An Adaptive Radar Target Detection Method, System, Device and Medium

Abstract

The present invention provides an adaptive radar target detection method, system, device and medium, belonging to the technical field of adaptive radar target detection. It performs high-resolution decomposition and Gaussian fitting on radar echo signals, analyzes the spectral centroid and spectral width characteristics, and determines the frequency domain interval of the radar echo signals; decomposes the radar echo signals into multiple intrinsic mode components, and screens out the mode components falling within the target frequency domain interval for reconstruction; extracts key features from the reconstructed radar echo signals from the time domain and frequency domain dimensions, and through a feature screening method, retains the features with discriminative power for target detection; divides the long-time signals in the radar echo signals into several small windows, performs VMD decomposition and reconstruction respectively; uses the CA-CFAR algorithm to perform target detection on the extracted key features. The method improves the accuracy of target detection under complex sea conditions and strong sea clutter backgrounds, and improves the accuracy and robustness of target detection.

Description

Technical Field

[0001] The present invention belongs to the technical field of adaptive radar target detection, and in particular relates to an adaptive radar target detection method, system, equipment and medium. Background Art

[0002] Marine adaptive radar target detection is widely used in fields such as ocean monitoring. However, under high sea conditions, the wave height is large and the wave fluctuations are significant, resulting in the target signal returning to the radar receiving end being often interfered by strong sea clutter. These sea clutters not only present complex spectrum expansion and multi-time and space scale changes, causing the characteristics of the target signal to be submerged or distorted, but also their statistical characteristics and distribution characteristics are significantly different from those under a stable background. The performance of traditional target detection algorithms drops sharply under high sea conditions, and there are a large number of missed detections.

[0003] Researchers have proposed a variety of signal processing methods for suppressing sea clutter under high sea conditions. Existing methods mostly rely on techniques such as bandpass filtering, pulse accumulation or spatial filtering, but their ability to suppress sea clutter is still limited. Subsequently, time-frequency adaptive analysis tools such as empirical mode decomposition (EMD) and local mean decomposition (LMD) were introduced into the field of signal processing to improve sea clutter suppression performance. However, these methods are prone to modal aliasing when processing complex multi-component signals, affecting the extraction accuracy and stability of the target signal.

[0004] Therefore, how to effectively extract target signal features and achieve stable and reliable detection in the violently fluctuating sea clutter background is a technical problem that needs to be urgently solved in the field of radar signal processing. Summary of the invention

[0005] The present invention provides an adaptive radar target detection method, which is aimed at channel buoy detection under high sea conditions. The method effectively suppresses sea clutter and enhances target signals by accurately formulating frequency domain intervals and screening target modal components, combined with the Doppler frequency shift characteristics of the buoy.

[0006] Methods include:

[0007] Acquire radar echo signals and perform high-resolution decomposition of radar echo signals;

[0008] Gaussian fitting is performed on the clutter spectrum in the radar echo signal, and the spectrum centroid and spectrum width characteristics are analyzed. Combined with the Doppler frequency shift characteristics of the radar echo signal, the frequency domain interval of the radar echo signal is determined;

[0009] Decompose the radar echo signal into multiple eigenmode components, and select the mode components falling within the target frequency domain interval for reconstruction;

[0010] Extract key features from the reconstructed radar echo signal in the time domain and frequency domain dimensions, and retain the features with discrimination for target detection through feature screening methods;

[0011] Segment the long-time signal in the radar echo signal into several small windows, and perform VMD decomposition and reconstruction respectively;

[0012] Send the reconstructed signal into the CA-CFAR detection module, and use the CA-CFAR algorithm to perform target detection on the extracted key features.

[0013] It should be further noted that the high-resolution decomposition of the radar echo signal in the steps also includes:

[0014] The radar echo signal is decomposed into modal components ;

[0015] Each modal component has a central frequency ;

[0016] Define the objective function of high-resolution decomposition as: The constraint condition is:

[0017]

[0018] where represents the time differential operator;

[0019] In the method, a Lagrangian function is also constructed;

[0020]

[0021] where is the frequency constraint parameter, is the Lagrange multiplier, represents the inner product operation;

[0022] By alternately updating , and , iteratively solve the optimization problem;

[0023]

[0024]

[0025]

[0026] where is the gradient descent learning rate; is the Fourier transform of the original signal f(t); is the kth modal component Fourier transform of; is the Lagrange multiplier Fourier transform of; is the k-th modal component Fourier transform of;

[0027] The iterative process stops when the following conditions are met;

[0028]

[0029] where, is the number of iterations, is the maximum number of iterations, is a preset convergence threshold. After VMD decomposition, modal components are obtained .

[0030] It should be further noted that the step of determining the frequency domain interval of the radar echo signal in combination with the Doppler frequency shift characteristic of the radar echo signal further includes:

[0031] Calculate the corresponding Doppler frequency range through the known radial velocity of the buoy. The relationship between the Doppler frequency shift and the radial velocity is as follows:

[0032]

[0033]

[0034] where, is the Doppler frequency shift, is the radial velocity of the buoy, is the wavelength of the signal, is the radar transmission frequency.

[0035] It should be further noted that the frequency domain interval of the target signal is determined by the following strategy:

[0036]

[0037] In the method, the modal components u k (t) of the radar echo signal are also superimposed to realize the reconstruction of the target signal;

[0038]

[0039] where, S target (t) is the reconstructed target signal, f k is the center frequency of the k-th modal signal, [f low , f high is the frequency interval.

