Space-based radar sea surface target identification method

Abstract

The invention provides a space-based radar sea surface target identification method, and the method comprises the steps: obtaining a sea surface high-resolution range profile HRRP matrix of a ship, carrying out the amplitude normalization of the HRRP matrix, and carrying out the superposition and accumulation of the normalized HRRP matrix in the azimuth direction; extracting the structural characteristics of the accumulated HRRP; respectively using short-time Fourier transform and Fourier coefficient fitting to extract micro-motion features from the accumulated HRRP azimuth; the structural features and the micro-motion features are input into a meta classifier, a sea surface target recognition result is obtained, the meta classifier is formed by stacking a first base classifier and a second base classifier, the first base classifier is obtained through training according to structural feature samples, and the second base classifier is obtained through training according to micro-motion feature samples. According to the method, the structural features and the micro-motion features of the HRRP echo matrix are extracted respectively, feature fusion and feature classification are performed by using the stacking classifier, the problem that high-similarity false targets are difficult to distinguish in one-dimensional HRRP recognition is solved, and the recognition accuracy and reliability are improved.

Description

Technical Field

[0001] The present application relates to the technical field of radar target recognition, and in particular to a method for identifying sea surface targets with space-based radar. Background Art

[0002] Radar automatic target recognition (RATR) is extremely valuable in maritime transport management and navigation safety, and has far-reaching significance in both military and civilian fields. Space-based radar has a wide field of view and is completely unconstrained by airspace and territorial waters. It can continuously observe the sea for a long time and monitor space targets of interest all day and all weather. It has the potential to be widely used in military and civilian fields, and is of vital strategic significance for my country to ensure national security and promote economic and social development.

[0003] At present, there has been extensive research on the recognition of surface ship targets. However, the existing recognition based on the high-resolution range profile (HRRP) of the surface ship mainly focuses on the structural characteristics of the ship, but the time domain echo of the radar ship target HRRP also contains rich micro-motion characteristics. From the perspective of space-based radar, the shaking of the ship will change the attitude angle of the ship in the radar, causing the scattering points to move regularly across the range unit. Summary of the invention

[0004] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0005] To this end, the first purpose of this application is to propose a method for identifying sea surface targets using space-based radar.

[0006] The second objective of this application is to provide a space-based radar sea surface target recognition device.

[0007] The third objective of the present application is to provide an electronic device.

[0008] A fourth objective of the present application is to provide a computer-readable storage medium.

[0009] A fifth object of the present application is to provide a computer program product.

[0010] To achieve the above-mentioned purpose, the first embodiment of the present application proposes a space-based radar sea surface target recognition method, including:

[0011] Acquire a high-resolution range image HRRP matrix of the ship's sea surface, perform amplitude normalization processing on the HRRP matrix, and stack and accumulate the normalized HRRP matrix along the azimuth direction;

[0012] Extracting structural features of the accumulated HRRP, wherein the structural features include equivalent scattering center dimension, equivalent size, entropy, standard deviation, bias, irregularity, frequency domain incoherent range image and central moment features;

[0013] Short-time Fourier transform and Fourier coefficient fitting are used to extract micro-motion features from the accumulated HRRP azimuths.

[0014] The structural features and the micro-motion features are input into a meta-classifier to obtain a sea surface target recognition result, wherein the meta-classifier is composed of a first base classifier and a second base classifier stacked together, the first base classifier is trained based on structural feature samples, and the second base classifier is trained based on micro-motion feature samples.

[0015] Optionally, the acquiring of a high-resolution range image HRRP matrix of a ship on the sea surface, performing amplitude normalization processing on the HRRP matrix, and stacking and accumulating the normalized HRRP matrix along the azimuth direction includes:

[0016] Get the HRRP matrix of the ship, recorded as:

[0017]

[0018] Wherein, X is the HRRP matrix, M is the number of rows of the HRRP matrix, indicating the number of time frames; N is the number of columns of the HRRP matrix, indicating the number of distance units; x ij represents a single element in the ith row and jth column of the HRRP matrix, that is, the signal strength of the jth distance unit in the ith frame;

[0019] The HRRP matrix is ​​normalized row by row, where the i-th row HRRP is denoted as X i =[x i1 …x iN ], the elements in the HRRP matrix are normalized according to the following formula, the expression is:

[0020]

[0021] The normalized HRRP matrix is ​​superimposed along the azimuth direction, and the expression is:

[0022]

[0023] Among them, X i ' represents the accumulated HRRP.

