Multifunction radar wave cell extraction method and system based on dot matrix model

Abstract

The application discloses a multifunctional radar wave element extraction method and system based on a dot matrix model, and the method comprises the following steps: for a multifunctional radar (MFR) pulse sequence, a dot matrix (D) based on a pulse repetition interval (PRI) value, a pulse width (PW) value and a radio frequency (RF) value is respectively constructed according to the threshold values of the corresponding PRI value, PW value and RF value of the MFR pulse sequence; PRI ,D RF ,D PW ; the constructed dot matrix (D PRI ,D RF ,D PW is respectively subjected to image preprocessing, image hole filling and image closing operation; an L-shaped corner point on a main diagonal line of the preprocessed dot matrix is detected by using a Harris corner point detection operator, that is, a wave element starting point; and a union set of the detection results is obtained, that is, a wave element starting pulse set. The application does not depend on cluster parameter setting, can fully utilize the overall modulation law of the wave element, reduces the false alarm probability while improving the wave element detection probability, and still has good performance in the case that the pulse is lost or the parameter measurement error is large.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing technology, and in particular to a multifunctional radar waveform unit extraction method and system based on a dot matrix model. Background Technology

[0002] With the rapid development of radar technology, especially multi-function radar (MFR) represented by active phased array radar, its widespread deployment and application in the military have resulted in radar signals in the battlefield electromagnetic environment exhibiting characteristics of high density, complex forms, and diverse functions, posing significant challenges to electronic reconnaissance. Waveform elements, as a fundamental component of MFR signals, are essential for acquiring electronic intelligence. The quality of waveform element extraction directly affects the performance of radar radiation source identification, and consequently, the level of battlefield electromagnetic situational awareness and threat assessment.

[0003] Based on the characteristics of different signal models, there are currently three main approaches to MFR waveform unit extraction: First, waveform unit extraction is based on Hidden Markov Models (HMMs). Second, waveform unit extraction is based on Syntactic Models, which constructs the MFR signal into a three-level model from bottom to top: pulse, radar word, and radar phrase, where the radar word corresponds to the waveform unit of the MFR signal. Third, waveform unit extraction is based on hierarchical models, which divides the MFR signal into three levels from top to bottom: functional layer, task layer, and waveform layer.

[0004] The aforementioned waveform unit extraction methods mainly suffer from the following problems: Both Hidden Markov Model (HMM) and Syntactic Model (SMM)-based methods require prior knowledge such as waveform libraries and waveform transition rules, which limits their effectiveness in detecting non-cooperative MFRs (Multi-Frequency Patterns). The lack of prior knowledge significantly degrades their recognition performance. Hierarchical Model-based methods primarily employ unsupervised algorithms such as clustering to identify waveform unit change points, over-relying on clustering parameter settings and neglecting the overall modulation rules of the waveform units. Summary of the Invention

[0005] This invention addresses the problem of poor recognition performance in existing MFR waveform unit extraction methods, which over-rely on clustering parameter settings and neglect the overall modulation pattern of waveform units. It proposes a multifunctional radar waveform unit extraction method and system based on a dot matrix model. This method incorporates biological sequence analysis technology, utilizing the repeatability of pulses within the same waveform unit and the differences in pulses between different waveform units to reconstruct a two-dimensional MFR waveform unit into a dot matrix model. A corner detection algorithm is then used to identify the starting point of the waveform unit. This invention does not rely on clustering parameter settings and can fully utilize the overall modulation pattern of waveform units, improving the detection probability of waveform units while reducing the false alarm probability. It maintains good performance even under conditions of pulse loss and large parameter measurement errors.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] This invention proposes a multifunctional radar waveform unit extraction method based on a dot matrix model, comprising:

[0008] Step 1: For the multi-function radar MFR pulse sequence, construct a dot matrix D based on the corresponding pulse repetition interval (PRI), pulse width (PW), and radio frequency (RF) values, respectively. PRI D RF D PW ;

[0009] Step 2: Apply the constructed raster image D PRI D RF D PW Image preprocessing is performed separately, followed by image hole filling and image closing operations;

[0010] Step 3: Use the Harris corner detection operator to detect the L-shaped corner points on the main diagonal of the preprocessed dot matrix, which are the starting points of the waveform units.

[0011] Step 4: Take the union of the detection results from Step 3 to obtain the set of initial pulses for the waveform unit.