[0040] Further, it should be noted that the method quantitatively measures the discrimination ability between the target and the background clutter by calculating the relative feature value between the target distance unit and the clutter distance unit. The relative feature value RX is defined as follows:

[0041]

[0042] Wherein, and respectively represent the corresponding feature values of the target unit and the clutter unit;

[0043] Define the change amount of the relative feature value , and express the difference in the relative feature value between the original radar echo signal and the VMD reconstructed signal on a logarithmic scale;

[0044]

[0045] Wherein, represents the relative feature value of the VMD reconstructed signal, represents the relative feature value of the original signal.

[0046] Further, it should be noted that in the method, the long-time signal S(t) is divided into several windows, each window having a length of L win , and sliding with a preset step size L step ;

[0047] Define the overlap ratio between windows as m, then the expressions for the sliding step size and the length of the overlapping region are:

[0048]

[0049]

[0050] Stitch the results of all windows into a complete signal, that is, in the overlapping region, add the overlapping part signals of the previous window and the next window and take the average value:

[0051]

[0052] Wherein, S prev (t) is the overlapping part signal of the previous window, and S next (t) is the overlapping part signal of the next window.

[0053] Further, it should be noted that the CA-CFAR algorithm includes: the number of training units and the number of guard units ;

[0054] The number of training units The number of units for estimating clutter power, the number of guard units To prevent the influence of target signals near the detection unit on clutter estimation;

[0055] Estimate the clutter power level through the reference units in the front and back windows , the clutter power level is calculated as follows:

[0056]

[0057] where is the measurement value of the th unit in the front window, is the measurement value of the th unit in the back window;

[0058] The threshold factor is an important parameter for adjusting the detection threshold and is determined by the set false alarm probability ;

[0059]

[0060] The detection threshold is calculated through the threshold factor and the clutter estimation value ;

[0061]

[0062] Finally, the radar echo signal intensity is compared with the detection threshold , and the detection decision is based on the following formula;

[0063]

[0064] If the radar echo signal intensity is greater than the detection threshold, it is determined that a target is detected.

[0065] This application also provides an adaptive radar target detection system, and the system includes:

[0066] A signal decomposition module for obtaining the radar echo signal and performing high-resolution decomposition on the radar echo signal;

[0067] A frequency domain interval selection module for performing Gaussian fitting on the clutter spectrum in the radar echo signal, analyzing the spectral centroid and spectral width characteristics, and combining the Doppler frequency shift characteristics of the radar echo signal to determine the frequency domain interval of the radar echo signal;

[0068] A decomposition and reconstruction module, which is used to decompose the radar echo signal into multiple intrinsic mode components, and screen out the mode components falling within the target frequency domain interval for reconstruction;

[0069] A feature extraction and screening module, which is used to extract key features from the reconstructed radar echo signal in the time domain and frequency domain dimensions, and retain the features with discriminative power for target detection through feature screening methods;

[0070] A windowing processing module, which is used to divide the long-time signal in the radar echo signal into several small windows, and perform VMD decomposition and reconstruction respectively;

[0071] A CA-CFAR detection module, which is used to send the reconstructed signal into the CA-CFAR detection module, and use the CA-CFAR algorithm to detect targets for the extracted key features.

[0072] According to another embodiment of the present application, there is provided an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the adaptive radar target detection method are implemented.

[0073] According to still another embodiment of the present application, there is further provided a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the adaptive radar target detection method are implemented.

[0074] It can be seen from the above technical solutions that the present invention has the following advantages:

[0075] The adaptive radar target detection method provided by the present application aims to solve the problem that the radar target echo signal is strongly submerged by sea clutter in high sea states. The present application adaptively decomposes the radar echo signal through VMD, decomposes the complex signal into multiple intrinsic mode components, and effectively separates the target signal from the sea clutter. Subsequently, based on the frequency domain interval selection strategy based on Gaussian fitting, the target mode components are accurately locked, realizing the enhancement of the target signal and clutter suppression. On the reconstructed signal, key features are extracted in multiple dimensions, including average amplitude, peak height, and Doppler peak height, and these features can effectively distinguish the target from the background clutter. Finally, a constant false alarm rate detection algorithm is used to detect targets for the extracted key features, significantly improving the accuracy and robustness of target detection.

[0076] The main implementation method of the method of the present application is based on the efficient signal decomposition of VMD: using the variational mode decomposition technique, the radar echo signal is decomposed with high resolution, avoiding the mode mixing phenomenon in traditional methods, and ensuring the effective separation of the target signal and the clutter.

[0077] Gaussian fitting frequency domain interval selection strategy: By performing Gaussian fitting on the clutter spectrum and combining statistical analysis of the spectral centroid and spectral width, the frequency domain interval of the target signal is accurately determined, improving the locking accuracy of the target modal components.

[0078] Windowing processing strategy: The long-time signal is divided into several small windows, and VMD decomposition and reconstruction are performed separately, further enhancing the detection ability of the target signal, especially performing excellently at the target edge units. Brief Description of the Drawings

[0079] To more clearly illustrate the technical solutions of the present invention, the drawings required for description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0080] Figure 1 It is a flowchart of the adaptive radar target detection method;

[0081] Figure 2 It is a spectrogram when reconstructing the signal;

[0082] Figure 3 It is a flowchart of the windowing processing;

[0083] Figure 4 It is a schematic diagram of the electronic device. Detailed Embodiments

[0084] The adaptive radar target detection method provided in this application is for detecting channel buoys in high sea states. By accurately determining the frequency domain interval and screening the target modal components, combined with the Doppler frequency shift characteristics of the buoys, sea clutter is effectively suppressed and the target signal is enhanced. Based on multi-feature extraction, through feature analysis and screening, the key features that significantly affect the target detection performance are determined, optimizing the efficiency and accuracy of the detection algorithm. The method of this application significantly improves the accuracy of target detection and the adaptability of the system in complex sea conditions and strong sea clutter backgrounds, providing a more reliable signal processing and detection method for sea surface target detection.