[0024] Optionally, the structural features of the extracted and accumulated HRRP include:

[0025] For a HRRP after extraction and accumulation, that is, Xi ′=[x i1 …x iN ], denoted as X i =[x(1)…x(N)], the structural features are calculated by the following formulas, including:

[0026] Calculate the equivalent scattering center dimension N scatter , the mathematical expression is:

[0027]

[0028] Where ε(·) is the unit step function, defined as:

[0029]

[0030] Calculate the equivalent size E size , the mathematical expression is:

[0031] E size =S(N scatter )-S(1)

[0032] Among them, S is the position vector of the distance unit where the echo intensity in the HRRP signal is greater than half of the maximum value, and its calculation expression is:

[0033] S={i|x(i)≥max(X) / 2}

[0034] Calculate the entropy, the mathematical expression is:

[0035]

[0036] Calculate the standard deviation std(dB), the mathematical expression is:

[0037]

[0038] Calculate the deviation (dB), the mathematical expression is:

[0039]

[0040] Calculate the irregularity (dB), the mathematical expression is:

[0041]

[0042] Calculate the frequency domain incoherent range image The mathematical expression is:

[0043]

[0044] Calculate the central moment characteristics and assume the first-order origin moment is: Then the p-order central moment C p The mathematical expression is:

[0045]

[0046] Optionally, the extracting micro-motion features from the accumulated HRRP azimuth using Fourier coefficient fitting includes:

[0047] The change of the distance unit of the strongest scattering point over time is extracted, and the change of the position of the strong scattering point in the distance unit with the ship attitude angle is fitted using the Fourier series, and the coefficient of the Fourier series is used as the micro-motion feature of the ship. The calculation formula of the Fourier series is:

[0048]

[0049] Where, f(t) represents the function of the HRRP signal, with time t as the independent variable, a0 is the constant term in the Fourier series, and a n is the cosine term coefficient of the Fourier series, b n is the sinusoidal term coefficient of the Fourier series, ω is the angular frequency of the fundamental frequency, and n is the harmonic number.

[0050] Optionally, extracting micro-motion features from the accumulated HRRP orientation using short-time Fourier transform includes:

[0051] Select multiple distance units near the strong scattering point, use short-time Fourier transform to monitor the changes of multiple distance units simultaneously, and set the i-th HRRP after accumulation as X i , is the amplitude of the nth distance unit of the i-th HRRP. The signal data on each distance unit is short-time Fourier transformed, and the calculation formula is:

[0052]

[0053] Among them, w(m) is the selected window function;

[0054] The multiple short-time Fourier transform matrices calculated by the selected multiple distance units are added together, and the matrix obtained by the addition is recorded as STFT. The matrix STFT is subjected to singular value decomposition to reduce its characteristic dimension, that is:

[0055]

[0056] Among them, STFT m×n represents the matrix after short-time Fourier transform; U m×k is an m×k matrix containing the left singular vectors of the STFT matrix; Σ k×k is a k×k diagonal matrix containing singular values; It is the transpose of the V matrix, with a size of k×n. The V matrix contains the right singular vectors of the STFT matrix;

[0057] According to the singular values ​​obtained by decomposition, the micro-motion characteristics of the ship are extracted.

[0058] To achieve the above-mentioned purpose, the second embodiment of the present application proposes a space-based radar sea surface target recognition device, comprising:

[0059] A preprocessing module is used to obtain a high-resolution range image HRRP matrix of the ship's sea surface, perform amplitude normalization processing on the HRRP matrix, and stack and accumulate the normalized HRRP matrix along the azimuth direction;

[0060] A structural feature extraction module is used to extract the structural features of the accumulated HRRP, wherein the structural features include equivalent scattering center dimension, equivalent size, entropy, standard deviation, deviation, irregularity, frequency domain incoherent range image and central moment features;

[0061] A micro-motion feature extraction module is used to extract micro-motion features from the accumulated HRRP azimuth using short-time Fourier transform and Fourier coefficient fitting respectively;

[0062] The target recognition module is used to input the structural features and the micro-motion features into a meta-classifier to obtain a sea surface target recognition result, wherein the meta-classifier is composed of a first base classifier and a second base classifier stacked together, the first base classifier is trained based on the structural feature samples, and the second base classifier is trained based on the micro-motion feature samples.

