A method for detecting and intelligently identifying a target of a missile-borne radar in a hypersonic speed

Abstract

The application belongs to the field of radar target detection and recognition, and is suitable for solving the problem of detecting low-speed targets by missile-borne radar with large incident residual angle and high sea conditions in the terminal guidance phase of hypersonic vehicles. Firstly, the sea clutter spectrum is approximated by a Gaussian curve to obtain the prior spectrum under different incident residual angles and different sea conditions. The block median method is used for real-time spectrum measurement of sea clutter. The KL divergence is used for sea clutter spectrum estimation and normalization suppression. The spectrum of sea clutter is whitened. After short-time Fourier transform, a new two-dimensional matrix signal is obtained. The support vector machine is used for sea clutter discrimination, so as to realize the anti-sea clutter of terminal guidance radar. The application combines the prior and posterior of sea clutter spectrum, improves the stability and accuracy of detection under clutter, and selects the shape, duration and RCS of sea clutter as characteristic quantities, so that the intelligibility of intelligent identification is strong.

Description

Technical Field

[0001] This invention belongs to the field of radar target detection and identification research, and is applicable to solving the problem of detecting and identifying low-speed targets in the background of sea clutter when hypersonic vehicles attack at a large incident angle at the terminal stage. Background Technology

[0002] Hypersonic missile-borne radar has wide applications in space, with a key focus and challenge being the detection of low-speed targets on the sea surface, such as ships, buoys, and floating objects. In anti-ship applications, radar guidance is the commonly used terminal guidance method for aircraft. To achieve rapid strikes and high-altitude penetration, the incident covariance angle (HGA) of missile-borne radar during the terminal guidance phase of hypersonic vehicles is generally large. Under large HGA conditions, the sea clutter environment faced by the terminal guidance radar is much harsher than that of traditional low-altitude penetration anti-ship aircraft. The increased backscattering intensity of sea clutter poses a risk to the radar seeker in capturing sea clutter, making it difficult to stably detect and track targets. Traditional frequency-domain constant false alarm rate (CFAR) detection methods are ineffective in detecting targets, making the detection of floating targets on the sea surface a significant challenge.

[0003] Sea clutter generally differs from Gaussian white noise, and its characteristics are related to factors such as the incident angle and sea state. The uncertainty and gradual change of sea clutter pose a severe challenge to missile-borne radars under high sea state and large incident angle conditions. Sea clutter characteristics mainly include amplitude distribution, Doppler spectrum distribution, spatial correlation, temporal correlation, and polarization characteristics. Amplitude statistical models are the most widely studied, with classic models including Rayleigh distribution, log-normal distribution, Weibull distribution, and K-distribution. With the improvement of radar resolution, the K-distribution is generally considered to be the model that best approximates the sea clutter distribution. Weinberg GV analyzed the sea clutter characteristics under incident angles of 20-50° using airborne Ingara fully polarized X-band radar data. The study found that the Pareto distribution is not only a good model for sea clutter echoes from radars with small incident angles, but it is also effective as a model for sea clutter from radars with large incident angles. It is evident that sea clutter exhibits strong correlation, and improving the signal-to-noise ratio (SNR) of coherent signal accumulation under strong sea clutter conditions faces significant challenges. Therefore, suppressing sea clutter is crucial. Clutter whitening is a commonly used method, but the uncertainty and gradual changes in the sea clutter spectrum pose challenges to traditional methods. Thus, how to achieve sea clutter whitening before accumulation is an interesting topic. Furthermore, after coherent accumulation, sea clutter and targets need further differentiation and removal. Considering the significant differences between sea clutter's morphology, duration, and amplitude in the signal and those of maritime targets, these characteristics can be used for differentiation. Machine learning has become a very popular method for target recognition in recent years, with support vector machines being the most classic approach. Therefore, a classifier can be built using support vector machines after coherent accumulation to identify and remove sea clutter, thereby improving the anti-sea clutter capability of missile-borne radar. Summary of the Invention

[0004] To address the problem of detecting slow-moving sea targets by hypersonic vehicles against a large incident angle, this invention provides a method for anti-clutter detection and intelligent identification of hypersonic missile-borne radar targets. This method achieves anti-clutter performance through sea clutter spectrum estimation and normalization suppression using KL divergence, and sea clutter identification using support vector machines. The technical solution adopted by this invention to solve the aforementioned technical problem is as follows:

[0005] A method for anti-clutter detection and intelligent identification of hypersonic missile-borne radar targets includes the following technical measures:

[0006] Step 1: Conduct experimental data analysis of sea clutter to obtain the spectral characteristics of sea clutter data under different incident angles and sea conditions, obtain the prior distribution of the sea clutter spectrum, and approximate the sea clutter spectrum using a Gaussian curve.

