Marine mixture target identification method and system based on modal decomposition and reconstruction

Abstract

The invention belongs to the technical field of radar signal processing and target identification, and particularly relates to a maritime hybrid target identification method and system based on modal decomposition and reconstruction, and the method comprises the steps: carrying out the variational modal decomposition of a radar echo signal of a maritime hybrid target, and decomposing an original signal into a plurality of intrinsic modal signals; estimating a background noise energy reference and removing noise modals based on the decomposed modal signals, clustering the remaining modals according to the center frequency, and combining the modals with the frequency within the center frequency range into an independent single-target signal which is of the same target and is reconstructed into an independent single-target signal; respectively carrying out time-frequency analysis on each reconstructed single target signal to obtain a corresponding time-frequency diagram, and extracting time-frequency domain features from the time-frequency diagram; and inputting the extracted features into a trained support vector machine classifier to realize automatic identification of the ship target and the floating target. A ship target and a floating target can be distinguished more accurately in a mixture scene, and the stability of tracking identification is improved.

Description

Technical Field

[0001] The present application belongs to the technical field of radar signal processing and target recognition, and specifically relates to a method and system for identifying marine hybrid targets based on modal decomposition and reconstruction. Background Art

[0002] A mixed target at sea refers to a ship and a floating target within the same radar range cell, forming a "mixed" echo. Compared to a single target signal, a mixed echo exhibits greater structural complexity. In the frequency domain, its energy distribution often exhibits a multi-peak structure; in time-frequency plots, this manifests as multiple prominent micro-Doppler ridges, further complicating identification.

[0003] When facing the problem of mixed objects, the existing floating target recognition method mainly focuses on the Doppler channel with the strongest energy in the time-frequency diagram. If the ship echo energy is higher than that of the floating target, the system will often identify the mixed object as a ship. If the floating target echo energy is dominant, it may cause the system to mistakenly identify it as a floating target, miss the continuous tracking of the ship, and even cause the "loss of tracking" problem. If the energy of the two is close, the system will easily switch frequently between the two types of targets, seriously affecting the stability of tracking and recognition. Summary of the Invention

[0004] In order to achieve accurate separation and classification identification of each component target in a mixture and improve the integrity of system perception, the present invention provides a method and system for marine mixture target identification based on modal decomposition and reconstruction.

[0005] In a first aspect, the technical solution of the present invention provides a method for identifying a mixed target at sea based on modal decomposition and reconstruction, comprising the following steps: S1. Perform variational mode decomposition on the radar echo signal of the marine mixed target to decompose the original signal into multiple eigenmode signals; S2. Perform modal screening on the decomposed modal signal, including: Calculate the modal energy peak value of each modal signal obtained by decomposition; Calculate the background noise energy baseline of the original signal; Based on the background noise energy benchmark, the energy threshold is dynamically set in combination with the preset threshold factor, and the modes with modal energy peak values ​​less than the energy threshold are judged as noise modes and eliminated; S3. Cluster the remaining modes according to the center frequency, merge the modes with frequencies within the center frequency range as the same target, and reconstruct them into independent single target signals; S4. Perform time-frequency analysis on each reconstructed single target signal to obtain the corresponding time-frequency graph, and extract time-frequency domain features from the time-frequency graph, including instantaneous micro-Doppler frequency total variation features and main Doppler channel level entropy features; S5. Input the extracted features into the trained support vector machine classifier to realize automatic recognition of ship targets and floating targets.

[0006] By using variational mode decomposition (VMD) to decompose complex signals into multiple intrinsic mode signals, this method can effectively separate the different components in the mixed target and provide a clear signal basis for subsequent feature extraction and classification. The modal screening and clustering reconstruction steps can eliminate noise modes and merge modes belonging to the same target, effectively improving the signal integrity and signal-to-noise ratio, and enhancing the accuracy of target recognition. The instantaneous micro-Doppler frequency total variation features and main Doppler channel level entropy features extracted from the time-frequency domain can accurately reflect the target's motion characteristics and signal complexity, providing high-quality feature input for the classifier. Inputting the extracted features into the trained support vector machine classifier can realize the automatic recognition of ship targets and floating targets, improving recognition efficiency and accuracy, and reducing human intervention.

[0007] As a further limitation of the technical solution of the present invention, in S2, the step of calculating the modal energy peak of each modal signal includes: The modal decomposition of the k modal signal Perform magnitude square operation; Find the maximum value of the square of the amplitude in all time domains, which is the k The modal energy peak of a modal signal; the formula is as follows:

[0008] Where, For the k The modal energy peak of the modal signal, t is the time.

[0009] By performing amplitude squaring and maximum value extraction, the modal energy peak of each modal signal can be accurately calculated, providing a reliable basis for subsequent noise mode removal. The calculation process is simple and easy to implement, allowing for rapid evaluation of the energy characteristics of modal signals without increasing computational complexity.