[0012] Further, step 1 includes:

[0013] For an MFR pulse train containing n pulses, P = {P1, P2, ..., P...} n The PRI, RF, PW values ​​of each pulse contained therein are represented by the following sequence:

[0014]

[0015] In the formula, x i Let y represent the PRI value of the i-th pulse in pulse train P, where i = 1, 2, ..., n. iz represents the RF value of the i-th pulse. i This represents the PW value of the i-th pulse;

[0016] Improving the handling of pulse pairs with identical signal parameter values ​​by setting a threshold:

[0017]

[0018] In the formula, Indicates the pulse pair P i ,P j The PRI value matching situation, Threshold PRI Indicates the pulse pair P i ,P j The PRI values ​​are judged to be the same threshold; Indicates the pulse pair P i ,P j RF value matching status, Threshold RF Indicates the pulse pair P i ,P j The RF values ​​were judged to be the same threshold; Indicates the pulse pair P i ,P j PW value matching status, Threshold PW Indicates the pulse pair P i ,P j The PW values ​​were judged to be the same threshold;

[0019] Therefore, an N×N dot matrix D can be constructed from the PRI, PW, and RF values ​​respectively. PRI D RF D PW The expression is as follows:

[0020]

[0021] Furthermore, in step 2, the image hole filling adopts a seed filling algorithm, and a 3×3 convolution kernel is selected for image closing operation.

[0022] Further, step 3 includes:

[0023] For the pixel (i,i) located on the main diagonal in the preprocessed dot matrix image, calculate its corner response function value (CRF). If the CRF of the point is greater than 0, the point can be determined to be a corner point, i.e., the starting point of the waveform unit. The position number i corresponds to the starting pulse number of the waveform unit in the MFR pulse train.

[0024] Another aspect of this invention proposes a multifunctional radar waveform unit extraction system based on a dot matrix model, comprising:

[0025] The dot matrix construction module is used to construct a dot matrix D based on the corresponding pulse repetition interval (PRI), pulse width (PW), and radio frequency (RF) values ​​for a multi-function radar MFR pulse sequence. PRI D RF D PW ;

[0026] The image preprocessing module is used to process the constructed raster image D. PRI D RF D PW Image preprocessing is performed separately, followed by image hole filling and image closing operations;

[0027] The corner detection module is used to detect L-shaped corners on the main diagonal of the preprocessed dot matrix image using the Harris corner detection operator, i.e., the starting points of the waveform units.

[0028] The module for obtaining the initial pulse set is used to take the union of the detection results of the corner detection module, thus obtaining the initial pulse set of the waveform unit.

[0029] Furthermore, the dot matrix construction module is specifically used for:

[0030] For an MFR pulse train containing n pulses, P = {P1, P2, ..., P...} n The PRI, RF, PW values ​​of each pulse contained therein are represented by the following sequence:

[0031]

[0032] In the formula, x i Let y represent the PRI value of the i-th pulse in pulse train P, where i = 1, 2, ..., n. i z represents the RF value of the i-th pulse. i This represents the PW value of the i-th pulse;

[0033] Improving the handling of pulse pairs with identical signal parameter values ​​by setting a threshold:

[0034]

[0035] In the formula, Indicates the pulse pair P i ,P j The PRI value matching situation, Threshold PRI Indicates the pulse pair P i ,P j The PRI values ​​are judged to be the same threshold; Indicates the pulse pair P i ,P jRF value matching status, Threshold RF Indicates the pulse pair P i ,P j The RF values ​​were judged to be the same threshold; Indicates the pulse pair P i ,P j PW value matching status, Threshold PW Indicates the pulse pair P i ,P j The PW values ​​were judged to be the same threshold;

[0036] Therefore, an N×N dot matrix D can be constructed from the PRI, PW, and RF values ​​respectively. PRI D RF D PW The expression is as follows:

[0037]

[0038] Furthermore, in the image preprocessing module, the image hole filling adopts a seed filling algorithm, and a 3×3 convolution kernel is selected for image closing operation.

