A method for detecting small targets in sea clutter based on characteristic manifold

Abstract

The present invention discloses a method for detecting small targets in sea clutter based on characteristic manifolds, which relates to the technical field of small target detection. The extraction of the significant features: a series of coherent pulses are emitted on each transceiver channel, and a complex time series of each distance unit is received. The radar target detection problem in sea clutter can be transformed into the following binary hypothesis testing problem. The method for detecting small targets in sea clutter based on characteristic manifolds described in the present invention is tested using an SDRDSP data set and shows that it provides higher stability and higher detection probability compared with non-coherently accumulated CA‑CFAR and single feature detection methods; the geometric distance derived from the characteristic manifold effectively suppresses sea clutter and enhances the target discrimination capability; in summary, the characteristic manifold method represents a powerful and effective advancement in radar target detection under complex sea clutter conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of small target detection, and in particular to a method for detecting small targets in sea clutter based on characteristic manifolds. Background Art

[0002] Extracting various significant features from radar echoes to achieve radar detection of small targets such as boats, icebergs and frogmen is considered an effective detection method. Features usually intuitively or empirically represent certain statistical information extracted from radar echoes, which have a certain ability to classify targets and sea clutter. So far, dozens of explicit and interpretable empirical features extracted from time domain, frequency domain and transform domain have been applied to radar target detection;

[0003] Different features are suitable for different situations, and their complementarity is the basis for the robustness and good detection performance of feature-based detectors. Therefore, multi-feature-based object detection is an idea to improve detection performance. Commonly used multi-feature object detection methods are usually based on experience, selecting multiple features to construct a multidimensional feature space, and using specific classifiers for training. However, these methods usually rely on empirical analysis for feature selection and comprehensive use, and rarely consider the correlation between features;

[0004] Based on the information geometry radar target detection theory, the multi-dimensional feature samples are modeled into a high-dimensional matrix manifold. The corresponding clutter geometric center point is estimated using the reference unit, and the geometric topological distance with the manifold geometric point of the unit to be detected is calculated. The corresponding detector is designed using the new manifold distance statistics obtained by calculation. The proposed method has better performance than other feature-based target detection methods through actual data verification on the SDRDSP public radar dataset.

[0005] Target detection methods based on multidimensional features are often used to detect small targets in sea clutter, but existing algorithms do not fully utilize the correlation information between features, which limits their performance to a certain extent. Summary of the invention

[0006] The present invention provides a method for detecting small targets in sea clutter based on characteristic manifolds to solve the problems raised in the above background technology.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0008] A method for detecting small targets in sea clutter based on characteristic manifolds, including extraction of significant features and establishment of characteristic manifolds;

[0009] Extraction of the salient features:

[0010] By transmitting a series of coherent pulses on each transceiver channel and receiving a complex time series for each range cell, the radar target detection problem in sea clutter can be transformed into the following binary hypothesis testing problem;

[0011] ;

[0012] in, N represents the length of the received complex time series, K Represents the number of reference cells (RC), and It is the time series of radar echoes received at the cell under test (CUT) and the reference cell. and is the sea clutter time series in the unit to be detected and the reference unit, is an unknown radar echo; the null hypothesis H0 and the alternative hypothesis H1 correspond to the existence of only sea clutter or sea clutter and target in the unit to be detected respectively;

[0013] The establishment of the characteristic manifold: the eleven radar echo features used include: relative average amplitude (RAA), time domain entropy mean (TEM), relative peak height (RPH), relative Doppler peak height (RDPH), relative Doppler vector entropy (RVE), second moment of frequency domain entropy (SOFE), ridge integration (RI), maximum size of connected regions (MS), number of connected regions (NR), ridges radon transform maximum value (RRT-MV), and ridges radon transform band width (RRT-BW).

[0014] Preferably, the radar echo time series is used and Calculating the features of the echo and using a classifier to complete the decision of the inspection problem is called a feature-based target detection method.

[0015] Preferably, a sliding window is used to extract the above features from the radar echo data of the detection unit and the reference unit to obtain a feature matrix of the corresponding unit. ,in For the i The corresponding features K The value of samples.