[0063] To achieve the above-mentioned purpose, the third aspect of the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

[0064] The memory stores computer-executable instructions;

[0065] The processor executes the computer-executable instructions stored in the memory to implement the method as described in any one of the first aspects.

[0066] To achieve the above-mentioned purpose, the fourth aspect embodiment of the present application proposes a computer-readable storage medium, in which computer-readable storage medium is stored computer execution instructions, and when the computer execution instructions are executed by a processor, they are used to implement the method as described in any one of the first aspects.

[0067] To achieve the above-mentioned purpose, the fifth aspect of the present application proposes a computer program product, which implements any method in the first aspect when executed by a processor.

[0068] The technical solution provided by the embodiments of the present application brings at least the following beneficial effects:

[0069] By extracting the structural features and micro-motion features of the HRRP echo matrix respectively and using a stacked classifier for feature fusion and feature classification, the problem of high-similarity false targets being difficult to distinguish in one-dimensional HRRP recognition is overcome, and the recognition accuracy and reliability are improved.

[0070] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0072] Figure 1 A schematic diagram of a flow chart of a method for identifying sea surface targets using a space-based radar provided in an embodiment of the present application;

[0073] Figure 2 A schematic diagram of the process of training and stacking a meta-classifier provided in an embodiment of the present application;

[0074] Figure 3 A schematic diagram of the structure of a space-based radar sea surface target recognition device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0075] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0076] Radar automatic target recognition (RATR) is extremely valuable in maritime transport management and navigation safety, and has far-reaching significance in both military and civilian fields. Space-based radar has a wide field of view and is completely unconstrained by airspace and territorial waters. It can continuously observe the sea for a long time and monitor space targets of interest all day and all weather. It has the potential to be widely used in military and civilian fields, and is of vital strategic significance for my country to ensure national security and promote economic and social development.

[0077] At present, there has been extensive research on the recognition of surface ship targets. However, the existing recognition based on the high-resolution range profile (HRRP) of the surface ship mainly focuses on the structural characteristics of the ship, but the time domain echo of the radar ship target HRRP also contains rich micro-motion characteristics. From the perspective of space-based radar, the shaking of the ship will change the attitude angle of the ship in the radar, causing the scattering points to move regularly across the range unit.

[0078] In addition, the existing target recognition technology still has the following two difficulties: on the one hand, the attitude angles observed by space-based radars are significantly different from those of other radars, and existing research has not fully demonstrated the applicability of recognition technology to space-based radars; on the other hand, current research has failed to combine high-range resolution image structure features and micro-motion features.

[0079] However, it is well known that in the feature extraction stage, the more comprehensive and representative the extracted information is, the better the accuracy and reliability of recognition will be. In order to overcome the current low accuracy of sea surface ship target recognition, this application proposes a space-based radar sea surface target recognition method, which combines micro-motion features with structural features, and performs feature fusion and recognition through stacked classifiers.

[0080] Figure 1 The following is a flow chart of a method for identifying sea surface targets using a space-based radar provided in an embodiment of the present application. Figure 1 As shown, the method comprises the following steps:

[0081] Step 101, obtaining the high-resolution range image HRRP matrix of the ship's sea surface, performing amplitude normalization processing on the HRRP matrix, and stacking and accumulating the normalized HRRP matrix along the azimuth direction.

[0082] In the embodiment of the present application, the HRRP (high resolution range profile) matrix of the ship is first obtained to represent the signal reflection intensity at different time frames and different distance units, which is recorded as:

[0083]

[0084] Where X is the HRRP matrix, M is the number of rows in the HRRP matrix, indicating the number of time frames; N is the number of columns in the HRRP matrix, indicating the number of distance units; x ij It represents the single element in the i-th row and j-th column of the HRRP matrix, that is, the signal strength of the j-th distance unit in the i-th frame.

