main data sorting module; Detection statistics calculation module, used to utilize and Obtain clutter covariance matrix estimate According to the generalized likelihood ratio detection criterion, the clutter covariance matrix is estimated Obtain range-extended target detection statistics; The target existence judgment module is used to judge whether the range-extended target detection statistic is greater than or equal to a predetermined threshold. If so, it is determined that a target exists in the range unit to be detected; otherwise, it is determined that no target exists in the range unit to be detected.
9. A terminal, characterized in that: include: A memory for storing a target intelligent fusion detection program that only requires support from the data to be inspected; A processor is used to implement the steps of the target intelligent fusion detection method that only requires support from data to be tested as claimed in any one of claims 1 to 7 when executing the target intelligent fusion detection program that only requires support from data to be tested.
10. A computer-readable storage medium, characterized in that The readable storage medium stores a target intelligent fusion detection program that only requires support from data to be tested. When the target intelligent fusion detection program that only requires support from data to be tested is executed by the processor, the steps of the target intelligent fusion detection method that only requires support from data to be tested as described in any one of claims 1 to 7 are implemented.