[0039] Furthermore, the corner detection module is specifically used for:

[0040] For the pixel (i,i) located on the main diagonal in the preprocessed dot matrix image, calculate its corner response function value (CRF). If the CRF of the point is greater than 0, the point can be determined to be a corner point, i.e., the starting point of the waveform unit. The position number i corresponds to the starting pulse number of the waveform unit in the MFR pulse train.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] This invention introduces biological sequence analysis technology, utilizing the repeatability of pulses within the same waveform unit and the differences in pulses between different waveform units to reconstruct a one-dimensional MFR waveform unit into a two-dimensional dot matrix model. A corner detection algorithm is then used to identify the starting point of the waveform unit. This invention does not rely on clustering parameter settings and can fully utilize the overall modulation pattern of the waveform unit, improving the detection probability of the waveform unit while reducing the false alarm probability. It maintains good performance even under conditions of pulse loss and large parameter measurement errors. Attached Figure Description

[0043] Figure 1 This is a basic flowchart of a multifunctional radar waveform unit extraction method based on a dot matrix model according to an embodiment of the present invention;

[0044] Figure 2 This is an example diagram of the MFR pulse train dot matrix according to an embodiment of the present invention;

[0045] Figure 3 This is a schematic diagram illustrating the effect of noise on the main diagonal pattern structure features according to an embodiment of the present invention;

[0046] Figure 4 This is the bitmap preprocessing flow according to an embodiment of the present invention;

[0047] Figure 5 This is an MFR pulse train dot matrix diagram containing 150 waveform units, as described in an embodiment of the present invention.

[0048] Figure 6 Different σ for embodiments of the present invention 2 TPR change curves under RDP values;

[0049] Figure 7 Different σ for embodiments of the present invention 2 FPR change curve under RDP value;

[0050] Figure 8 This is a schematic diagram of the architecture of a multifunctional radar waveform unit extraction system based on a dot matrix model, according to an embodiment of the present invention. Detailed Implementation

[0051] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments:

[0052] like Figure 1 As shown, a multifunctional radar waveform unit extraction method based on a dot matrix model includes:

[0053] Step 1: For the multi-function radar MFR pulse sequence, construct a dot matrix D based on the corresponding Pulse Repetition Interval (PRI), Pulse Width (PW), and Radio Frequency (RF) values, respectively. PRI D RF D PW ;

[0054] Step 2: Apply the constructed raster image D PRI D RF D PW Image preprocessing is performed separately, followed by image hole filling and image closing operations;

[0055] Step 3: Use the Harris corner detection operator to detect the L-shaped corner points on the main diagonal of the preprocessed dot matrix, which are the starting points of the waveform units.

[0056] Step 4: Take the union of the detection results from Step 3 to obtain the set of initial pulses for the waveform unit.

[0057] Specifically, this embodiment constructs an MFR pulse train dot matrix model based on the dot matrix principle. This model can intuitively display the variation patterns of modulation type, parameter values, and the number of pulses among different waveform units in the MFR pulse train through the combination of patterns of different shapes and sizes on the main diagonal.

[0058] (a) Principle of dot matrix diagrams

[0059] Dot plots are a simple and effective method in biological sequence analysis, often used to detect similar regions within a single sequence (i.e., repetitive sequences) or between different sequences. A dot plot is a two-dimensional representation method; when detecting similar points within a single sequence, both the x-axis and y-axis of the dot plot are represented using the same sequence. If points at different positions within the sequence can match, then that position is marked on the dot plot.

[0060] In a two-dimensional space, a sample sequence {X1,X2,…,X} of length L can be represented. L Arrange the samples from left to right along the X-axis and from top to bottom along the Y-axis. For any position (i,j) in space, 1≤i≤L, 1≤j≤L, if the corresponding sample value X i and X j If the positions are the same, then the position is marked with a point. Therefore, a binary function can be defined to represent this marking rule.

[0061]

[0062] In the formula, the binary function T i,j This represents the case where the position (i,j) in two-dimensional space is marked.

[0063] Since each position in the two-dimensional space can be marked with a point according to the binary function, a dot matrix D of size L×L can be obtained. The dot matrix D can also be regarded as a binary image containing L×L pixels.

[0064]

[0065] (2) MFR pulse train dot matrix model

[0066] During the operation of the MFR (Medium-Range Array), its waveform units are sequentially linked according to a certain timing sequence under the control of the signal generation mechanism to form the final pulse train signal transmitted by the radar. Therefore, by constructing an MFR pulse train dot matrix model, we can not only reflect its overall variation law, but also analyze individual waveform units independently.

[0067] For an MFR pulse train {P1, P2, ..., P} containing n pulses... n In this context, the PRI, RF, PW values ​​of each pulse contained therein can be represented by the following sequence:

[0068]

[0069] In the formula, x i y represents the PRI value of the i-th (i = 1, 2, ..., n) pulse in pulse train P. i z represents the RF value of the i-th pulse. i This represents the PW value of the i-th pulse.