[0085] In order to enhance the comparability and signal-to-noise ratio of the data and extract effective features, the HRRP matrix is ​​first normalized by row, where the i-th row of HRRP is denoted as X i =[xi1 …x iN ], the elements in the HRRP matrix are normalized according to the following formula, the expression is:

[0086]

[0087] It should be noted that in radar signal processing, different targets and acquisition conditions may cause amplitude changes. The main purpose of normalization is to eliminate the differences in the amplitudes of different HRRP signals and make the scales of different samples consistent. This helps to enhance the contrast of features and make subsequent processing and classification more stable.

[0088] Finally, the normalized HRRP matrix is ​​superimposed along the azimuth direction, and the expression is:

[0089]

[0090] Among them, X i ' represents the accumulated HRRP.

[0091] In a possible embodiment, N=10 is set.

[0092] It is understandable that in practical applications, the azimuth of the target may vary slightly, and analyzing each HRRP separately may result in the influence of slight deviations. The superposition operation can average the values ​​in multiple azimuths, thereby better suppressing the noise and retaining the main features of the target.

[0093] Step 102: extracting the structural features of the accumulated HRRP.

[0094] For a HRRP after extraction and accumulation, that is, X i ′=[x i1 …x iN ], denoted as X i =[x(1)…x(N)], the structural features can be extracted by the following formulas, including:

[0095] 1. Calculate the equivalent scattering center dimension N scatter ,This feature is to quantify the number of distance cells in HRRP that exceed a certain threshold. It can remove the influence of some clutter without eliminating most of the signal features. Its mathematical expression is:

[0096]

[0097] Where ε(·) is the unit step function, defined as:

[0098]

[0099] In this feature extraction scenario, it can be used to determine whether the value of each distance cell is greater than zero, and set the result to 1 or 0 so that the distance cell that meets the condition is counted into the count of the scattering center.

[0100] 2. Calculate the equivalent size E size This quantitative index is used to describe the radial size of the target. If S is defined as the position vector of the distance unit where the echo intensity in the HRRP signal is greater than half of the maximum value, that is:

[0101] S={i|x(i)≥max(X) / 2}

[0102] The mathematical expression of the equivalent size is:

[0103] E size =S(N scatter )-S(1)

[0104] 3. Calculate the entropy of the HRRP signal, which is used to quantify the randomness of the variable. Its mathematical expression is:

[0105]

[0106] 4. Calculate the standard deviation std(dB) of HRRP, which is used to quantify the degree of dispersion of data. Its mathematical expression is:

[0107]

[0108] 5. Calculate the deviation (dB) of HRRP, which is used to quantify the distribution characteristics of the data. Its mathematical expression is:

[0109]

[0110] 6. Calculate the HRRP irregularity (dB). This index is used to quantify the local deviation of the HRRP signal and reflects the relative size relationship between the current distance unit and the previous and next distance units. Its mathematical expression is:

[0111]

[0112] 7. Calculate the frequency domain incoherent range image of HRRP This indicator can suppress the azimuth sensitivity of HRRP. Its mathematical expression is:

[0113]

[0114] 8. Calculate the central moment feature of HRRP. The central moment is a basic translation-invariant feature that can effectively reveal the structural information of the ship.

[0115] Assume the first-order origin moment to be: Then the p-order central moment C p The mathematical expression is:

[0116]

[0117] In the embodiment of the present application, 2nd to 6th order central moment features are selected.

[0118] Step 103 , extracting micro-motion features from the accumulated HRRP azimuths using short-time Fourier transform and Fourier coefficient fitting respectively.

[0119] The existing recognition based on HRRP of sea surface ships mainly focuses on the structural characteristics of ships, but the time domain echo of radar ship target HRRP also contains rich micro-motion characteristics. From the perspective of space-based radar, the shaking of the ship will change the attitude angle of the ship in the radar, causing the regular movement of scattering points across distance units.

[0120] For the ship motion model, this application mainly considers the roll of the ship and the linear motion between the ship and the radar, which can cause HRRP to flicker and wander. For this micro-motion, this application uses Fourier series to fit the position of the strong scattering point in the distance unit as the ship's attitude angle changes; on the other hand, this application uses short-time Fourier transform to extract the time-frequency characteristics of specific units.