[0010] As a further limitation of the technical solution of the present invention, in S2, the step of calculating the background noise energy baseline of the original signal includes: Extract the signal segment corresponding to the set frequency in the original signal spectrum as the noise segment , calculate the average energy of the noise segment as the background noise energy benchmark; the set frequency is the set percentage frequency of the highest frequency in the original signal spectrum;

[0011] Where, is the background noise energy benchmark, M is the number of sampling points in the noise segment, and t is the time.

[0012] By extracting the signal segment corresponding to the set frequency in the original signal spectrum as the noise segment and calculating its average energy as the background noise energy benchmark, this method of setting the frequency can more reasonably select the noise segment, make the calculated background noise energy benchmark more in line with the actual situation, provide an accurate basis for dynamically setting the energy threshold, and enhance the accuracy of modal screening.

[0013] As a further limitation of the technical solution of the present invention, in S2, the modal signal satisfies the following conditions:

[0014] It is determined to be a noise mode and is removed, where is the modal energy peak of the i-th modal signal, is the threshold factor.

[0015] Specific conditions for determining noise modes are given. Based on the background noise energy baseline and combined with a preset threshold factor, an energy threshold is dynamically set. Modes with modal energy peaks below the energy threshold are identified as noise modes and eliminated. This dynamic threshold setting approach adapts to noise conditions in different environments and signal conditions, more accurately eliminating noise modes, reducing the impact of noise on subsequent target recognition, and improving recognition accuracy.

[0016] As a further limitation of the technical solution of the present invention, S3 specifically includes: Calculate the center frequency of each residual modal component; A center frequency deviation Δf is preset; all remaining modes are clustered according to the center frequency, and the modal components with a center frequency difference less than Δf are classified into the same cluster; For each cluster, all modal components belonging to the cluster are weighted summed to obtain the reconstructed single target signal; The reconstructed signal obtained from each cluster is subjected to inverse Fourier transform to obtain a single target signal in the time domain.

[0017] As a further limitation of the technical solution of the present invention, in S4, the method for calculating the instantaneous micro-Doppler frequency total variation characteristic includes: Extract the frequency position sequence of the main frequency ridge in the time-frequency diagram; The instantaneous micro-Doppler frequency total variation characteristics are obtained by performing the first-order difference of the frequency position sequence and calculating the sum of the absolute values.

[0018] This feature is derived by extracting the frequency position sequence of the dominant frequency ridges in the time-frequency graph and performing a first-order difference to sum their absolute values. This feature effectively reflects the instantaneous variation of target signals in the time-frequency domain, helping to distinguish between ships and floating targets. It provides the classifier with more discriminative feature information, improving recognition accuracy.

[0019] As a further limitation of the technical solution of the present invention, in S4, the method for calculating the main Doppler channel level entropy feature includes: Integrate the time-frequency graph along the time direction to obtain the total energy of each Doppler channel; The channel with the largest energy is selected as the main Doppler channel, and its energy sequence in the time direction is extracted; The rank entropy characteristics of the main Doppler channel are obtained by calculating the rank entropy of the energy sequence.

[0020] First, the total energy of each Doppler channel is obtained by integrating the time-frequency graph along the time direction. The channel with the largest energy is selected as the main Doppler channel. The energy sequence in the time direction is then extracted and the hierarchical entropy is calculated. This feature can characterize the complexity of the main Doppler channel energy sequence, providing the classifier with important characteristics of the target signal energy, further enhancing the classifier's ability to distinguish different targets.

[0021] As a further limitation of the technical solution of the present invention, the step of performing level entropy calculation on the energy sequence to obtain the level entropy feature of the main Doppler channel includes: Preset embedding dimension m and reconstruct the energy sequence into a matrix; Sort the elements in the matrix and get a vector consisting of the number of swaps; Calculate the probabilities of various numbers of exchanges; Based on the calculated probability, the entropy value is calculated by Shannon entropy and ; The entropy value and Normalization is performed to obtain the rank entropy, that is, the rank entropy feature of the main Doppler channel is obtained.

[0022] The matrix is ​​reconstructed by presetting the embedding dimensions, sorting the matrix elements to obtain a swap count vector, calculating the probabilities of various swap counts, and then calculating the entropy value based on Shannon entropy and normalizing it to obtain the rank entropy. This detailed calculation step ensures the accuracy and consistency of the rank entropy feature calculation, providing a guarantee for accurately extracting the characteristics of the target signal and helping to improve target recognition performance.