[0085] The following will detail the specific steps of the adaptive radar target detection method provided in this application. For illustration rather than limitation, specific details such as specific system structures and technologies are proposed to thoroughly understand the embodiments of this application. However, those skilled in the art should clearly understand that this application can also be implemented in other embodiments without these specific details.

[0086] Statements such as "an embodiment" or "some embodiments" described in this application mean that the specific features, structures, or characteristics described in the embodiment are included in one or more embodiments of this application. Thus, statements such as "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments", etc., which appear at different places in this application, do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways.

[0087] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0088] Please refer to Figure 1 The following is a flowchart of an adaptive radar target detection method in a specific embodiment. The method includes:

[0089] S101: Obtain a radar echo signal and perform high-resolution decomposition on the radar echo signal.

[0090] In some embodiments, an electromagnetic wave signal reflected and received from a target object can be obtained through a related radar receiving device. The electromagnetic wave signal may include, but is not limited to, information such as the distance, speed, and angle of the target object.

[0091] Considering that there are various noises and interferences in the electromagnetic wave signal, high-resolution decomposition can be performed on the radar echo signal. Specifically, a high-resolution waveform decomposition method based on the lidar physical model can be used. Other methods such as wavelet transform can also be used to decompose the radar echo signal.

[0092] For example, in the manner of high-resolution waveform decomposition based on the lidar physical model, the light spot is micro-differentiated into adjacent small light spots, and the superposition of two Gaussian functions is used to establish an emission pulse model, and a waveform model of the received echo signal is established. Then, the lidar emission pulse and the echo signal are separated to obtain a narrow emission pulse and an echo signal. By using the deconvolution algorithm to restore the target scattering cross-section function and processing the narrow emission pulse and the echo signal respectively, the sum of the scattering cross-section functions of two targets can be decomposed and cubic spline decomposition can be used. After the target scattering cross-section function is decomposed in sequence, the sub-echo signal is obtained by convolving the target scattering cross-section function with the emission pulse, realizing high-resolution decomposition.

[0093] It can be seen that the high-resolution decomposition in this embodiment can separate components with different frequencies in the radar echo signal, and analyze the clutter and target characteristics in the signal. By observing the signal at different scales, the change characteristics of the signal in different frequency ranges can be captured, so as to extract the target information in the radar echo.

[0094] S102: Perform Gaussian fitting on the clutter spectrum in the radar echo signal, analyze the spectrum centroid and spectrum width characteristics, and combine the Doppler frequency shift characteristics of the radar echo signal to determine the frequency domain interval of the radar echo signal.

[0095] According to the embodiment of the present application, for the signal after high-resolution decomposition, the power spectral density of the clutter can be calculated by a spectrum estimation method. Then, use a Gaussian function to fit the power spectrum, and determine the mean, standard deviation, etc. of the Gaussian function by minimizing the fitting error.

[0096] The spectrum centroid of this embodiment can be obtained by the weighted average of the power spectral density and frequency, which reflects the central frequency position of the clutter spectrum. The spectrum width can be determined by calculating the distribution width of the power spectral density on both sides of the spectrum centroid. For example, the full width at half maximum is used to measure the spectrum width. By analyzing the spectrum centroid and spectrum width, the frequency concentration degree and distribution range of the clutter can be understood. In this embodiment, through the analysis of the radar, combined with the spectrum centroid and spectrum width of the clutter, a frequency domain interval including the possible frequency range of the target signal is determined.

[0097] This embodiment uses Gaussian fitting to describe the spectrum characteristics of the clutter, and also obtains the frequency distribution of the clutter through the analysis of the spectrum centroid and spectrum width. Then, combined with the Doppler frequency shift characteristics, the frequency domain interval of the target can be determined, improving the accuracy of target detection.

[0098] S103: Decompose the radar echo signal into multiple intrinsic mode components, and screen out the mode components falling within the target frequency domain interval for reconstruction.

[0099] In this embodiment, the empirical mode decomposition method is used to decompose the radar echo signal. Here, by continuously screening the extreme points in the signal, the upper and lower envelope lines are constructed, and then the intrinsic mode components are obtained.

[0100] Optionally, the empirical mode decomposition method is used to find all the maximum and minimum points of the signal, use the cubic spline function to fit the upper and lower envelope lines respectively, calculate the mean of the upper and lower envelope lines, subtract the mean from the original signal to obtain a new signal, and repeat the above process for the new signal until a certain stop condition is met, to obtain a series of IMFs.

[0101] In this embodiment, the frequency characteristics of each IMF are further compared with the target frequency domain interval determined in step S102, and the IMFs with frequencies falling within this interval are screened out. Then, these screened-out IMFs are superimposed and reconstructed to obtain a signal that only contains the target-related frequency components. According to the frequency domain interval determined in step S102, the modal components falling within the target frequency domain interval are screened out. These components may contain target information. The screened-out modal components are reconstructed to obtain a reconstructed version of the target signal.

[0102] In this way, by screening out and reconstructing the modal components falling within the target frequency domain interval, clutter and other irrelevant frequency components can be removed, the characteristics of the target signal can be highlighted, and the signal-to-noise ratio of the target signal can be improved.

[0103] S104: Extract the key features from the reconstructed radar echo signal in the time domain and frequency domain dimensions, and retain the features with discriminative power for target detection through a feature screening method.

[0104] In this embodiment, in the time domain, features such as the mean, variance, peak value, and zero-crossing rate of the signal can be extracted. The mean reflects the average intensity of the signal, the variance represents the degree of signal fluctuation, the peak value reflects the maximum amplitude of the signal, and the zero-crossing rate describes the number of times the signal crosses the zero level per unit time.