[0121] Specifically, the steps of extracting micro-motion features from the accumulated HRRP azimuths by Fourier coefficient fitting include:

[0122] The change of the distance unit of the strongest scattering point over time is extracted, and the change of the position of the strong scattering point in the distance unit with the ship attitude angle is fitted by Fourier series. The coefficient of Fourier series is used as the micro-motion feature of the ship. The calculation formula of Fourier series is:

[0123]

[0124] Where, f(t) represents the function of the HRRP signal, with time t as the independent variable, a0 is the constant term in the Fourier series, and a n is the cosine term coefficient of the Fourier series, b n is the coefficient of the sine term of the Fourier series, ω is the angular frequency of the fundamental frequency, and n is the harmonic number.

[0125] It is understandable that different types of ships have different masses, centers of gravity, and shapes, and their micro-motion characteristics, such as sway amplitude and sway period, will also vary.

[0126] In a possible embodiment, the order of the Fourier series is set to 3, and its coefficients are used as the micro-motion characteristics of the ship.

[0127] Specifically, the steps of extracting micro-motion features from the accumulated HRRP orientations by short-time Fourier transform include:

[0128] Multiple distance units near the strong scattering point are selected to monitor the change of the reflected signal at different distance units over time, and the change of multiple distance units is monitored simultaneously by short-time Fourier transform.

[0129] Let the i-th HRRP after accumulation be X i , is the amplitude of the nth distance unit of the i-th HRRP. The signal data on each distance unit is short-time Fourier transformed, and the calculation formula is:

[0130]

[0131] Among them, w(m) is the selected window function, which is used to weight and smooth the data. By performing Fourier transform on the data in different time windows, the frequency domain features are extracted to capture the micro-motion information.

[0132] Furthermore, multiple short-time Fourier transform matrices calculated for the selected multiple distance units are added, and the matrix obtained by the addition is recorded as STFT. The matrix STFT is subjected to singular value decomposition to reduce its characteristic dimension, that is:

[0133]

[0134] Among them, STFT m×n It represents the matrix after short-time Fourier transformation, which is the result matrix obtained by transforming the signals of multiple frames of data and multiple range units, and is used to describe the micro-motion characteristics of the target; U m×k is an m×k matrix containing the left singular vectors of the STFT matrix, which is used to retain the main time or distance feature information in the matrix; Σ k×k is a k×k diagonal matrix containing singular values, which are arranged from large to small, representing the energy distribution and principal component strength in the data. The largest singular value corresponds to the most important feature, and the gradually decreasing singular values ​​correspond to the secondary features; It is the transpose of the V matrix, with a size of k×n. The V matrix contains the right singular vectors of the STFT matrix, which represent the contribution of different features of the signal in the frequency domain.

[0135] Finally, through singular value decomposition, the present application can choose to retain larger singular values ​​and their corresponding vectors, thereby reducing the data dimension while retaining the main features of the signal to describe the micro-motion characteristics of the target.

[0136] Step 104, input the structural features and micro-motion features into a meta-classifier to obtain a sea surface target recognition result.

[0137] In the embodiment of the present application, the meta-classifier is composed of a stack of a first base classifier and a second base classifier, the first base classifier is trained based on structural feature samples, and the second base classifier is trained based on micro-motion feature samples.

[0138] It can be understood that the meta-classifier is a stacked classifier. The stacked classifier is an integrated learning technology that uses multiple base classifiers to learn and predict different features, and then uses the predicted probability as a new feature to train a new meta-classifier to achieve feature fusion and classification.

[0139] In a possible embodiment, the flow chart of the training and stacking process of the meta-classifier is as follows: Figure 2 shown.

[0140] Reference Figure 2 ,Firstly collect training samples and preprocess them.

[0141] It can be understood that the training samples are other sea surface high-resolution range image HRRP matrix samples of ships, and the preprocessing steps can be synchronously referred to the normalization step and the accumulation step in step 101, which will not be repeated in this application.

[0142] Then, based on the preprocessed training samples, the structural features and micro-motion features are extracted, and different base classifiers are trained respectively to learn and identify the structural features of sea surface targets and the micro-motion features of targets. Each base classifier is trained separately to optimally identify and classify different target types within its specific feature domain.