[0007]

[0008] in, f 0 represents the center frequency of sea clutter. Let be the standard deviation of the sea clutter spectrum, and This represents the root mean square value of the sea clutter velocity distribution. The wavelength of radar electromagnetic waves;

[0009] Step 2: Sea clutter spectrum estimation and normalization suppression based on KL divergence, the specific method is as follows:

[0010] A block median adaptive spectral estimation method is employed. Based on the parameters of the sea clutter estimation filter using the nearest-neighbor range gate, the spectral estimate of the Doppler frequency cell in the time-frequency domain is used to obtain the corresponding adaptive spectral filter. Sea clutter suppression based on the adaptive spectral filter is then performed. nk The spectral estimate is expressed by the following formula.

[0011]

[0012] The Doppler value at each frequency point is obtained by median estimation of multiple units, completing the Doppler spectrum estimation of sea clutter. The sea clutter in the signal is then normalized using the estimated Doppler spectrum. The spectral distribution of sea clutter under corresponding incident angles and sea clarity conditions, obtained a priori, is then compared with... P ( nk To compare the fits, the KL divergence was used to calculate the goodness of fit between the two methods.

[0013]

[0014] When the two distributions are completely equal =0; The center frequency of the sea clutter with Q distribution parameters is determined by minimizing the KL divergence. f 0 Spectral standard deviation of sea clutter and coefficients c 0 Thus, when our distribution best matches the prior distribution, Take the minimum value to obtain the spectrum distribution in real time. P The estimated spectrum was then used to normalize the sea clutter.

[0015] Step 3: After the sea clutter spectrum normalization and suppression processing in Step 2, perform a short-time Fourier transform on the signal to obtain a new two-dimensional matrix signal in the frequency domain. After accumulation, a new two-dimensional matrix signal is obtained.

[0016] Step 4: Intelligent sea clutter identification based on support vector machines. The specific method is as follows:

[0017] CFAR is performed on the two-dimensional matrix obtained in step three, and the constant false alarm threshold is determined using a support vector machine-based method. Since there are no targets in most of the sea clutter experimental data, a support vector machine is used to train the sea clutter data to obtain a classifier. During online CFAR detection of sea clutter, the obtained classifier is used to classify targets and sea clutter.

[0018] Specifically, the method for obtaining the classifier in step four is as follows:

[0019] (1) Pure sea clutter datasets, target signal datasets and preset false alarm probabilities for different incident angles and sea conditions;

[0020] (2) Extract features from the training data to construct a feature vector. Select three features as feature quantities: sea clutter morphology, sea clutter duration, and sea clutter RCS. Their definitions are as follows:

[0021] (a) Morphology of sea clutter: Using morphological theory, determine the number of range cells and azimuth cells occupied by sea clutter;

[0022] (b) Sea clutter duration: By utilizing the certainty of the signal and the fluctuation of the clutter, the fluctuation characteristics of the signal between different frames are compared;

[0023] (c) RCS of sea clutter: The RCS of sea clutter is inferred from the radar equation based on the distance of the sea clutter and is used as a characteristic quantity.

[0024] (3) The extracted pure sea clutter feature vectors are used to generate an SVDD classifier through the SVDD training algorithm, and the classifier thresholds for different incident angles and different sea conditions are used to form a data table;

[0025] (4) Extract the features of the test data and construct the feature vector. Calculate the feature vector and extract the SVDD classifier threshold corresponding to the current time of the platform. By comparing the feature vector and the classifier threshold, the target can be detected.