[0105] For the method of frequency domain feature extraction, the reconstructed signal is subjected to Fourier transform. After being transformed to the frequency domain, features such as the peak frequency, bandwidth, and centroid frequency of the spectrum can be extracted. The spectral peak frequency represents the frequency position where the signal energy is concentrated, the bandwidth reflects the distribution range of the signal frequency components, and the centroid frequency is similar to the spectral centroid and measures the central position of the spectrum.

[0106] The extraction of time domain and frequency domain features in this embodiment can reflect the characteristics of the signal and improve the accuracy of target detection. Feature screening can reduce redundant information and reduce the computational complexity.

[0107] S105: Divide the long-time signal in the radar echo signal into several small windows, and perform VMD decomposition and reconstruction on each window respectively.

[0108] The signal window segmentation in this embodiment divides the long-time radar echo signal into multiple small windows according to a certain time length, and each small window contains a certain number of sampling points. The selection of the window length needs to comprehensively consider the change characteristics of the signal and computational resources.

[0109] Perform variational mode decomposition (VMD) on the signals within each small window. VMD is a signal decomposition method based on the variational principle. By constructing and solving a variational model, the signal is decomposed into multiple modal components with different center frequencies. After decomposition, each modal component is analyzed, and the modal components related to the target are selected for reconstruction to obtain the reconstructed signals for each small window.

[0110] S106: Feed the reconstructed signals into the CA-CFAR detection module and use the CA-CFAR algorithm to perform target detection on the extracted key features.

[0111] In this embodiment, each reconstructed small window signal is fed into the CA-CFAR detection module. The CA-CFAR detection module calculates the noise power within the reference cells and computes the detection threshold according to the set false alarm probability. The signal amplitude of the cell to be detected is compared with the detection threshold. If the signal amplitude is greater than the threshold, it is determined as a target signal.

[0112] It can be seen that the above method realizes the adaptive processing and target detection of radar echo signals through steps such as high-resolution decomposition, clutter spectrum analysis, empirical mode decomposition, feature extraction and screening, and CA-CFAR detection. It can effectively separate targets and clutter in complex environments and improve the accuracy and robustness of target detection.

[0113] Furthermore, in combination with the execution steps of the above adaptive radar target detection method, as a refinement and extension of the specific implementation manner of the above embodiment, in order to fully illustrate the specific implementation process in this embodiment, the following provides a more specific method.

[0114] The method includes:

[0115] Step 201: Variational mode decomposition and reconstruction.

[0116] In this embodiment, VMD is an adaptive and completely non-recursive signal processing method. By solving the variational optimization problem, the signal is decomposed into several modal components, and each modal component has the minimum bandwidth in the frequency domain. The VMD method can effectively suppress the mode mixing phenomenon in the EMD method and has better robustness.

[0117] Assume the original signal is decomposed into modal components , and each modal component has a center frequency .

[0118] The goal here is to minimize the bandwidth of each modal component while satisfying that the sum of all modal components is equal to the original signal. The objective function and constraints are shown as follows.

[0119]

[0120]

[0121] Among them, represents the time differential operator. In order to introduce the constraint conditions into the optimization problem, a Lagrangian function is constructed.

[0122]

[0123] Among them, is the frequency constraint parameter, is the Lagrange multiplier, represents the inner product operation.

[0124] By alternately updating , and , the optimization problem is solved iteratively.

[0125]

[0126]

[0127]

[0128] Among them, is the gradient descent learning rate. is the Fourier transform of the original signal f(t), which converts the original time-domain signal to the frequency domain and reflects the frequency component distribution of the original signal at different frequencies . is the Fourier transform of the k-th mode component . is the Fourier transform of the Lagrange multiplier . is the Fourier transform of the k-th mode component , which reflects the frequency component distribution of the k-th mode component in the frequency domain.

[0129] When the following conditions are met, the iterative process stops.

[0130]

[0131] Among them, is the number of iterations, is the maximum number of iterations, is the preset convergence threshold. After VMD decomposition, mode components are obtained.

[0132] Step 202: Select the frequency domain interval.

[0133] The Doppler shift of the buoy in this embodiment is closely related to its radial velocity. The radial velocity of the buoy causes a frequency shift in the radar echo, which is a direct manifestation of the Doppler effect.

[0134] Under high sea state conditions, the radial velocity of the buoy is affected by the sea waves. In particular, the undulation of the sea waves and the directional change of the wind waves will affect the movement trajectory of the buoy, thereby causing changes in the Doppler shift.

[0135] The change in the radial velocity is not only related to the movement speed of the buoy but also closely related to the movement direction of the buoy relative to the radar. For example, when the buoy is moving towards or away from the radar, its radial velocity will affect the shift of the echo frequency, thereby affecting the radar's recognition of the target.

[0136] By knowing the radial velocity of the buoy, the corresponding Doppler frequency range can be calculated. The relationship between the Doppler shift and the radial velocity is as follows.

[0137]

[0138]

[0139] Where, is the Doppler shift. is the radial velocity of the buoy. A positive value indicates that the target is moving away from the radar, and a negative value indicates that the target is approaching the radar. is the wavelength of the signal. is the radar transmit frequency.

[0140] In a high sea state environment, the radar echo signal is usually affected by complex sea clutter, resulting in spectral broadening and irregular fluctuations of the echo signal.

[0141] The spectral centroid is the weighted average frequency of the spectrum and can describe the "center" position of the signal energy distribution. It represents the spectral energy center of the signal and can reflect the spectral distribution characteristics of the signal. In the clutter signal, the spectral centroid is usually located in the lower frequency band because the clutter signal in the ocean environment often has lower frequency components. The spectral width describes the extent of the signal spectrum expansion, that is, the width of the signal energy distribution. For the clutter signal, it usually exhibits a wider spectral width, meaning that its energy is distributed over a wider frequency range. This characteristic enables the clutter signal and the target signal to be distinguished in the frequency domain by the spectral width.