[0016] Preferably, based on the information geometry target detection theory, the correlation between features can be used to model it as a Toeplitz positive definite matrix:

[0017] ;

[0018] in, Representation characteristics i With features j The correlation between Represents statistical expectation. In addition, the sample average is used to estimate the statistical expectation. The matrix The essence of is the covariance matrix obtained by summing up the multi-dimensional eigenvector samples. According to statistical principles, it is a positive definite matrix.

[0019] Preferably, the characteristic positive definite matrix of the unit to be detected can be estimated and the characteristic positive definite matrix of the reference unit ; Such a set of topolithic Hermitian positive definite matrices constitutes a matrix manifold:

[0020] ;

[0021] in, Represents an 11×11 real matrix.

[0022] Preferably, based on the information geometry target detection method, the target detection problem is transformed into a matrix manifold The geometric problem on the manifold is solved, and the geometric distance measurement on the manifold is used to distinguish the target from the clutter, and finally determine whether the target exists or not.

[0023] The corresponding detection decision formula is:

[0024] ;

[0025] in, represents the manifold geometric distance, Represents the detection threshold, Represented by the reference unit characteristic positive definite matrix The estimated clutter geometric center can be expressed as the solution to the following optimization problem:

[0026] .

[0027] Preferably, since the matrix manifold is a high-dimensional manifold with a certain geometric structure, the difference of sample data in the manifold needs to be measured by geodesic distance. In this paper, the log-Euclidean (LE) distance is used for two points on the matrix manifold. , the LE distance between them is:

[0028] ;

[0029] in, represents the matrix norm, Represents matrix logarithm operation; LE distance is used as the distance measure between two points on the matrix manifold. n Reference Unit , its LE geometric center is:

[0030] ;

[0031] in, Represents matrix exponential operation.

[0032] Due to the adoption of the above technical solution, the present invention has the following technical advances compared with the prior art:

[0033] The present invention provides a method for detecting small targets in sea clutter based on characteristic manifolds. Tests using the SDRDSP dataset show that the method provides higher stability and higher detection probability than the non-coherently accumulated CA-CFAR and single-feature detection methods. The geometric distance derived from the characteristic manifold effectively suppresses sea clutter and enhances the target discrimination capability. Overall, the characteristic manifold method represents a powerful and effective advancement in radar target detection under complex sea clutter conditions.

[0034] The present invention provides a method for detecting small targets in sea clutter based on characteristic manifolds. The application of the characteristic manifold method significantly enhances the performance of feature-based detection technology and provides higher reliability and stability in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the geometric meaning of the geometric center and manifold distance on the characteristic manifold of the present invention;

[0036] Figure 2 It is a complete process schematic diagram of the method proposed by the present invention;

[0037] Figure 3 It is a schematic diagram of detection probability under different sea conditions of the present invention;

[0038] Figure 4 It is a schematic diagram for comparing the detection threshold of the present invention under 5 levels of sea conditions with the detection statistics of the proposed algorithm;

[0039] Figure 5 It is a schematic diagram for comparing the detection threshold and the RAA detection statistic under level 5 sea conditions of the present invention. DETAILED DESCRIPTION

[0040] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0041] like Figure 1-Figure 5 As shown, a method for detecting small targets in sea clutter based on characteristic manifolds includes extraction of significant features and establishment of characteristic manifolds;

[0042] Extraction of salient features:

[0043] By transmitting a series of coherent pulses on each transceiver channel and receiving a complex time series for each range cell, the radar target detection problem in sea clutter can be transformed into the following binary hypothesis testing problem;

[0044] ;

[0045] in, N represents the length of the received complex time series, K Represents the number of reference cells (RC), and It is the time series of radar echoes received at the cell under test (CUT) and the reference cell. and is the sea clutter time series in the unit to be detected and the reference unit, is an unknown radar echo; the null hypothesis H0 and the alternative hypothesis H1 correspond to the existence of only sea clutter or sea clutter and target in the unit to be detected respectively;

[0046] Establishment of feature manifold: The eleven radar echo features used include: Relative Average Amplitude (RAA), Time domain Entropy Mean (TEM), Relative Peak Height (RPH), Relative Doppler Peak Height (RDPH), Relative Vector Entropy (RVE), Second moment of Frequency domain Entropy (SOFE), Ridge Integration (RI), Maximum Size of connected regions (MS), Number of connected Regions (NR), Ridges Radon Transform Maximum Value (RRT-MV), and Ridges Radon Transform Band Width (RRT-BW).