[0023] As a further limitation of the technical solution of the present invention, the energy sequence

[0024] matrix

[0025] A vector of swap counts

[0026] Probability of various numbers of exchanges

[0027] Entropy

[0028] Level Entropy

[0029] In the formula, s represents the energy sequence, m represents the embedding dimension, and N is the length of the energy sequence. Indicates when exist is the number of occurrences of the same value, c represents the number of exchanges, and are the entropy values ​​when embedding dimensions m and m+1 respectively.

[0030] As a further limitation of the technical solution of the present invention, in S5, the support vector machine classifier is a polynomial kernel function support vector machine classifier, which projects the input features into a high-dimensional space through nonlinear mapping, and characterizes the high-order correlation between the features for automatic identification of ship targets and floating targets.

[0031] In a second aspect, the technical solution of the present invention further provides a marine hybrid target recognition system based on modal decomposition and reconstruction, comprising: The modal decomposition module is used to perform variational modal decomposition on the radar echo signal of the marine mixed target, decomposing the original signal into multiple eigenmode signals; The modal screening module is used to perform modal screening on the decomposed modal signals, specifically to calculate the modal energy peak of each modal signal; calculate the background noise energy baseline of the original signal; based on the background noise energy baseline, dynamically set the energy threshold in combination with a preset threshold factor, and determine the mode with a modal energy peak value less than the energy threshold as a noise mode and eliminate it; The clustering and reconstruction module is used to cluster the remaining modes according to the center frequency, merge the modes with frequencies within the center frequency range into the same target, and reconstruct them into independent single target signals; The time-frequency analysis module is used to perform time-frequency analysis on each reconstructed single target signal to obtain the corresponding time-frequency graph and extract time-frequency domain features from the time-frequency graph, including the instantaneous micro-Doppler frequency total variation feature and the main Doppler channel level entropy feature; The classifier module is used to input the extracted features into the trained support vector machine classifier to realize automatic recognition of ship targets and floating targets.

[0032] As a further limitation of the technical solution of the present invention, the modality screening module includes: The modal energy peak unit is used to perform an amplitude square operation on the kth modal signal obtained by modal decomposition; it is used to find the maximum value of the squared amplitude in all time domains as the modal energy peak of the kth modal signal.

[0033] The background noise energy benchmark calculation unit is used to extract the signal segment corresponding to the set frequency in the original signal spectrum as the noise segment, where the set frequency is the set percentage frequency of the highest frequency in the original signal spectrum; and calculate the average energy of the noise segment as the background noise energy benchmark.

[0034] The screening and filtering unit is used to dynamically set the energy threshold based on the relationship between the energy peak of the modal signal and the background noise energy benchmark, combined with a preset threshold factor, and judge the mode with a modal energy peak less than the energy threshold as a noise mode and eliminate it.

[0035] The clustering reconstruction module is specifically used to calculate the center frequency of each residual modal component; preset a center frequency deviation Δf; cluster all residual modes according to the center frequency, and classify the modal components with a center frequency difference less than Δf into the same cluster; for each cluster, all modal components belonging to the cluster are weighted summed to obtain a reconstructed single target signal; the reconstructed signal obtained from each cluster is subjected to an inverse Fourier transform to obtain a single target signal in the time domain.

[0036] As a further limitation of the technical solution of the present invention, the time-frequency analysis module includes: The time-frequency analysis unit is used to perform time-frequency analysis on each reconstructed single target signal to obtain the corresponding time-frequency graph; The first calculation unit is used to extract the frequency position sequence of the main frequency ridge in the time-frequency diagram; perform first-order difference on the frequency position sequence and calculate the sum of absolute values ​​to obtain the instantaneous micro-Doppler frequency total variation feature.

[0037] The second calculation unit is used to integrate the time-frequency diagram along the time direction to obtain the total energy of each Doppler channel, select the channel with the largest energy as the main Doppler channel, and extract its energy sequence in the time direction; perform level entropy calculation on the energy sequence to obtain the level entropy feature of the main Doppler channel.

[0038] As a further limitation of the technical solution of the present invention, the second computing unit is specifically used to preset the embedding dimension m, reconstruct the energy sequence into a matrix; sort the elements in the matrix to obtain a vector consisting of the number of exchanges; calculate the probabilities of various exchange numbers; based on the calculated probability, calculate the entropy value through Shannon entropy; normalize the entropy value to obtain the level entropy, that is, obtain the level entropy characteristics of the main Doppler channel.

[0039] As a further limitation of the technical solution of the present invention, the classifier module is a polynomial kernel function support vector machine classifier, comprising: A feature input unit, used for inputting the extracted features into a polynomial kernel function support vector machine classifier; The nonlinear mapping unit is used to project the input features into a high-dimensional space through nonlinear mapping to characterize the high-order correlation between features; Classification unit, used to realize automatic identification of ship targets and floating targets.