[0143] After training, the two base classifiers are stacked to form a meta-classifier. The meta-classifier further analyzes and makes decisions based on the information obtained from the first and second base classifiers. By combining structural features and micro-motion features, the meta-classifier can more comprehensively identify and classify sea surface targets, improving classification accuracy and robustness.

[0144] It is understandable that after each classifier is trained, it needs to be tested with test samples. Through the evaluation of test samples, the performance indicators such as accuracy, precision, and recall of each base classifier can be obtained, and potential overfitting or underfitting problems can be identified.

[0145] Finally, the meta-classifier obtains the final sea surface target recognition result based on the output of the two base classifiers and the test results. In addition, through the feedback of the test samples, the weight or stacking method of the meta-classifier can be further tuned to improve its comprehensive performance.

[0146] In the embodiments of the present application, support vector machines, random forests, and logistic regression classifiers can be used as base classifiers of the present application, and logistic regression classifiers can be used as meta-classifiers of the present application, and the present application does not make any specific limitations on this.

[0147] In order to further demonstrate the beneficial effects of the present application, the present application also conducted a group of simulation experiments, and the specific results are shown in the following examples.

[0148] Example 1

[0149] (1) Simulation experiment conditions:

[0150] The present invention uses electromagnetic simulation software CST to simulate the simulation data of a certain X-band broadband space-based radar on sea surface ships. The four models are small civilian ships (18 meters), Fujian ship (178 meters), 055 missile destroyer (158 meters) and diluted corner reflector array (25 meters), all of which are adjusted and scaled using Hypermesh software.

[0151] (2) Experimental data setting:

[0152] The training data set contains 280 groups of HRRP matrix data per category, and each group of HRRP matrix contains 1280 frames of one-dimensional HRRP; the test data set contains 120 groups of HRRP matrix data per category, and each group of HRRP matrix contains 1280 frames of one-dimensional HRRP.

[0153] This application uses three different classifiers as base classifiers for comparison, namely support vector machine, random forest, and logistic regression classifiers. The specific setting parameters are as follows:

[0154] 1) Support vector machine: SVM uses grid search method to find the best solution. The kernel function of the base classifier for structural feature training is set to radial basis function, and the kernel function coefficient gamma is set to 0.1; the kernel function of the base classifier for micro-motion feature training is set to radial basis function, and the kernel function coefficient gamma is set to 1.

[0155] 2) Random Forest: The number of decision trees of the base classifier trained with structural features is set to 100; the number of decision trees of the base classifier trained with micro-motion features is set to 1000.

[0156] 3) Logistic regression: The maximum number of iterations of the logistic regression classifier is 10,000.

[0157] 4) Meta-classifier: Meta-classifiers all use logistic regression classifiers with a maximum number of iterations of 1000.

[0158] (3) Experimental results

[0159] Table 1, Table 2, and Table 3 are the recognition accuracy rates of the base classifier trained only with structural features, the base classifier trained only with micro-motion features, and the meta-classifier trained with both structural features and micro-motion features, respectively. From the data in Table 1, Table 2, and Table 3, it can be seen that after fusing structural features and micro-motion features, the recognition accuracy of the classifier is significantly improved, which proves the effectiveness of this fusion method.

[0160] Table 1

[0161] Classifier Methods Accuracy SVM 80.3％ Random Forest 85.4％ Logistic Regression 64.0％

[0162] As shown in Table 1, when only structural features are used, random forest performs best, achieving 85.4% accuracy, followed by SVM, while logistic regression has the lowest accuracy.

[0163] Table 2

[0164] Classifier Methods Accuracy SVM 72.0％ Random Forest 70.6％ Logistic Regression 66.9％

[0165] As shown in Table 2, when only micro-motion features are used, the recognition accuracy is generally low, especially the accuracy of random forest and SVM is below 72%. Micro-motion features may be more unstable in recognition than structural features, especially when the structural features of the target are not obvious, the classification effect of relying solely on micro-motion features is not ideal.