[0026] The beneficial effects of this invention are:

[0027] Compared to existing technologies, the missile-borne radar target detection method based on support vector machine threshold surfaces described in this technical solution has the following advantages:

[0028] (1) By combining the prior and posterior of sea clutter, the KL divergence is used to calculate the fit between the two under the guidance of the prior, and the sea clutter spectrum normalization is carried out in real time, which improves the stability and accuracy of clutter detection.

[0029] (2) By introducing the three quantities of sea clutter morphology, sea clutter duration and sea clutter RCS as feature quantities, the interpretability of intelligent recognition is stronger.

[0030] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can extend the scope of the technology disclosed in the present invention to other modifications, variations and applications, all of which should be covered within the scope of the present invention. Attached Figure Description

[0031] Appendix Figure 1 Flowchart of the method steps of the present invention Detailed Implementation

[0032] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings. Figure 1 The specific steps of the present invention include:

[0033] A method for anti-clutter detection and intelligent identification of hypersonic missile-borne radar targets includes the following technical measures:

[0034] Step 1: Conduct experimental data analysis of sea clutter to obtain the spectral characteristics of sea clutter data under different incident angles and sea conditions, obtain the prior distribution of the sea clutter spectrum, and approximate the sea clutter spectrum using a Gaussian curve.

[0035]

[0036] in, f 0 represents the center frequency of sea clutter. Let be the standard deviation of the sea clutter spectrum, and This represents the root mean square value of the sea clutter velocity distribution. The wavelength of radar electromagnetic waves;

[0037] Step 2: Sea clutter spectrum estimation and normalization suppression based on KL divergence, the specific method is as follows:

[0038] A block median adaptive spectral estimation method is employed. Based on the parameters of the sea clutter estimation filter using the nearest-neighbor range gate, the spectral estimate of the Doppler frequency cell in the time-frequency domain is used to obtain the corresponding adaptive spectral filter. Sea clutter suppression based on the adaptive spectral filter is then performed. nk The spectral estimate is expressed by the following formula.

[0039]

[0040] The Doppler value at each frequency point is obtained by median estimation of multiple units, completing the Doppler spectrum estimation of sea clutter. The sea clutter in the signal is then normalized using the estimated Doppler spectrum. The spectral distribution of sea clutter under corresponding incident angles and sea clarity conditions, obtained a priori, is then compared with... P ( nk To compare the fits, the KL divergence was used to calculate the goodness of fit between the two methods.

[0041]

[0042] When the two distributions are completely equal =0; The center frequency of the sea clutter with Q distribution parameters is determined by minimizing the KL divergence. f 0 Spectral standard deviation of sea clutter and coefficients c 0 Thus, when our distribution best matches the prior distribution, Take the minimum value to obtain the spectrum distribution in real time. P The estimated spectrum was then used to normalize the sea clutter.

[0043] Step 3: After the sea clutter spectrum normalization and suppression processing in Step 2, perform a short-time Fourier transform on the signal to obtain a new two-dimensional matrix signal in the frequency domain. After accumulation, a new two-dimensional matrix signal is obtained.

[0044] Step 4: Intelligent sea clutter identification based on support vector machines. The specific method is as follows:

[0045] Since the distribution of sea clutter varies depending on the sea clutter environment, CFAR is performed on the two-dimensional matrix obtained in step three, and a support vector machine-based method is used to determine the constant false alarm threshold. Because the sea clutter experimental data often lacks targets, a support vector machine is used to train the sea clutter data to obtain a classifier. During online CFAR detection of sea clutter, the acquired classifier is used to classify targets and sea clutter. The specific method for obtaining the classifier is as follows:

[0046] (1) Pure sea clutter datasets, target signal datasets and preset false alarm probabilities for different incident angles and sea conditions;

[0047] (2) Extract features from the training data to construct a feature vector. Select three features as feature quantities: sea clutter morphology, sea clutter duration, and sea clutter RCS. Their definitions are as follows:

[0048] (a) Morphology of sea clutter: Using morphological theory, determine the number of range cells and azimuth cells occupied by sea clutter;

[0049] (b) Sea clutter duration: By utilizing the certainty of the signal and the fluctuation of the clutter, the fluctuation characteristics of the signal between different frames are compared;

[0050] (c) RCS of sea clutter: The RCS of sea clutter is inferred from the radar equation based on the distance of the sea clutter and is used as a characteristic quantity.