[0040] The above technical solution demonstrates the following advantages: By performing variational modal decomposition on radar echo signals, the complex mixed signal is decomposed into multiple intrinsic mode signals, which are then reconstructed into independent single-target signals through screening and clustering. Finally, time-frequency analysis and feature extraction are performed, and automatic recognition is achieved using a support vector machine classifier. This processing method effectively solves the recognition difficulties caused by the complex structure of mixed signals, enabling more accurate distinction between ship targets and floating targets, and improving the stability of tracking and recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for the description. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0042] Figure 1 A flowchart of a method provided in an embodiment of the present invention.

[0043] Figure 2 Spectrum diagram of the mixed target.

[0044] Figure 3 This is the spectrum diagram after modal decomposition.

[0045] Figure 4 This is the spectrum after filtering.

[0046] Figure 5 is the spectrum of the reconstructed signal.

[0047] Figure 6 A system block diagram provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to achieve accurate separation and classification of each component target in a mixture, this application proposes a target recognition method based on multi-target modal decomposition and reconstruction and time-frequency feature differences. The core idea of ​​the method is: by introducing the VMD (Variational Mode Decomposition) algorithm, the radar echo signal is decomposed into multiple modal components to achieve Doppler channel separation of the signal. In order to solve the problems of modal redundancy caused by traditional modal decomposition after the preset number of modes and the fragmented expression of target information in multiple modes, an energy-constrained filtering module and a spectral consistency-based clustering module are designed after VMD processing to achieve automatic screening and effective integration of modal signals. Specifically, the filtering module cuts in from the energy dimension, identifies and eliminates redundant modes with a level close to that of background noise, thereby suppressing the information interference introduced by non-target components and effectively improving the signal-to-noise ratio and feature separability of the reconstructed signal. Furthermore, in order to solve the fragmentation problem caused by the cross-modal distribution of target information, this application designs a clustering module to cluster the decomposition results according to the modal center frequency, and judge the modes with adjacent frequencies as being generated by the same target, thereby completing modal merging and complete reconstruction.

[0049] The signal obtained by preprocessing is called the reconstructed signal, and the reconstructed signal is subjected to short-time Fourier transform to obtain its time-frequency information. Starting from the image level, the total variation of instantaneous micro-Doppler frequency (VF) is used to measure the degree of fluctuation of the "ridge" or "energy trajectory" in the target micro-Doppler image. From the perspective of data information structure, this application introduces the rating entropy of the dominant Doppler channel (REDDC) to measure the degree of chaos of the signal's temporal energy distribution on the dominant frequency channel. Finally, the feature vector is composed and input into the polynomial kernel SVM to achieve target classification.

[0050] To make the objectives, features, and advantages of this application more apparent and understandable, the technical solutions protected by this application will be described clearly and completely below using specific embodiments and accompanying drawings. Obviously, the embodiments described below are only a portion of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by persons of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0051] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as those commonly understood by those skilled in the art to which the present invention pertains. The terms used in this application and in the specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention.

[0052] like Figure 1 As shown, an embodiment of the present invention provides a method for identifying a marine hybrid target based on modal decomposition and reconstruction, comprising the following steps: S1. Perform variational mode decomposition on the radar echo signal of the marine mixed target to decompose the original signal into multiple eigenmode signals; This application uses existing technology to collect radar echo signals from mixed targets at sea. In this step, the radar echo signals are subjected to variational modal decomposition, breaking the original signals into a set of intrinsic mode components with specific frequency characteristics. These specific frequency characteristics refer to the unique frequency distribution characteristics of each intrinsic mode function (IMF), including center frequency, bandwidth, frequency distribution uniformity, and frequency adaptability.

[0053] In order to ensure the effective separation of the components in the signal while retaining the modal redundancy space, the number of modal components K is set to a value greater than the target number, for example, 4.

[0054] Taking the measured data as an example, Figure 2 The spectrum of the mixed target shows two energy peaks. The energy peak at zero frequency and with a wide bandwidth indicates a floating target, while the energy peak at a frequency other than zero frequency and with a narrow bandwidth indicates a ship target. Figure 3 is the spectrum diagram after modal decomposition, from Figure 3 It can be seen that mode 1 corresponds to the ship mode, mode 3 and mode 4 correspond to floating targets, and mode 2 is the background noise mode.