[0070] When constructing a dot matrix diagram for the actual received MFR pulse train based on the dot matrix diagram principle, it is also necessary to consider the impact of noise on the PRI, RF, and PW values. The situation of incorrect labeling of pulse pairs with the same signal parameter values ​​can be improved by setting a certain threshold for the binary function.

[0071]

[0072] In the formula, Indicates the pulse pair P i ,P j The PRI value matching situation, Threshold PRI Indicates the pulse pair P i ,P j The PRI values ​​are judged to be the same threshold; Indicates the pulse pair P i ,P j RF value matching status, Threshold RF Indicates the pulse pair P i ,P j The RF values ​​were judged to be the same threshold; Indicates the pulse pair P i ,P j PW value matching status, Threshold PW Indicates the pulse pair P i ,P j The PW values ​​are judged to be the same threshold.

[0073] Therefore, an N×N dot matrix D can be constructed from the PRI, PW, and RF values ​​respectively. PRI D RF D PW The expression is as follows:

[0074]

[0075] In practical signal reception environments, the PRI, RF, and PW thresholds can be adjusted based on factors such as receiver measurement errors. Generally, the threshold is set to... PRI Set the Threshold to 6-10us. RFSet to 6-10Hz, and set Threshold. PW Set to 1~3us.

[0076] The dot plot of the MFR sequence for task P1, constructed based on PRI, RF, and PW, is as follows: Figure 2 As shown.

[0077] Figure 2 In the diagram, an asterisk (*) represents the starting pulse of a waveform unit, which is the intersection of image edges. White dots represent pulses within the waveform unit. Since the MFR pulse train is composed of successive waveform units, differences in the modulation type, parameter values, and number of pulses within each waveform unit result in patterns of different shapes and sizes appearing on the main diagonal of the MFR pulse train dot matrix. These patterns can be distinguished by the intersections of their edges. These intersections, as feature points on the image, are generally referred to as corner points in digital image processing.

[0078] Furthermore, in step 2, the image hole filling adopts a seed filling algorithm, and a 3×3 convolution kernel is selected for image closing operation.

[0079] Specifically, raster image preprocessing includes:

[0080] Because noise can affect the actual detection of MFR, the received MFR pulse sequence may experience pulse loss and parameter measurement errors. This results in holes inside the solid squares representing individual waveform units on the main diagonal of the dot matrix, and blurred edges; the striped squares no longer have obvious striped characteristics; and solid noise blocks appear on the line segments, such as... Figure 3 As shown.

[0081] The disruption of structural features along the main diagonal of a raster image can affect the accuracy of subsequent corner detection. Therefore, image preprocessing of the raster image is necessary: ​​hole filling and closing operations are used to suppress noise blocks and highlight image corners. The preprocessing flow is as follows: Figure 4 As shown.

[0082] Image hole filling uses a seed filling algorithm: assuming a seed pixel is pre-set in the area to be filled, then starting from that pixel, the search for the next pixel is performed using a four-way or eight-way algorithm, and so on until all pixels in the area are found.

[0083] Image dilation and erosion are fundamental operations in mathematical morphology. The main function of image dilation is to incorporate points from surrounding areas into the image; here, it's used to connect discontinuous points in striped squares and fill holes in solid squares. The mathematical definition is as follows:

[0084]

[0085] In the formula, f(x,y) represents the input image, and D f Let f represent the domain of f, b(s,t) represent the structuring element, and D represent the domain of f. b This represents the domain of b.

[0086] The main function of image erosion is to eliminate the boundaries and protrusions on the boundaries of the input image. Here, it is used to remove noise blocks on line segments and to repair blurred boundaries caused by image dilation. The mathematical definition is as follows:

[0087] (fΘb)(x,y)=min{f(xs,yt)-b(s,t)(xs,yt)∈D f ;(s,t)∈D b} (7)

[0088] The process of dilating and then eroding an image is called the closing operation. It is used to remove small (relative to the size of the structuring element) dark details in an image while keeping the bright parts relatively unaffected. The mathematical definition is as follows:

[0089]

[0090] Since the same waveform unit contains three or more pulses, and the dot matrix is ​​a binary image formed by comparing the MFR pulse sequence with itself, a 3×3 convolution kernel is selected for closing operation to fill the pattern holes on the main diagonal of the MFR pulse dot matrix and repair the pattern boundary.