[0142] The Gaussian distribution is widely used to describe the amplitude distribution of random phenomena or signals, and its mathematical expression is as follows.

[0143]

[0144] Where, represents the maximum value of the curve, is the center position of the curve, is the standard deviation. x is the independent variable used to estimate the center frequency of the radar echo signal and the frequency variable in the context of the 3dB bandwidth BW. By fitting the radar echo signal data at different frequency values to the Gaussian model, the relevant parameters can be estimated.

[0145] To estimate the center frequency of the radar echo signal and the 3dB bandwidth , it is necessary to fit the Gaussian model. This fitting process is completed by minimizing the error between the data and the Gaussian model.

[0146] The center frequency is usually used to describe the main frequency position of the signal, that is, the main energy of the signal is concentrated near this frequency. The parameter in the Gaussian model can represent the center frequency .

[0147]

[0148] The 3dB bandwidth represents the frequency range when the spectral energy drops to half of the peak value. This bandwidth reflects the spectral width of the signal and is usually used to describe the degree of signal expansion. Through the mathematical properties of the Gaussian distribution, the 3dB bandwidth can be calculated by the standard deviation of the Gaussian curve.

[0149]

[0150] Among them, 2.355 is the conversion coefficient from the standard deviation to the 3dB bandwidth, which is derived from the logarithmic properties of the Gaussian distribution.

[0151] In high sea states, there is usually a certain correlation between the spectral centroid and spectral width of clutter signals. Due to the diversity of ocean clutter, its spectrum often shows large fluctuations, manifested as the continuous change of the position of the spectral centroid and a large spectral width, and the spectrum shows a wide expansion. By analyzing the spectral centroid and spectral width, clutter and target signals can be effectively distinguished. Clutter signals usually show a lower spectral centroid and a wider spectral width, while target signals may show a narrower spectral width and a more concentrated spectral centroid. This analysis can provide a strong basis for subsequent target signal separation and enhancement.

[0152] In a typical high sea state environment, the spectral characteristics of clutter signals are significantly affected by the wind and wave directionality and the motion of the buoy. To effectively separate target signals from background clutter in radar echoes, a reasonable selection of the frequency domain interval becomes a key step in signal processing. The selection strategy of the frequency domain interval is based on the center frequency f0 and bandwidth BW of the clutter spectrum, and takes into account the influence of the Doppler shift fd of the buoy on the spectral distribution.

[0153] In this application, based on the experimental environment being in typical high-sea state conditions, since the dominant propagation direction of the wind and waves is relatively fixed with respect to the position of the receiving device, the energy of the clutter spectrum is mainly concentrated in the positive frequency region, while the negative frequency components are relatively weak.

[0154] Therefore, the frequency domain interval of the target signal can be determined through the following strategy.

[0155]

[0156] In the frequency domain interval selection strategy, the spectrum of the target signal is concentrated in a specific frequency interval [f low , f high , while the clutter signal is usually distributed outside this region. Through the central frequency of each modal component, it can be determined whether it belongs to the target signal. By superimposing the modal components u k (t) that belong to the target signal, the reconstruction of the target signal can be achieved.

[0157]

[0158] Among them, S target (t) is the reconstructed target signal, and f k is the central frequency of the k-th modal signal.

[0159] This frequency domain interval selection strategy can effectively suppress the clutter signal and enhance the relative intensity of the target signal. Under high-sea conditions, the spectral characteristics of the wind and waves cause the frequency of the clutter to be concentrated within a certain range. By accurately selecting the frequency domain interval of the target signal, the interference of the clutter can be avoided, while fully retaining the spectral information of the target signal. This is crucial for improving the accuracy and robustness of target detection. Especially under the influence of complex sea conditions and buoy movement, it can significantly enhance the radar system's ability to identify targets.

[0160] According to this frequency domain interval selection strategy, determine whether each modal signal is a target signal. Reconstruct the target modal signals to obtain the reconstructed signal and draw the time-frequency spectrum diagram, as Figure 2 shown.

[0161] Step 203: Feature extraction and screening.

[0162] The features involved in this embodiment include average amplitude (AA), peak height (PH), and Doppler peak height (DPH). These features can effectively distinguish the target from the background clutter. The relative average amplitude measures the average amplitude difference between the unit to be detected and the reference unit, and improves the detection accuracy by enhancing the contrast between the target and the background.

[0163] The relative peak height evaluates the relative significance of significant peaks in the signal for identifying abnormal peak features.

[0164]

[0165]

[0166] Wherein, represents the number of units in the set and defines the pulse range participating in the ratio calculation.

[0167] DPH is used to quantify the significance of peaks in the Doppler spectrum. Specifically, the calculation process of DPH is as follows.

[0168]

[0169] Wherein, represents the number of channels, is the Doppler amplitude spectrum in the frequency domain.

[0170]

[0171] The Doppler frequency peak is determined by the following formula.

[0172]

[0173] Wherein, represents the Doppler channel range participating in the ratio operation.

[0174] The method of this embodiment quantitatively measures the discrimination ability between the target and the background clutter by calculating the relative feature value between the target range cell and the clutter range cell. The relative feature value RX is defined as follows.

[0175]

[0176] Wherein, represents the feature to be evaluated (such as AA, PH, DPH), and represent the corresponding feature values of the target cell and the clutter cell respectively. is the relative feature value, indicating the degree of feature contrast between the target and the clutter.