[0166] Table 3

[0167] Classifier Methods Accuracy SVM 87.1％ Random Forest 88.5％ Logistic Regression 75.2％

[0168] As shown in Table 3, after fusing structural features and micro-motion features, the accuracy of all classifiers has been significantly improved. The accuracy of random forest reached 88.5%, followed by SVM at 87.1%, and the accuracy of logistic regression also increased to 75.2%. Compared with the data in Tables 1 and 2, these values ​​show a significant improvement, especially for random forest and SVM.

[0169] By comparing the above data, we can conclude that the fusion of structural features and micro-motion features can significantly improve the recognition accuracy of the classifier. This fusion method utilizes the stability of structural features and the dynamic information of micro-motion features, so as to more comprehensively describe the target and improve the classification effect. This improvement has practical significance for applications such as sea surface target recognition.

[0170] In order to implement the above embodiments, the present application also proposes a space-based radar sea surface target recognition device. Figure 3 A schematic structural diagram of a space-based radar sea surface target recognition device 10 provided in an embodiment of the present application.

[0171] like Figure 3 As shown, the device comprises:

[0172] The preprocessing module 100 is used to obtain the high-resolution range image HRRP matrix of the ship's sea surface, perform amplitude normalization processing on the HRRP matrix, and stack and accumulate the normalized HRRP matrix along the azimuth direction;

[0173] The structural feature extraction module 200 is used to extract the structural features of the accumulated HRRP, wherein the structural features include equivalent scattering center dimension, equivalent size, entropy, standard deviation, deviation, irregularity, frequency domain incoherent range image and central moment feature;

[0174] A micro-motion feature extraction module 300 is used to extract micro-motion features from the accumulated HRRP azimuth using short-time Fourier transform and Fourier coefficient fitting respectively;

[0175] The target recognition module 400 is used to input the structural features and micro-motion features into a meta-classifier to obtain a sea surface target recognition result, wherein the meta-classifier is composed of a first base classifier and a second base classifier stacked together, the first base classifier is trained based on the structural feature samples, and the second base classifier is trained based on the micro-motion feature samples.

[0176] In order to implement the above embodiments, the present application also proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.

[0177] In order to implement the above embodiments, the present application also proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided by the above embodiments.

[0178] In order to implement the above embodiments, the present application also proposes a computer program product, including a computer program, which implements the methods provided by the above embodiments when executed by a processor.

[0179] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this application are in compliance with relevant laws and regulations and do not violate public order and good morals.

[0180] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.

[0181] The present application is expected to provide an implementation scheme for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.

[0182] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0183] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0184] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0185] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0186] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0187] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0188] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0189] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

[0190] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution of this application can be achieved, and this document is not limited here.

[0191] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A method for identifying sea surface targets using space-based radar, characterized in that: The following steps are involved: Acquire a high-resolution range image HRRP matrix of the ship's sea surface, perform amplitude normalization processing on the HRRP matrix, and stack and accumulate the normalized HRRP matrix along the azimuth direction; Extracting structural features of the accumulated HRRP, wherein the structural features include equivalent scattering center dimension, equivalent size, entropy, standard deviation, bias, irregularity, frequency domain incoherent range image and central moment features; Short-time Fourier transform and Fourier coefficient fitting are used to extract micro-motion features from the accumulated HRRP azimuths. The structural features and the micro-motion features are input into a meta-classifier to obtain a sea surface target recognition result, wherein the meta-classifier is composed of a first base classifier and a second base classifier stacked together, the first base classifier is trained based on structural feature samples, and the second base classifier is trained based on micro-motion feature samples.

2. The method according to claim 1, characterized in that The step of obtaining a high-resolution range image HRRP matrix of the ship's sea surface, performing amplitude normalization processing on the HRRP matrix, and stacking and accumulating the normalized HRRP matrix along the azimuth direction includes: Get the HRRP matrix of the ship, recorded as: Wherein, X is the HRRP matrix, M is the number of rows of the HRRP matrix, indicating the number of time frames; N is the number of columns of the HRRP matrix, indicating the number of distance units; x ij represents a single element in the ith row and jth column of the HRRP matrix, that is, the signal strength of the jth distance unit in the ith frame; The HRRP matrix is ​​normalized row by row, where the i-th row HRRP is denoted as X i =[x i1 …x iN ], the elements in the HRRP matrix are normalized according to the following formula, the expression is: The normalized HRRP matrix is ​​superimposed along the azimuth direction, and the expression is: Among them, X i ' represents the accumulated HRRP.