[0055] S2. Modal screening is performed on the decomposed modal signal. In order to ensure that the decomposed modes can accurately correspond to ships and floating targets, a two-stage processing method is designed. First, the modes with energy close to the background noise are filtered out through energy discrimination to eliminate the noise components that do not contain effective target information; then, the modes with similar frequencies are merged according to the center frequency characteristics to ensure that targets with a wide spectrum such as floating targets can be presented in the form of complete modes, thereby improving the reconstruction quality and recognition reliability of the signal. Figure 3 As can be seen, the energy of background noise mode 2 is nearly identical to the energy of the background noise. This characteristic of the peak energy of the spectrum of this mode being approximately identical to the energy of the background noise can be exploited to set up a filtering module. This module can filter out irrelevant modes that represent background noise. This step specifically includes: Calculate the modal energy peak of each decomposed modal signal; calculate the background noise energy baseline of the original signal; based on the background noise energy baseline, dynamically set the energy threshold in combination with a preset threshold factor, and determine the mode with a modal energy peak value less than the energy threshold as a noise mode and eliminate it; Since the modal energy of background noise is close to the energy of background noise, and the center frequency of target signal is different. This application designs filtering and clustering modules based on this. Calculate the energy peak of all modal components, and select the signal segment corresponding to the 20% frequency of high frequency in the original signal spectrum as the noise segment, calculate the average energy of the noise segment as the noise energy benchmark, and set the energy threshold according to the noise energy benchmark. The mode with modal energy peak less than the energy threshold is identified as noise mode, and screened out.

[0056] First, extract the signal segment corresponding to the highest 20% frequency in its spectrum from the original signal and record it as the noise segment Calculate the average energy of the noise segment as the background noise energy benchmark:

[0057] Where M is the number of sampling points in the noise segment and t is the time. k modal signal Calculate its peak energy:

[0058] If the mode meets the following conditions:

[0059] It is determined to be a noise mode and is removed, where is the modal energy peak of the i-th modal signal, is the threshold factor, which is dynamically adjusted according to the relative amplitude of the highest modal peak and the noise floor. Figure 3 The spectrum diagram after modal decomposition and Figure 4 Compare the filtered spectrum graph in . Figure 4 As can be seen from the figure, mode 2 representing the background noise has been successfully filtered out.

[0060] S3. Cluster the remaining modes according to the center frequency, merge the modes with frequencies within the center frequency range as the same target, and reconstruct them into independent single target signals; Modes with similar frequencies are merged based on their center frequency characteristics to ensure that targets with a wide spectrum, such as floating targets, can be presented in a complete modal form, thereby improving the signal reconstruction quality and recognition reliability.

[0061] For the case where the center frequencies of the ship target and the floating target are far apart (if the ship and the floating target are aliased in the distance unit and aliased in the frequency dimension, they do not fall within the scope of this application), and the two modes decomposed by a single target have relatively close center frequencies in the spectrum, this application introduces a clustering module to cluster the modes with similar center frequencies and merge them into a complete mode. Figure 4 The spectrum diagram after filtering is actually the spectrum diagram before modal clustering. Figure 5 is the spectrum diagram of the reconstructed signal. Figure 4 and Figure 5 Comparison shows that this method effectively solves the fragmentation problem of floating target signals being split into multiple modes during the modal decomposition process, thereby improving the integrity of the reconstructed signal.

[0062] In summary, for mixed targets, this application effectively separates the mixed echo signal into multiple independent single target signals through modal decomposition combined with a frequency domain energy filtering module and a clustering module.

[0063] S4. Perform time-frequency analysis on each reconstructed single target signal to obtain the corresponding time-frequency graph, and extract time-frequency domain features from the time-frequency graph, including instantaneous micro-Doppler frequency total variation features and main Doppler channel level entropy features; In order to quantify the structural differences between ship targets and floating targets in the time-frequency domain, this application extracts two types of micro-motion features from the image structure level and the time series data level, respectively, to characterize the motion regularity and frequency stability of the target.

[0064] Step a: Extract instantaneous micro-Doppler frequency total variation features; Ⅰ: First, we need to find the frequency position sequence of the main frequency ridge in the time-frequency graph:

[0065] It is a time-frequency diagram matrix, where the vertical axis is frequency f and the horizontal axis is time t; is the frequency value at time t, and T is the time length.

[0066] II: Take the first-order difference of the frequency position sequence and calculate the sum of the absolute values ​​to obtain the total variation of the main frequency ridge over time:

[0067] is the frequency value at time t-1.

[0068] Step b: extracting the main Doppler channel level entropy features; Ⅰ: First, extract the data of the main Doppler channel and convert the time-frequency matrix Integrate along the time direction to obtain the total energy of each Doppler channel:

[0069] II: Select the channel with the largest energy Take the main Doppler channel as the input and extract the energy sequence of the channel in the time direction:

[0070] Ⅲ: After obtaining the main Doppler channel energy sequence , N is the length of the sequence, and after the preset embedding dimension m, the matrix is ​​reconstructed It can be expressed as:

[0071] IV: For Sort each element in ascending order using bubble sort to obtain a vector consisting of the number of swaps :

[0072] V: Find the probability of various exchanges through normalization ,in Indicates when exist The number of occurrences of the same value in .