[0091] Further, step 3 includes:

[0092] For the pixel (i,i) located on the main diagonal in the preprocessed dot matrix image, calculate its corner response function value (CRF). If the CRF of the point is greater than 0, the point can be determined to be a corner point, i.e., the starting point of the waveform unit. The position number i corresponds to the starting pulse number of the waveform unit in the MFR pulse train.

[0093] Specifically, since the starting point of the waveform unit in the MFR pulse train is the corner point on the main diagonal of the dot matrix image, the starting point of the waveform unit can be found by performing corner detection on the dot matrix image, thereby extracting the waveform unit. After preprocessing, the dot matrix image is fully filled and the edges are relatively clear, which meets the conditions for corner detection. This invention uses the Harris corner detection operator, the principle of which is as follows.

[0094] Suppose a small window centered at the target pixel (x, y) moves u in the X direction and v in the Y direction, then the analytical expression for its grayscale change is:

[0095]

[0096] In the formula, E x,y W represents the amount of grayscale change within the window. x,y represents the window function, and I represents the grayscale function of the image.

[0097] After neglecting infinitesimal terms, E(x,y) is reduced to a quadratic form:

[0098]

[0099] In the formula,

[0100]

[0101] In the formula, M represents the real symmetric local structure matrix, and I x I represents the gradient in the x-direction of image I. y This represents the gradient in the y-direction of image I.

[0102] Analysis of matrix M reveals that its eigenvalues ​​are the first-order curvatures of its autocorrelation function. If both curvature values ​​are high, then the point is considered a corner. Therefore, by finding the eigenvalues ​​λ1 and λ2 of matrix M, the corner response function (CRF) is defined as follows:

[0103] CRF = det(M) - k·trace 2 (M) (12)

[0104] In the formula,

[0105]

[0106] In the formula, det(M) represents the determinant of matrix M, trace(M) represents the trace of matrix M, and k is a constant, usually taken as 0.04.

[0107] Calculate the CRF of the pixel (i,i) located on the main diagonal of the dot matrix. If the CRF of the pixel is greater than 0, the pixel is identified as a corner point, and the position number i corresponds to the starting pulse number of the waveform unit in the MFR pulse train.

[0108] Based on the concept of data fusion, a more complete and accurate waveform unit extraction result can be obtained than using a single parameter value. Therefore, the joint modulation pattern of the PRI, RF, and PW values ​​of the MFR pulse train should be fully utilized. For MFR, search and tracking are its most commonly used functions. Fixed modulation types are often used in search and tracking. PRI, RF, and PW often have at least one fixed value. The Harris corner detection operator has the best recognition effect on the starting point of waveform units of this modulation type. Therefore, the starting pulse of a waveform unit that cannot be successfully extracted by a single signal parameter value can be supplemented by the extraction results of the other two parameter values. By taking the union of the waveform unit starting pulse indices of PRI, RF, and PW, the final waveform unit starting point extraction result is obtained, thereby improving the detection probability.

[0109]

[0110] Specifically, the process for extracting MFR waveform units based on the dot matrix diagram is as follows:

[0111] Algorithm 1: MFR Waveform Unit Extraction Algorithm Based on Dot Array

[0112] Input: MFR pulse train P, dot plot threshold Threshold = {Threshold PRI Threshold RF Threshold PW};

[0113] Output: Set of initial pulses for waveform units, C;

[0114] 1. Based on the Threshold, construct the dot matrix D based on PRI, PW, and RF respectively. PRI D RF D PW ;

[0115] 2. Perform image preprocessing on the constructed raster image, including image hole filling and image closing operation based on 3×3 convolution kernels;

[0116] 3. Calculate the CRF of the elements on the main diagonal of the preprocessed raster image;

[0117] 4. For the dot matrix D PRI D RF D PW The union of the calculated pulse sequence numbers with CRF > 0 is the set of starting pulses C for the waveform unit.

[0118] To verify the effectiveness of the present invention, the following experiment was conducted:

[0119] The experimental data for this invention comes from a waveform library of a two-dimensional phased array radar given in reference 1 [Yang Liu, Zhu Weigang, Lü Shouye, et al. Waveform unit extraction method for non-cooperative multi-functional radar [J]. Systems Engineering and Electronics Technology. 2021, 43(10):2843-2850.]. 150 waveform units were randomly and uniformly selected from the simulation waveform library to generate MFR pulse trains for experiments. Dot matrix diagrams based on PRI, RF, and PW were constructed respectively, as shown below. Figure 5 As shown.