[0177] It should be noted that the relative average amplitude (RAA) is applicable to target detection in complex backgrounds under high sea conditions by comparing the average amplitude differences between target cells and clutter cells. By enhancing the contrast between the target and the background, RAA can effectively improve the detection accuracy. The relative peak height (RPH) reflects the significance of the target signal by comparing the peak characteristics of target cells and clutter cells, thus distinguishing the target from the clutter. For frequency domain features, the relative Doppler peak height depicts the significant characteristics of the target in the Doppler frequency domain and can reflect the energy proportion and variation of the Doppler frequency peak. The relative Doppler vector entropy is used to evaluate the complexity of the signal spectrum and is suitable for analyzing the spectral chaos degree of the signal.

[0178] To further quantify the effect of the VMD method on enhancing feature performance, the change amount of the relative feature value is defined , and the relative feature value difference between the original signal and the VMD reconstructed signal is expressed on a logarithmic scale.

[0179]

[0180] Among them, represents the relative feature value of the VMD reconstructed signal, represents the relative feature value of the original signal. Through the above formula, the improvement of the relative feature value can be quantified as a dB value, intuitively reflecting the improvement effect of the VMD method on different features. If the of a certain feature is significantly greater than zero, it indicates that the VMD method has a greater improvement in the target and clutter discrimination ability of this feature; if is close to zero or negative, it means that this feature has a weak response to the VMD method or no improvement.

[0181] Through the calculation of the relative feature value and the change amount of the relative feature value, this embodiment can clearly quantify the improvement effect of the VMD method on the target and clutter discrimination ability. This feature-based evaluation method provides data support for the subsequent optimization of target detection and classification algorithms.

[0182] Step 204: Windowing processing.

[0183] Under high sea conditions, the buoy target signal is often affected by complex environmental interference, which makes it difficult to effectively separate the target characteristics through global signal processing. To improve the accuracy and robustness of target detection, a windowing processing strategy is adopted. By segmenting and overlappingly splicing the signal, the local characteristics of the signal can be captured more accurately, while reducing the influence of clutter.

[0184] As Figure 3 shown, the windowing processing strategy is to divide the long-time signal into several small windows, and independently complete the VMD processing and reconstruction within each window. The long-time signal S(t) is divided into several windows, and the length of each window is Lwin and slide with a certain step length L step The window length L win The selection directly affects the frequency resolution. Assuming that the overlap ratio between windows is m, the expressions for the sliding step length and the length of the overlapping region are as follows.

[0185]

[0186]

[0187] For example, when m is 50%, it means that half of the length between windows overlaps. The smaller the sliding step length, the higher the overlap ratio, and the better the continuity of signal processing, but the computational complexity also increases accordingly.

[0188] The signals within each window are separately subjected to frequency-domain analysis, VMD decomposition, and reconstruction to ensure the accuracy of signal feature extraction. After processing, the results of all windows are stitched together into a complete signal, that is, within the overlapping region, the overlapping part signals of the previous window and the next window are added and averaged.

[0189]

[0190] Among them, S prev (t) is the overlapping part signal of the previous window, and S next (t) is the overlapping part signal of the next window.

[0191] Step 205: Signal detection and processing.

[0192] The sliding window of CA-CFAR consists of three parts, namely the front window, the rear window, and the detection unit. The front and rear windows are respectively used to estimate the clutter power, and the detection unit is the current target position to be detected.

[0193] To accurately estimate the clutter power, the CA-CFAR algorithm introduces two important parameters, the number of training cells and the number of guard cells. The number of training cells determines the number of cells used to estimate the clutter power, and the number of guard cells is used to prevent the target signal near the detection unit from affecting the clutter estimation. The number of guard cells does not directly participate in the calculation of the clutter power, but its position and number are crucial for ensuring the accuracy of the clutter estimation. The clutter power level is estimated through the reference cells of the front and rear windows .

[0194]

[0195] Among them, is the measurement value of the th cell in the front window, and is the Measurement value of one unit.

[0196] Threshold factor Is an important parameter for adjusting the detection threshold and is determined by the set false alarm probability To decide.

[0197]

[0198] Detection threshold Is calculated through the threshold factor and the clutter estimate To calculate.

[0199]

[0200] Finally, the signal strength of the detection unit Needs to be compared with the detection threshold For detection decision according to the following formula. If this formula holds, it is determined that the target is detected.

[0201]

[0202] The target detection probability is an important indicator to measure the performance of the detection algorithm and is used to evaluate the success rate of correctly detecting the target. In this embodiment, through segmented processing, the detection performance of the target unit in different pulse segments is quantitatively analyzed, so as to calculate the overall detection probability.

[0203] Assume that the signal data length of each range unit is N, which is divided into multiple pulse segments, and the length of each segment is L win , and the total number of segments M can be expressed by the following formula.

[0204]

[0205] For each segment of data of each range unit, target detection is performed through CA-CFAR, and based on the comparison between the in-segment eigenvalue X and the detection threshold T, the detection result is generated.

[0206]

[0207] For a single target range unit i , its detection probability P di Is defined by the following formula.

[0208]

[0209] Where, n success,i Represents the number of times the target is successfully detected in the M segments for this target unit. The overall detection probability P d Is the average value of the detection probabilities of all target units.

[0210]

[0211] Among them, K is the number of target units. Through the above formula calculation, the performance of each target unit in multi-segment detection can be quantified. The calculation of the overall detection probability can comprehensively evaluate the detection performance of the algorithm on all targets, providing a reference basis for the optimization and improvement of the detection algorithm.