[0073]

[0074] Where c represents the number of exchanges, which ranges from 0 to .

[0075] VI: Probability obtained based on the above formula, entropy value It can be obtained by Shannon entropy:

[0076] VII: Set m to m+1 and repeat again to get , will get and Normalized to get the rank entropy:

[0077] S5. Input the extracted features into the trained support vector machine classifier to realize automatic recognition of ship targets and floating targets.

[0078] In target classification tasks, traditional linear SVMs are limited in their ability to distinguish features due to the complex nonlinear relationships between features. To this end, this application introduces a polynomial kernel function that projects the original data into a high-dimensional space through nonlinear mapping, thereby characterizing the high-order correlations between features. The decision boundary constructed in this space can more effectively distinguish different categories, especially in low signal-to-noise ratio environments, showing stronger anti-interference capabilities and significantly improving classification performance.

[0079] It should be noted that the classifier training process involves combining the extracted instantaneous micro-Doppler frequency total variation (VF) and the main Doppler channel rank entropy (REDDC) features to form a feature vector for classification. A polynomial kernel function is selected as the kernel function of the support vector machine (SVM). This kernel function maps the feature vector from the original low-dimensional space to a high-dimensional space to capture the nonlinear relationships between features. The constructed SVM model is trained using labeled sample data containing ship targets and floating targets to determine the optimal model parameters, enabling the model to accurately distinguish between the two types of targets. The feature vector extracted from the reconstructed signal to be identified is input into the trained SVM model. The model makes a classification judgment based on the distribution of the feature vector in the high-dimensional space and outputs the target category corresponding to the feature vector (ship target or floating target), thereby achieving automatic recognition.

[0080] The proposed method was verified using measured data, and a mixed target data set and an independent target data set were selected to verify the performance of the proposed method.

[0081] Several sets of mixed target and single target data from multi-target scenarios were selected to verify the performance of the proposed algorithm in detecting the presence of mixed targets. The specific data description is shown in Table 1, which includes three sets of mixed target data and one set of independent target data.

[0082] Table 1: Dataset description

[0083] To evaluate the algorithm's performance in detecting the presence of mixed objects, we tested mixed objects from Datasets 1, 2, and 3. If the algorithm determined that they were mixed objects, the detection was considered correct. For Dataset 4, we tested individual ship and buoy targets. If the algorithm determined that they were not mixed objects, the detection was considered correct, thus evaluating the algorithm's false alarm rate.

[0084] Table 2: Performance analysis of mixed object presence detection

[0085] Table 2 shows that in Datasets 1 and 2, the mixed targets were located in sea state level 2. While the detection accuracy decreased with decreasing observation time, the average accuracy still reached 94.96%. In Dataset 3, the mixed targets were located in sea state level 4, and the detection accuracy significantly improved compared to sea state level 2. For Dataset 4, which contained only independent ship and floating targets and no mixed targets, the results showed an extremely low false alarm rate of only 0.1% for non-mixed targets, demonstrating its excellent misjudgment suppression effectiveness.

[0086] In the embodiment of the present invention, in S2, the step of calculating the modal energy peak value of each modal signal includes: The kth modal signal obtained by modal decomposition Perform magnitude square operation; Find the maximum value of the square of the amplitude in all time domains, which is the k The modal energy peak of the modal signal.

[0087] In some embodiments, S3 specifically includes: Calculate the center frequency of each residual modal component; A center frequency deviation Δf is preset; all remaining modes are clustered according to the center frequency, and the modal components with a center frequency difference less than Δf are classified into the same cluster; For each cluster, all modal components belonging to the cluster are weighted summed to obtain the reconstructed single target signal; The reconstructed signal obtained from each cluster is subjected to inverse Fourier transform to obtain a single target signal in the time domain.

[0088] like Figure 6 As shown, an embodiment of the present invention further provides a marine hybrid target recognition system based on modal decomposition and reconstruction, comprising: The modal decomposition module is used to perform variational modal decomposition on the radar echo signal of the marine mixed target, decomposing the original signal into multiple eigenmode signals; The modal screening module is used to perform modal screening on the decomposed modal signals, specifically to calculate the modal energy peak of each modal signal; calculate the background noise energy baseline of the original signal; based on the background noise energy baseline, dynamically set the energy threshold in combination with a preset threshold factor, and determine the mode with a modal energy peak value less than the energy threshold as a noise mode and eliminate it; The clustering and reconstruction module is used to cluster the remaining modes according to the center frequency, merge the modes with frequencies within the center frequency range into the same target, and reconstruct them into independent single target signals; The time-frequency analysis module is used to perform time-frequency analysis on each reconstructed single target signal to obtain the corresponding time-frequency graph and extract time-frequency domain features from the time-frequency graph, including the instantaneous micro-Doppler frequency total variation feature and the main Doppler channel level entropy feature; The classifier module is used to input the extracted features into the trained support vector machine classifier to realize automatic recognition of ship targets and floating targets.