[0120] The confusion matrix is ​​calculated based on the corner detection results. The four values ​​of the confusion matrix corresponding to the waveform unit extraction are: True Positive (TP), indicating that the sample is the starting pulse of the waveform unit and is marked as such; False Positive (FP), indicating that the sample is an internal pulse of the waveform unit but is marked as the starting pulse; True Negative (TN), indicating that the sample is an internal pulse of the waveform unit and is marked as such; and False Negative (FN), indicating that the sample is the starting pulse of the waveform unit but is marked as such. The performance of the waveform unit extraction algorithm is measured using relevant evaluation metrics based on the confusion matrix, defined as follows.

[0121]

[0122] TPR (True Positive Rate) measures the algorithm's ability to detect the start pulse of a waveform unit, FPR (False Positive Rate) measures the false alarm rate of misclassifying non-start pulses as start pulses, and F-score comprehensively evaluates the ability of the recognition result to approximate the actual data type label. Therefore, the performance of this algorithm can be evaluated based on TPR, FPR, and F-score.

[0123] 1. Comparative Experiment on Waveform Unit Extraction Effect

[0124] To verify the performance improvement of the waveform unit extraction algorithm based on the dot matrix model proposed in this invention, 500 Monte Carlo simulations were performed on the DA-DBSCAN algorithm proposed in Reference 1 and the method of this invention, respectively. The results are shown in Table 1.

[0125] Table 1 Performance Comparison of Different Waveform Unit Extraction Algorithms

[0126]

[0127]

[0128] As shown in Table 1, compared with Reference 1, the method proposed in this invention improves the TPR index by 1.15%, thus having a better detection capability for the start point of waveform units. It reduces the FPR index by 49.00%, greatly reducing the false alarm rate of waveform unit start point detection. It also improves the F-score index by 5.22%, demonstrating higher detection performance.

[0129] 2. The impact of pulse loss and measurement error on extraction results

[0130] To verify the adaptability of this algorithm in harsh real-world reconnaissance environments, this experiment involved adding zero-mean and σ values ​​to the PRI, RF, and PW values ​​of the MFR pulse train. 2 Gaussian white noise is used to simulate measurement error; pulse loss is simulated by randomly deleting parameter values ​​from the MFR pulse train. The Ratio of Dropped Pulses (RDP) is defined as follows:

[0131]

[0132] In the formula, N i n represents the number of pulses contained in the sequence of PRI, RF, and PW parameters in the MFR pulse train; i The number of pulses randomly selected.

[0133] For the PRI value, σ 2 The value of σ ranges from 0 to 4 μs. For the RF value, σ 2 The value ranges from 0 to 4 MHz. For the PW value, σ 2 The values ​​range from 0 to 4 μs, and the values ​​of RDP range from 0% to 16%. The results of 500 Monte Carlo simulations are shown in Table 2. Figure 6 , Figure 7 As shown.

[0134] Table 2 Different σ 2 Extraction performance under combined action with RDP

[0135]

[0136]

[0137] Figure 6 (a) Figure 7 (a) This visually illustrates the vertical variation of TPR and FPR values ​​in Table 2. It can be seen that the slope of the TPR curve is relatively small, indicating that σ 2 The effect on TPR is not significant, while the slope of the FPR curve is relatively large, indicating that σ 2The FPR is significantly affected, resulting in a noticeable decrease in the F value. This is because when the measurement noise is high, a pattern on the main diagonal of the dot matrix may be divided into two, and two starting points will be detected in the same waveform unit, leading to an increase in the false alarm rate.

[0138] Figure 6 (b) Figure 7 (b) The horizontal variation pattern of TPR and FPR values ​​in Table 2 is intuitively shown. It can be seen that the slope of the TPR and FPR variation curves is not large, indicating that RDP has little effect on TPR and FPR. The decrease in F value is not obvious. This is because the loss of pulse does not change the structural features of the pattern on the main diagonal of the dot matrix, but only changes the size of the pattern, which has little impact on the detection effect of the subsequent Harris corner detection operator.