[0051] (3) The extracted pure sea clutter feature vectors are used to generate an SVDD classifier through the SVDD training algorithm, and the classifier thresholds for different incident angles and different sea conditions are used to form a data table.

[0052] (4) Extract the features of the test data and construct the feature vector. Calculate the feature vector and extract the SVDD classifier threshold corresponding to the current time of the platform. By comparing the feature vector and the classifier threshold, the target can be detected.

Claims

1. A method for anti-clutter detection and intelligent identification of hypersonic missile-borne radar targets, characterized in that... include: Step 1: Conduct experimental data analysis of sea clutter to obtain the spectral characteristics of sea clutter data under different incident angles and sea conditions, obtain the prior distribution of the sea clutter spectrum, and approximate the sea clutter spectrum using a Gaussian curve. ； in, f 0 represents the center frequency of sea clutter. Let be the standard deviation of the sea clutter spectrum, and Let be the root mean square value of the sea clutter velocity distribution. The wavelength of radar electromagnetic waves; Step 2: Sea clutter spectrum estimation and normalization suppression based on KL divergence, the specific method is as follows: A block median adaptive spectral estimation method is employed. Based on the parameters of the sea clutter estimation filter using the nearest-neighbor range gate, the spectral estimate of the Doppler frequency cell in the time-frequency domain is used to obtain the corresponding adaptive spectral filter. Sea clutter suppression based on the adaptive spectral filter is then performed. nk The spectral estimate is expressed by the following formula. ； The Doppler value at each frequency point is obtained by median estimation of multiple units, completing the Doppler spectrum estimation of sea clutter. The sea clutter in the signal is then normalized using the estimated Doppler spectrum. The spectral distribution of sea clutter under corresponding incident angles and sea clarity conditions, obtained a priori, is then compared with... P ( nk To compare the fits, the KL divergence was used to calculate the goodness of fit between the two methods. ； When the two distributions are completely equal =0; The center frequency of the sea clutter with Q distribution parameters is determined by minimizing the KL divergence. f 0 Spectral standard deviation of sea clutter and coefficients c 0 Thus, when our distribution best matches the prior distribution, Take the minimum value to obtain the spectrum distribution in real time. P The estimated spectrum was then used to normalize the sea clutter. Step 3: After the sea clutter spectrum normalization and suppression processing in Step 2, perform a short-time Fourier transform on the signal to obtain a new two-dimensional matrix signal in the frequency domain. After accumulation, a new two-dimensional matrix signal is obtained. Step 4: Intelligent sea clutter identification based on support vector machines. The specific method is as follows: CFAR is performed on the two-dimensional matrix obtained in step three, and the constant false alarm threshold is determined using a support vector machine-based method. Since there are no targets in most of the sea clutter experimental data, a support vector machine is used to train the sea clutter data to obtain a classifier. During online CFAR detection of sea clutter, the obtained classifier is used to classify targets and sea clutter.

2. The method for anti-clutter detection and intelligent identification of hypersonic missile-borne radar targets according to claim 1, characterized in that... The specific method for obtaining the classifier in step four is as follows: (1) Pure sea clutter datasets, target signal datasets and preset false alarm probabilities for different incident angles and sea conditions; (2) Extract features from the training data to construct a feature vector. Select three features as feature quantities: sea clutter morphology, sea clutter duration, and sea clutter RCS. Their definitions are as follows: (a) Morphology of sea clutter: Using morphological theory, determine the number of range cells and azimuth cells occupied by sea clutter; (b) Sea clutter duration: By utilizing the certainty of the signal and the fluctuation of the clutter, the fluctuation characteristics of the signal between different frames are compared; (c) RCS of sea clutter: The RCS of sea clutter is inferred from the radar equation based on the distance of the sea clutter and is used as a characteristic quantity. (3) The extracted pure sea clutter feature vectors are used to generate an SVDD classifier through the SVDD training algorithm, and the classifier thresholds for different incident angles and different sea conditions are used to form a data table; (4) Extract the features of the test data and construct the feature vector. Calculate the feature vector and extract the SVDD classifier threshold corresponding to the current time of the platform. By comparing the feature vector and the classifier threshold, the target can be detected.

Images