[0047] Using radar echo time series and Calculating the features of the echo and using a classifier to complete the decision of the inspection problem is called a feature-based target detection method.

[0048] Use sliding windows to extract the above features from the radar echo data of the detection unit and the reference unit to obtain the feature matrix of the corresponding unit ,in For the i The corresponding features K Here, Indicates d The original data of the reference unit, Indicates d The feature matrix of the reference unit, Indicates d The characteristic matrix manifold of reference cells. Represents the characteristic matrix manifold of the unit to be detected, Represents the original data of the unit to be detected, Represents the feature matrix of the unit to be detected.

[0049] Based on the information geometry target detection theory, the correlation between features can be modeled as a Toeplitz positive definite matrix:

[0050] ;

[0051] in, Representation characteristics i With features j The correlation between represents the statistical expectation. In addition, we use the sample average to estimate the statistical expectation; the matrix The essence of is the covariance matrix obtained by summing up the multi-dimensional eigenvector samples. According to statistical principles, it is a positive definite matrix.

[0052] The characteristic positive definite matrix of the unit to be detected can be estimated and the characteristic positive definite matrix of the reference unit ; Such a set of topolithic Hermitian positive definite matrices constitutes a matrix manifold:

[0053] ;

[0054] in, Represents an 11×11 real matrix.

[0055] Based on the information geometry target detection method, the target detection problem is transformed into a matrix manifold The geometric problem on the manifold is solved, and the geometric distance measurement on the manifold is used to distinguish the target from the clutter, and finally determine whether the target exists or not.

[0056] The corresponding detection decision formula is:

[0057] ;

[0058] in, represents the manifold geometric distance, Represents the detection threshold, Represented by the reference unit characteristic positive definite matrix The estimated clutter geometric center can be expressed as the solution to the following optimization problem:

[0059] .

[0060] Since the matrix manifold is a high-dimensional manifold with a certain geometric structure, the difference of sample data in the manifold needs to be measured by geodesic distance. In this paper, the log-Euclidean (LE) distance is used to measure the difference between two points on the matrix manifold. , the LE distance between them is:

[0061] ;

[0062] in, represents the matrix norm, Represents matrix logarithm operation; LE distance is used as the distance measure between two points on the matrix manifold. n Reference Unit , its LE geometric center is:

[0063] ;

[0064] in, Represents matrix exponential operation.

[0065] like Figure 3 As shown in Figure 2, compared with other public target detection radar datasets, the SDRDSP dataset covers a wider range of sea conditions and contains more pure sea clutter units for reference. The performance of the proposed method is evaluated under different sea conditions with a false alarm probability of 10 -2 The performance of the proposed method is compared with the non-coherently accumulated cell-averaged constant false alarm rate (CA-CFAR) detection method and the detection method using RAA as the detection statistic.

[0066] like Figure 4 and Figure 5 As shown in the figure, the detection performance curves of each method show that the two feature-based small target detection techniques perform better than the non-coherent accumulation CA-CFAR detection method under high sea conditions; in addition, the detection curve of the proposed method shows the smallest change with the improvement of sea conditions, indicating that it has the highest stability; in particular, under sea conditions of level 5, the detection probability of the proposed method is more than 10% higher than that of the detection method based on RAA features;

[0067] The detection statistics and detection threshold curves of a single frame are plotted. The results show that using the geometric distance derived from the feature manifold as the detection statistic can effectively suppress sea clutter, thereby enhancing the distinction between the target and sea clutter and reducing the probability of false alarms. In contrast, the detection performance using a single feature as the detection statistic is poor, which is intuitively manifested in the small difference between the values ​​of the target unit and the clutter unit, and the characteristic value of the target unit cannot reach the threshold calculated by CA-CFAR. Therefore, the application of the feature manifold method significantly enhances the performance of feature-based detection technology and provides higher reliability and stability in complex environments.