[0139] In summary, this invention addresses the problem of MFR waveform unit extraction. First, an MFR pulse train dot matrix model is constructed to visually display the modulation rules and starting pulse positions of the waveform units. Then, image preprocessing is used to eliminate noise. Finally, the Harris corner detection operator is used to detect the starting pulse positions of the waveform units, and data fusion is employed to effectively extract the MFR pulse train waveform units. Simulation experiments show that the waveform unit extraction algorithm proposed in this invention significantly improves detection performance compared to traditional clustering waveform unit extraction algorithms, and it can better adapt to pulse loss and measurement errors when conducting actual reconnaissance of non-cooperative MFRs.

[0140] Based on the above embodiments, such as Figure 8 As shown, another aspect of the present invention proposes a multifunctional radar waveform unit extraction system based on a dot matrix model, comprising:

[0141] The dot matrix construction module is used to construct a dot matrix D based on the corresponding pulse repetition interval (PRI), pulse width (PW), and radio frequency (RF) values ​​for a multi-function radar MFR pulse sequence. PRI D RF D PW ;

[0142] The image preprocessing module is used to process the constructed raster image D. PRI D RF D PW Image preprocessing is performed separately, followed by image hole filling and image closing operations;

[0143] The corner detection module is used to detect L-shaped corners on the main diagonal of the preprocessed dot matrix image using the Harris corner detection operator, i.e., the starting points of the waveform units.

[0144] The module for obtaining the initial pulse set is used to take the union of the detection results of the corner detection module, thus obtaining the initial pulse set of the waveform unit.

[0145] Furthermore, the dot matrix construction module is specifically used for:

[0146] For an MFR pulse train containing n pulses, P = {P1, P2, ..., P...} n The PRI, RF, PW values ​​of each pulse contained therein are represented by the following sequence:

[0147]

[0148] In the formula, x i Let y represent the PRI value of the i-th pulse in pulse train P, where i = 1, 2, ..., n. i z represents the RF value of the i-th pulse. i This represents the PW value of the i-th pulse;

[0149] Improving the handling of pulse pairs with identical signal parameter values ​​by setting a threshold:

[0150]

[0151] In the formula, Indicates the pulse pair P i ,P j The PRI value matching situation, Threshold PRI Indicates the pulse pair P i ,P j The PRI values ​​are judged to be the same threshold; Indicates the pulse pair P i ,P j RF value matching status, Threshold RF Indicates the pulse pair P i ,P j The RF values ​​were judged to be the same threshold; Indicates the pulse pair P i ,P j PW value matching status, Threshold PW Indicates the pulse pair P i ,P j The PW values ​​were judged to be the same threshold;

[0152] Therefore, an N×N dot matrix D can be constructed from the PRI, PW, and RF values ​​respectively. PRI D RF D PW The expression is as follows:

[0153]

[0154] Furthermore, in the image preprocessing module, the image hole filling adopts a seed filling algorithm, and a 3×3 convolution kernel is selected for image closing operation.

[0155] Furthermore, the corner detection module is specifically used for:

[0156] For the pixel (i,i) located on the main diagonal in the preprocessed dot matrix image, calculate its corner response function value (CRF). If the CRF of the point is greater than 0, the point can be determined to be a corner point, i.e., the starting point of the waveform unit. The position number i corresponds to the starting pulse number of the waveform unit in the MFR pulse train.

[0157] In summary, this invention introduces biological sequence analysis technology, utilizing the repeatability of pulses within the same waveform unit and the differences in pulses between different waveform units to reconstruct a one-dimensional MFR waveform unit into a two-dimensional dot matrix model. A corner detection algorithm is then used to identify the starting point of the waveform unit. This invention does not rely on clustering parameter settings and can fully utilize the overall modulation pattern of the waveform unit, improving the detection probability of the waveform unit while reducing the false alarm probability. It maintains good performance even under conditions of pulse loss and large parameter measurement errors.