[0212] Combined with the above method and the average detection probability, the AA feature detection performs better than coherent integration detection in both the original signal and the VMD-reconstructed signal. The average detection probability of AA feature detection in the original signal is 0.6805, and the average detection probability of AA feature detection in the VMD-reconstructed signal is 0.7438. The VMD processing improves the detection probability of AA feature detection by 9.3%. The AA feature detection performs better than coherent integration detection in both the original signal and the VMD-reconstructed signal. The average detection probability of coherent integration detection in the original signal is 0.3566, and the average detection probability of coherent integration detection in the VMD-reconstructed signal is 0.5915. VMD can not only optimize the time-domain characteristics and further improve the performance of AA feature detection, but also significantly enhance the frequency-domain characteristics, thus particularly significantly improving the performance of coherent integration detection.

[0213] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution is prior or subsequent. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0214] The following are embodiments of the adaptive radar target detection system provided by the embodiments of the present disclosure. This system and the adaptive radar target detection method of the above embodiments belong to the same inventive concept. For the details not described in detail in the embodiments of the adaptive radar target detection system, reference can be made to the embodiments of the above adaptive radar target detection method.

[0215] The system includes: a signal decomposition module, configured to obtain a radar echo signal and perform high-resolution decomposition on the radar echo signal.

[0216] A frequency-domain interval selection module, configured to perform Gaussian fitting on the clutter spectrum in the radar echo signal, analyze the spectrum centroid and spectrum width characteristics, and determine the frequency-domain interval of the radar echo signal in combination with the Doppler frequency shift characteristics of the radar echo signal.

[0217] A decomposition and reconstruction module, configured to decompose the radar echo signal into multiple intrinsic mode components, and screen out the mode components falling within the target frequency-domain interval for reconstruction.

[0218] The feature extraction and screening module is used to extract key features from the reconstructed radar echo signal in the time domain and frequency domain dimensions, and through feature screening methods, retain the features that have discriminative power for target detection.

[0219] The windowing processing module is used to divide the long-time signal in the radar echo signal into several small windows, and perform VMD decomposition and reconstruction separately.

[0220] The CA-CFAR detection module is used to send the reconstructed signal into the CA-CFAR detection module, and use the CA-CFAR algorithm to perform target detection on the extracted key features.

[0221] The adaptive radar target detection system involved in this application can receive radar echo signals under high sea conditions, and through the frequency domain interval selection module, accurately determine the frequency domain interval of the target signal based on the spectral centroid and spectral width characteristics of sea clutter, combined with the Doppler frequency shift characteristics of the target.

[0222] This embodiment also uses variational mode decomposition technology to decompose the complex echo signal into multiple intrinsic mode components, and screens out the mode components that fall within the target frequency domain interval for reconstruction, thereby effectively enhancing the target signal and suppressing sea clutter. Subsequently, the feature extraction and screening module extracts key features from the reconstructed signal from multiple dimensions such as the time domain and frequency domain, such as average amplitude, peak height, and Doppler peak height, and through feature screening methods, retains the features that are most discriminative for target detection. To further improve the detection performance, the system adopts a windowing processing module to divide the long-time signal into several small windows, and perform VMD decomposition and reconstruction separately to improve the feature performance of the edge target units.

[0223] Finally, the signal optimized by features is sent into the CA-CFAR detection module, and the constant false alarm rate is used to perform target detection on the extracted key features to ensure high-accuracy and robust target recognition under complex sea conditions.

[0224] Through the collaborative work of the above modules, the entire system can effectively distinguish and detect radar targets under high sea conditions, significantly improving the performance and reliability of target detection. The four groups of data used in the experiment are from the experimental data set collected at the First Bathing Beach in Yantai, Shandong Province, China. The data was collected by two X-band solid-state power amplifier test radars of the SPPR50P model. The data covers two sea condition levels of 4 and 5, and the wave height ranges from 1.8 meters to 2.7 meters, fully reflecting the dynamic change characteristics from medium sea conditions to high sea conditions. These data provide important experimental support for the research of target detection algorithms under high sea conditions, and also help to analyze the law of clutter characteristics changing with sea conditions.

[0225] Such as Figure 4As shown, the present application also provides an electronic device, including a display module 103, a memory 102, a processor 101, and a computer program stored on the memory and executable on the processor 101. When the processor 101 executes the program, the steps of the adaptive radar target detection method are implemented.

[0226] In the embodiments of the present invention, the electronic device includes, but is not limited to, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as, a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components, their connections and relationships, and their functions shown in this embodiment are only examples and are not intended to limit the implementation of the embodiments of the present application described and / or claimed in this embodiment.

[0227] In the embodiments of the present application, the processor 101 may be implemented by using at least one of an application specific integrated circuit, a programmable logic device, a processor, a controller, a microcontroller, a microprocessor, and an electronic unit designed to execute the functions described herein. In some cases, such an implementation may be implemented in the controller. For a software implementation, an implementation of a process or function may be implemented with a separate software module that permits execution of at least one function or operation. The software code may be implemented by a software application (or program) written in any appropriate programming language. The software code may be stored in the memory and executed by the controller.

[0228] The display module 103 is used to display information input by the user or information provided to the user. The display module 103 may include a display panel, and the display panel may be configured in the form of a liquid crystal display, an organic light emitting diode, or the like.

[0229] The memory 102 may be used to store software programs and various data. The memory 102 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other volatile solid state storage devices.

[0230] The present application also provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the adaptive radar target detection method are implemented.

[0231] The storage medium may adopt any combination of one or more readable media. The readable media may be a readable signal medium or a readable storage medium. The readable storage medium may, for example, but not be limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0232] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined in these embodiments can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown in this specification, but rather will be accorded the widest scope consistent with the principles and novel features disclosed in these embodiments.