[0089] The entire system can automatically complete the entire process from signal input to target recognition, effectively solve the problem of mixed target recognition at sea, and improve the stability and accuracy of target tracking and recognition at sea.

[0090] In some embodiments, the modality screening module includes: The modal energy peak unit is used to perform an amplitude square operation on the kth modal signal obtained by modal decomposition; it is used to find the maximum value of the squared amplitude in all time domains as the modal energy peak of the kth modal signal.

[0091] The background noise energy benchmark calculation unit is used to extract the signal segment corresponding to the set frequency in the original signal spectrum as the noise segment, where the set frequency is the set percentage frequency of the highest frequency in the original signal spectrum; and calculate the average energy of the noise segment as the background noise energy benchmark.

[0092] The screening and filtering unit is used to dynamically set the energy threshold based on the relationship between the energy peak of the modal signal and the background noise energy benchmark, combined with a preset threshold factor, and judge the mode with a modal energy peak less than the energy threshold as a noise mode and eliminate it.

[0093] The clustering reconstruction module is specifically used to calculate the center frequency of each residual modal component; preset a center frequency deviation Δf; cluster all residual modes according to the center frequency, and classify the modal components with a center frequency difference less than Δf into the same cluster; for each cluster, all modal components belonging to the cluster are weighted summed to obtain a reconstructed single target signal; the reconstructed signal obtained from each cluster is subjected to an inverse Fourier transform to obtain a single target signal in the time domain.

[0094] In some embodiments, the time-frequency analysis module includes: The time-frequency analysis unit is used to perform time-frequency analysis on each reconstructed single target signal to obtain the corresponding time-frequency graph; The first calculation unit is used to extract the frequency position sequence of the main frequency ridge in the time-frequency diagram; perform first-order difference on the frequency position sequence and calculate the sum of absolute values ​​to obtain the instantaneous micro-Doppler frequency total variation feature.

[0095] The second calculation unit is used to integrate the time-frequency diagram along the time direction to obtain the total energy of each Doppler channel, select the channel with the largest energy as the main Doppler channel, and extract its energy sequence in the time direction; perform level entropy calculation on the energy sequence to obtain the level entropy feature of the main Doppler channel.

[0096] In some embodiments, the second computing unit is specifically used to preset the embedding dimension m, reconstruct the energy sequence into a matrix; sort the elements in the matrix to obtain a vector consisting of the number of exchanges; calculate the probabilities of various exchange numbers; based on the calculated probabilities, calculate the entropy value through Shannon entropy; normalize the entropy value to obtain the level entropy, that is, obtain the level entropy characteristics of the main Doppler channel.

[0097] In some embodiments, the classifier module is a polynomial kernel support vector machine classifier, comprising: A feature input unit, used for inputting the extracted features into a polynomial kernel function support vector machine classifier; The nonlinear mapping unit is used to project the input features into a high-dimensional space through nonlinear mapping to characterize the high-order correlation between features; Classification unit, used to realize automatic identification of ship targets and floating targets.

[0098] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein, but is intended to be construed in the widest manner consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying marine hybrid targets based on modal decomposition and reconstruction, characterized in that: The following steps are involved: S1. Perform variational mode decomposition on the radar echo signal of the marine mixed target to decompose the original signal into multiple eigenmode signals; S2. Perform modal screening on the decomposed modal signal, including: Calculate the modal energy peak value of each modal signal obtained by decomposition; Calculate the background noise energy baseline of the original signal; Based on the background noise energy benchmark, the energy threshold is dynamically set in combination with the preset threshold factor, and the modes with modal energy peak values ​​less than the energy threshold are judged as noise modes and eliminated; S3. Cluster the remaining modes according to the center frequency, merge the modes with frequencies within the center frequency range as the same target, and reconstruct them into independent single target signals; S4. Perform time-frequency analysis on each reconstructed single target signal to obtain the corresponding time-frequency graph, and extract time-frequency domain features from the time-frequency graph, including instantaneous micro-Doppler frequency total variation features and main Doppler channel level entropy features; S5. Input the extracted features into the trained support vector machine classifier to realize automatic recognition of ship targets and floating targets.

2. The method for identifying marine hybrid targets based on modal decomposition and reconstruction according to claim 1 is characterized in that: In S2, the steps of calculating the modal energy peak of each modal signal include: The modal decomposition of the k modal signal Perform magnitude square operation; Find the maximum value of the square of the amplitude in all time domains, which is the k The modal energy peak of a modal signal; the formula is as follows: Where, is the modal energy peak of the kth modal signal, and t is the time.