[0068] The working principle of the present invention is as follows: by converting radar echo feature data into high-dimensional manifolds, the radar target detection problem is reformulated as the problem of distinguishing geometric points on the feature matrix manifold; this method integrates multi-dimensional features and their correlations, significantly improving the detection performance; tests using the SDRDSP dataset show that our method provides higher stability and higher detection probability compared with the non-coherently accumulated CA-CFAR and single-feature detection methods; the geometric distance derived from the feature manifold effectively suppresses sea clutter and enhances the target distinction capability; in summary, the feature manifold method represents a powerful and effective advancement in radar target detection under complex sea clutter conditions.

[0069] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for detecting small targets in sea clutter based on characteristic manifolds, characterized by: Including the extraction of significant features and the establishment of feature manifolds; Extraction of the salient features: By transmitting a series of coherent pulses on each transceiver channel and receiving a complex time series for each range cell, the radar target detection problem in sea clutter can be transformed into the following binary hypothesis testing problem; ； in, N represents the length of the received complex time series, K represents the number of reference cells, and It is the time series of radar echoes received at the cell under test (CUT) and the reference cell. and is the sea clutter time series in the unit to be detected and the reference unit, is an unknown radar echo; the null hypothesis H0 and the alternative hypothesis H1 correspond to the existence of only sea clutter or sea clutter and target in the unit to be detected respectively; The establishment of the characteristic manifold: the eleven radar echo characteristics utilized include: relative average amplitude, time domain entropy value mean, relative peak height, relative Doppler peak height, relative Doppler vector entropy, second-order moment of frequency domain entropy value, time-frequency ridge accumulation, maximum connected area size, number of connected areas, time-frequency ridge transform domain peak value, and time-frequency ridge transform domain bandwidth.

2. The method for detecting small targets in sea clutter based on characteristic manifold according to claim 1, characterized in that: Using radar echo time series and Calculating the features of the echo and using a classifier to complete the decision of the inspection problem is called a feature-based target detection method.

3. The method for detecting small targets in sea clutter based on characteristic manifold according to claim 1, characterized in that: Use sliding windows to extract the above features from the radar echo data of the detection unit and the reference unit to obtain the feature matrix of the corresponding unit ,in For the i The corresponding features K The value of samples.

4. The method for detecting small targets in sea clutter based on characteristic manifold according to claim 3, characterized in that: Based on the information geometry target detection theory, the correlation between features can be modeled as a Toeplitz positive definite matrix: ； in, Representation characteristics i With features j The correlation between Represents statistical expectation. In addition, the statistical expectation is estimated using the sample average. The matrix The essence of is the covariance matrix obtained by summing up the multi-dimensional eigenvector samples. According to statistical principles, it is a positive definite matrix.

5. The method for detecting small targets in sea clutter based on characteristic manifold according to claim 4, characterized in that: The characteristic positive definite matrix of the unit to be detected can be estimated and the characteristic positive definite matrix of the reference unit ; Such a set of topolithic Hermitian positive definite matrices constitutes a matrix manifold: ； in, Represents an 11×11 real matrix.

6. The method for detecting small targets in sea clutter based on characteristic manifold according to claim 5, characterized in that: Based on the information geometry target detection method, the target detection problem is transformed into a matrix manifold The geometric problem on the manifold is solved, and the geometric distance measurement on the manifold is used to distinguish the target from the clutter, and finally determine whether the target exists or not. The corresponding detection decision formula is: ； in, represents the manifold geometric distance, Represents the detection threshold, Represented by the reference unit characteristic positive definite matrix The estimated clutter geometric center can be expressed as the solution to the following optimization problem: 。 7. The method for detecting small targets in sea clutter based on characteristic manifold according to claim 6, characterized in that: Since the matrix manifold is a high-dimensional manifold with a certain geometric structure, the difference of sample data in the manifold needs to be measured by geodesic distance. In this paper, logarithmic Euclidean distance is used for two points on the matrix manifold. , the LE distance between them is: ； in, represents the matrix norm, Represents matrix logarithm operation; LE distance is used as the distance measure between two points on the matrix manifold. n Reference Unit , its LE geometric center is: ； in, Represents matrix exponential operation.

Images