Claims

1. An adaptive radar target detection method, characterized in that The method includes: Obtain a radar echo signal, and decompose the radar echo signal based on a high-resolution waveform decomposition method of a lidar physical model, or use a wavelet transform method to decompose the radar echo signal; Perform Gaussian fitting on the clutter spectrum in the radar echo signal, analyze the spectral centroid and spectral width characteristics, and combine the Doppler frequency shift characteristics of the radar echo signal to determine the frequency domain interval of the radar echo signal; Decompose the radar echo signal into multiple intrinsic mode components, and screen out the mode components falling within the target frequency domain interval for reconstruction; Extract key features from the reconstructed radar echo signal in the time domain and frequency domain dimensions, and retain the features with discriminative power for target detection through a feature screening method; Segment the long-time signal in the radar echo signal into several small windows, and perform VMD decomposition and reconstruction respectively; Send the reconstructed signal into the CA-CFAR detection module, and use the CA-CFAR algorithm to perform target detection on the extracted key features.

2. The adaptive radar target detection method according to claim 1, characterized in that The step of segmenting the long-time signal in the radar echo signal into several small windows and performing VMD decomposition and reconstruction respectively further includes: Decompose the radar echo signal into modal components ; Each modal component has a center frequency ; Define the objective function for high-resolution decomposition as follows: , and the constraint conditions are: wherein, represents a time differential operator; In the method, a Lagrangian function is also constructed; wherein, is a frequency constraint parameter, is a Lagrange multiplier, represents an inner product operation; By alternately updating , and , iteratively solve the optimization problem; Among them, is the gradient descent learning rate; is the Fourier transform of the original signal f(t); is the k-th modal component of the Fourier transform; is the Lagrange multiplier of the Fourier transform; is the k-th modal component of the Fourier transform; When the following conditions are met, the iterative process stops; Among them, is the number of iterations, is the maximum number of iterations, is a preset convergence threshold. After VMD decomposition, modal components are obtained .

3. The adaptive radar target detection method according to claim 1 or 2, wherein The step of combining the Doppler frequency shift characteristics of the radar echo signal to determine the frequency domain interval of the radar echo signal further includes: Calculate the corresponding Doppler frequency range through the known radial velocity of the buoy. The relationship between the Doppler frequency shift and the radial velocity is as follows: Among them, is the Doppler frequency shift, is the radial velocity of the buoy, is the wavelength of the signal, is the radar transmission frequency.

4. The adaptive radar target detection method according to claim 3, wherein Determine the frequency domain interval of the target signal through the following strategy: In the method, the modal components of the radar echo signal are also u k (t) superposed to realize the reconstruction of the target signal; Among them, S target (t) is the reconstructed target signal, f k is the center frequency of the k-th modal signal, [f low , f high is the frequency interval, BW is the 3dB bandwidth, and fd is the Doppler shift.

5. The adaptive radar target detection method according to claim 1 or 2, wherein The method quantitatively measures the discrimination ability between the target and the background clutter by calculating the relative value of the features of the target distance unit and the clutter distance unit. The relative feature value RX is defined as follows: Among them, and represent the corresponding eigenvalues of the target unit and the clutter unit respectively; Define the variation of the relative feature value , and express the difference in the relative feature values of the original radar echo signal and the VMD reconstructed signal on a logarithmic scale; Among them, represents the relative value of the features of the VMD reconstructed signal, represents the relative value of the features of the original signal.

6. The adaptive radar target detection method according to claim 1 or 2, wherein In the method, the long-time signal S(t) is divided into a number of windows, each window having a length of L win , and with a preset step length L step sliding; Define the overlap ratio between windows as m. Then the expressions for the sliding step size and the length of the overlapping area are: Stitch the results of all windows into a complete signal, that is, within the overlapping area, add the overlapping part signals of the previous window and the next window and take the average value: Among them, S prev (t) is the overlapping part signal of the previous window, and S next (t) is the overlapping part signal of the next window.

7. The adaptive radar target detection method according to claim 1 or 2, wherein The CA-CFAR algorithm includes: the number of training cells and the number of guard cells ; Number of training units Number of units for estimating clutter power, number of guard units To prevent the influence of target signals near the detection unit on clutter estimation; Estimate the clutter power level through reference cells in the front and rear windows , the clutter power level is calculated as follows: Among them, is the measurement value of the th unit in the front window, is the measurement value of the th unit in the rear window; Threshold factor An important parameter used to adjust the detection threshold, determined by the set false alarm probability ; Detection threshold Calculated through the threshold factor and the clutter estimate ; Finally, the radar echo signal strength is compared with the detection threshold for detection decision according to the following formula; If the radar echo signal strength is greater than the detection threshold, it is determined that a target has been detected.

8. An adaptive radar target detection system, characterized in that, The system is used to implement the adaptive radar target detection method according to any one of claims 1 to 7; The system includes: A signal decomposition module for obtaining a radar echo signal and performing high-resolution decomposition on the radar echo signal; A frequency domain interval selection module for performing Gaussian fitting on the clutter spectrum in the radar echo signal, analyzing the spectral centroid and spectral width characteristics, and combining the Doppler frequency shift characteristics of the radar echo signal to determine the frequency domain interval of the radar echo signal; A decomposition and reconstruction module for decomposing the radar echo signal into multiple intrinsic mode components and screening out the mode components falling within the target frequency domain interval for reconstruction; A feature extraction and screening module, which is used to extract key features from the reconstructed radar echo signals in the time domain and frequency domain dimensions, and retain the features with discrimination for target detection through feature screening methods; A windowing processing module, which is used to divide the long-time signals in the radar echo signals into several small windows, and perform VMD decomposition and reconstruction respectively; A CA-CFAR detection module, which is used to send the reconstructed signals into the CA-CFAR detection module, and use the CA-CFAR algorithm to perform target detection on the extracted key features.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the adaptive radar target detection method according to any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the adaptive radar target detection method according to any one of claims 1 to 7.

Images