3. The method for identifying marine hybrid targets based on modal decomposition and reconstruction according to claim 2 is characterized in that: In S2, the step of calculating the background noise energy baseline of the original signal includes: Extract the signal segment corresponding to the set frequency in the original signal spectrum as the noise segment , calculate the average energy of the noise segment as the background noise energy benchmark; the set frequency is the set percentage frequency of the highest frequency in the original signal spectrum; Where, is the background noise energy benchmark, M is the number of sampling points in the noise segment, and t is the time.

4. The method for identifying marine hybrid targets based on modal decomposition and reconstruction according to claim 3 is characterized in that: In S2, the modal signal satisfies the following conditions: It is determined to be a noise mode and is removed, where is the modal energy peak of the i-th modal signal, is the threshold factor.

5. The method for identifying marine hybrid targets based on modal decomposition and reconstruction according to claim 4 is characterized in that: S3 specifically includes: Calculate the center frequency of each residual modal component; A center frequency deviation Δf is preset; all remaining modes are clustered according to the center frequency, and the modal components with a center frequency difference less than Δf are classified into the same cluster; For each cluster, all modal components belonging to the cluster are weighted summed to obtain the reconstructed single target signal; The reconstructed signal obtained from each cluster is subjected to inverse Fourier transform to obtain a single target signal in the time domain.

6. The method for identifying marine hybrid targets based on modal decomposition and reconstruction according to claim 5 is characterized in that: In S4, the method for calculating the instantaneous micro-Doppler frequency total variation characteristic includes: Extract the frequency position sequence of the main frequency ridge in the time-frequency diagram; The instantaneous micro-Doppler frequency total variation characteristics are obtained by performing the first-order difference of the frequency position sequence and calculating the sum of the absolute values.

7. The method for identifying marine hybrid targets based on modal decomposition and reconstruction according to claim 6 is characterized in that: In S4, the method for calculating the main Doppler channel level entropy feature includes: Integrate the time-frequency graph along the time direction to obtain the total energy of each Doppler channel; The channel with the largest energy is selected as the main Doppler channel, and its energy sequence in the time direction is extracted; The rank entropy characteristics of the main Doppler channel are obtained by calculating the rank entropy of the energy sequence.

8. The method for identifying marine hybrid targets based on modal decomposition and reconstruction according to claim 7 is characterized in that: The steps of calculating the level entropy of the energy sequence to obtain the level entropy characteristics of the main Doppler channel include: Preset embedding dimension m, the energy sequence Reconstruct into a matrix ; Sort the elements in the matrix and get a vector consisting of the number of swaps ; Calculate the probability of various numbers of exchanges ; Based on the calculated probability, the entropy value is calculated by Shannon entropy ; Embed the dimension m Increase to m +1, and calculated in the same way ; The entropy value and Normalization obtains the level entropy, that is, the level entropy characteristics of the main Doppler channel are obtained; level entropy In the formula, s represents the energy sequence, m represents the embedding dimension, and N is the length of the energy sequence. Indicates when exist is the number of occurrences of the same value, c represents the number of exchanges, and are the entropy values ​​when embedding dimensions m and m+1 respectively.

9. The method for identifying marine hybrid targets based on modal decomposition and reconstruction according to claim 8, characterized in that: In S4, the support vector machine classifier is a polynomial kernel function support vector machine classifier, which projects the input features into a high-dimensional space through nonlinear mapping, and characterizes the high-order correlation between features for automatic recognition of ship targets and floating targets.

10. A marine hybrid target recognition system based on modal decomposition and reconstruction, characterized in that: include: The modal decomposition module is used to perform variational modal decomposition on the radar echo signal of the marine mixed target, decomposing the original signal into multiple eigenmode signals; The modal screening module is used to perform modal screening on the decomposed modal signal, specifically to calculate the modal energy peak of each modal signal; and calculate the background noise energy baseline of the original signal; Based on the background noise energy benchmark, the energy threshold is dynamically set in combination with the preset threshold factor, and the modes with modal energy peak values ​​less than the energy threshold are judged as noise modes and eliminated; The clustering and reconstruction module is used to cluster the remaining modes according to the center frequency, merge the modes with frequencies within the center frequency range into the same target, and reconstruct them into independent single target signals; The time-frequency analysis module is used to perform time-frequency analysis on each reconstructed single target signal to obtain the corresponding time-frequency graph and extract time-frequency domain features from the time-frequency graph, including the instantaneous micro-Doppler frequency total variation feature and the main Doppler channel level entropy feature; The classifier module is used to input the extracted features into the trained support vector machine classifier to realize automatic recognition of ship targets and floating targets.

Images