[0158] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for extracting multifunctional radar waveform units based on a dot matrix model, characterized in that, include: Step 1: For the multi-function radar MFR pulse sequence, construct a dot matrix based on the corresponding pulse repetition interval (PRI), pulse width (PW), and radio frequency (RF) values, respectively. ;include: For a containing MFR pulse train of 1 pulse The PRI, RF, and PW values ​​of each pulse contained therein are represented by the following sequence: （3） In the formula, Indicates pulse train No. The PRI value of each pulse. , Indicates the first RF value of each pulse Indicates the first The PW value of each pulse; Improving the handling of pulse pairs with identical signal parameter values ​​by setting a threshold: （4） In the formula, Indicates pulse pair The matching status of PRI values, Indicates pulse pair The PRI values ​​are judged to be the same threshold; Indicates pulse pair RF value matching status, Indicates pulse pair The RF values ​​were judged to be the same threshold; Indicates pulse pair PW value matching status, Indicates pulse pair The PW values ​​were judged to be the same threshold; Set to 6~10us, Set to 6~10Hz, Set to 1~3us; Step 2: Process the constructed raster image Image preprocessing is performed separately, followed by image hole filling and image closing operations. This includes: filling the holes inside solid squares, diagonal squares, and solid noise blocks on the line segments on the main diagonal of the bitmap; and using a 3×3 convolution kernel to perform image closing operations to fill the pattern holes on the main diagonal of the MFR pulse sequence bitmap and repair the pattern boundaries. Step 3: Use the Harris corner detection operator to detect the L-shaped corners on the main diagonal of the preprocessed bitmap, i.e., the starting points of the waveform units; specifically, this includes: detecting the pixels located on the main diagonal of the preprocessed bitmap. Calculate its corner response function value If the point This allows us to determine that the point is a corner point, i.e., the starting point of the waveform unit, and the position number is [not specified]. The starting pulse number of the waveform unit in the corresponding MFR pulse train; Step 4: Take the union of the detection results from Step 3 to obtain the set of initial pulses for the waveform unit.

2. The method for extracting multifunctional radar waveform units based on a dot matrix model according to claim 1, characterized in that, Step 1 includes: Constructed from PRI, PW, and RF values ​​respectively Raster graphics of different sizes The expression is as follows: （5）。 3. The method for extracting multifunctional radar waveform units based on a dot matrix model according to claim 1, characterized in that, In step 2, the image hole filling uses a seed filling algorithm.

4. A multifunctional radar waveform unit extraction system based on a dot matrix model, characterized in that, include: The dot matrix construction module is used to construct dot matrix patterns based on the corresponding pulse repetition interval (PRI), pulse width (PW), and radio frequency (RF) values ​​for a multi-function radar MFR pulse sequence. Specifically used for: For a containing MFR pulse train of 1 pulse The PRI, RF, and PW values ​​of each pulse contained therein are represented by the following sequence: （3） In the formula, Indicates pulse train No. The PRI value of each pulse. , Indicates the first RF value of each pulse Indicates the first The PW value of each pulse; Improving the handling of pulse pairs with identical signal parameter values ​​by setting a threshold: （4） In the formula, Indicates pulse pair The matching status of PRI values, Indicates pulse pair The PRI values ​​are judged to be the same threshold; Indicates pulse pair RF value matching status, Indicates pulse pair The RF values ​​were judged to be the same threshold; Indicates pulse pair PW value matching status, Indicates pulse pair The PW values ​​were judged to be the same threshold; Set to 6~10us, Set to 6~10Hz, Set to 1~3us; The image preprocessing module is used to process the constructed raster image. Image preprocessing is performed separately, followed by image hole filling and image closing operations. Specifically, this is used to fill image holes in solid squares, diagonal squares, and solid noise blocks on line segments on the main diagonal of the bitmap; and to perform image closing operations using a 3×3 convolution kernel to fill pattern holes on the main diagonal of the MFR pulse sequence bitmap and repair pattern boundaries. The corner detection module uses the Harris corner detection operator to detect L-shaped corners on the main diagonal of the preprocessed bitmap, i.e., the starting points of waveform units; specifically, it includes: detecting pixels located on the main diagonal of the preprocessed bitmap. Calculate its corner response function value If the point This allows us to determine that the point is a corner point, i.e., the starting point of the waveform unit, and the position number is [not specified]. The starting pulse number of the waveform unit in the corresponding MFR pulse train; The module for obtaining the initial pulse set is used to take the union of the detection results of the corner detection module, thus obtaining the initial pulse set of the waveform unit.

5. The multifunctional radar waveform unit extraction system based on a dot matrix model according to claim 4, characterized in that, The dot matrix diagram construction module is specifically used for: Constructed from PRI, PW, and RF values ​​respectively Raster graphics of different sizes The expression is as follows: （5）。 6. The multifunctional radar waveform unit extraction system based on a dot matrix model according to claim 4, characterized in that, In the image preprocessing module, the image hole filling uses a seed filling algorithm.

Images