3. The method according to claim 2, characterized in that The structural characteristics of the HRRP after the extraction and accumulation include: For a HRRP after extraction and accumulation, that is, X i ′=[x i1 …x iN ], denoted as X i =[x(1)…x(N)], the structural features are calculated by the following formulas, including: Calculate the equivalent scattering center dimension N scatter , the mathematical expression is: Where ε(·) is the unit step function, defined as: Calculate the equivalent size E size , the mathematical expression is: E size =S(N scatter )-S(1) Among them, S is the position vector of the distance unit where the echo intensity in the HRRP signal is greater than half of the maximum value, and its calculation expression is: S={i|x(i)≥max(X) / 2} Calculate the entropy, the mathematical expression is: Calculate the standard deviation std(dB), the mathematical expression is: Calculate the deviation (dB), the mathematical expression is: Calculate the irregularity (dB), the mathematical expression is: Calculate the frequency domain incoherent range image The mathematical expression is: Calculate the central moment characteristics and assume the first-order origin moment is: Then the p-order central moment C p The mathematical expression is:

4. The method according to claim 3, characterized in that The method of extracting micro-motion features from the accumulated HRRP azimuth using Fourier coefficient fitting includes: The change of the distance unit of the strongest scattering point over time is extracted, and the change of the position of the strong scattering point in the distance unit with the ship attitude angle is fitted using the Fourier series, and the coefficient of the Fourier series is used as the micro-motion feature of the ship. The calculation formula of the Fourier series is: Where, f(t) represents the function of the HRRP signal, with time t as the independent variable, a0 is the constant term in the Fourier series, and a n is the cosine term coefficient of the Fourier series, b n is the coefficient of the sine term of the Fourier series, ω is the angular frequency of the fundamental frequency, and n is the harmonic number.

5. The method according to claim 4, characterized in that The extracting micro-motion features from the accumulated HRRP orientation using short-time Fourier transform includes: Select multiple distance units near the strong scattering point, use short-time Fourier transform to monitor the changes of multiple distance units simultaneously, and set the i-th HRRP after accumulation as X i , is the amplitude of the nth distance unit of the i-th HRRP. The signal data on each distance unit is short-time Fourier transformed, and the calculation formula is: Among them, w(m) is the selected window function; The multiple short-time Fourier transform matrices calculated by the selected multiple distance units are added together, and the matrix obtained by the addition is recorded as STFT. The matrix STFT is subjected to singular value decomposition to reduce its characteristic dimension, that is: Among them, STFT m×n represents the matrix after short-time Fourier transform; U m×k is an m×k matrix containing the left singular vectors of the STFT matrix; Σ k×k is a k×k diagonal matrix containing singular values; It is the transpose of the V matrix, with a size of k×n. The V matrix contains the right singular vectors of the STFT matrix; According to the singular values ​​obtained by decomposition, the micro-motion characteristics of the ship are extracted.

6. A space-based radar sea surface target recognition device, characterized in that: include: A preprocessing module is used to obtain a high-resolution range image HRRP matrix of the ship's sea surface, perform amplitude normalization processing on the HRRP matrix, and stack and accumulate the normalized HRRP matrix along the azimuth direction; A structural feature extraction module is used to extract the structural features of the accumulated HRRP, wherein the structural features include equivalent scattering center dimension, equivalent size, entropy, standard deviation, deviation, irregularity, frequency domain incoherent range image and central moment features; A micro-motion feature extraction module is used to extract micro-motion features from the accumulated HRRP azimuth using short-time Fourier transform and Fourier coefficient fitting respectively; The target recognition module is used to input the structural features and the micro-motion features into a meta-classifier to obtain a sea surface target recognition result, wherein the meta-classifier is composed of a first base classifier and a second base classifier stacked together, the first base classifier is trained based on the structural feature samples, and the second base classifier is trained based on the micro-motion feature samples.

7. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 5 when executed by a processor.

9. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 5 when being executed by a processor.

Images