Marine target detection method and system, electronic equipment and storage medium

Abstract

The application provides a marine target detection method and system, electronic equipment and a storage medium. The method comprises the following steps: acquiring radar echo detection data; extracting feature points of the radar echo detection data to form a feature vector; matching the feature vector with a decision region in a feature space to detect whether a target exists in the radar echo detection data, and the decision region in the feature space is constructed based on a feature combination strategy generated according to sample features. Since the decision region is constructed according to the feature combination strategy, the matching of the feature points in the feature sequence with the decision region in the feature space can improve the overall separability of the to-be-detected radar echo data in the feature space, thereby improving the marine small target detection performance.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing technology, and in particular to a method, system, electronic device, and storage medium for detecting maritime targets. Background Technology

[0002] Sea clutter is the radar echo received by radar and reflected from the sea surface. Maritime search radars are inevitably affected by sea clutter when detecting small floating targets such as small boats, ice floes, and debris. The intensity of sea clutter varies depending on radar parameters, radar illumination direction, and sea state. In the field of radar-based maritime target detection, target detection is mainly performed using energy-based methods. In low sea states and scenarios with a large target radar cross-section (RCS), the signal-to-clutter ratio (SCR) of the radar signal is high, resulting in generally good detection performance. However, in high sea states and with a small target RCS, the SCR decreases significantly, leading to a severe decline in detection performance. To address the problem of small target detection in sea clutter under low SCR conditions, feature detection methods are typically employed. Feature extraction is the core technology of these methods. This approach starts from the characteristics of marine targets and sea clutter, mining the differential features between marine targets and sea clutter in different transform domains. This includes extracting differential features in multiple domains such as the time domain, frequency domain, time-frequency domain, and polarization domain. Currently, the total number of features extracted from different representation domains has reached more than 30. The dimension of the feature vector constructed using multi-domain features has also been extended from one dimension to three, seven, or even higher dimensions, forming a high-dimensional feature space. By increasing the feature dimension, multiple features can complement each other for different types of targets, enabling the identification of marine targets and sea clutter within the feature space.

[0003] Currently, mainstream feature extraction methods fuse features based on second-order correlation to remove redundancy. That is, if two features have a strong second-order correlation, they are considered redundant, and a redundant feature needs to be removed when constructing a high-dimensional feature space. However, when constructing two-dimensional or even high-dimensional feature spaces from one-dimensional feature spaces, the core of target detection lies in separating two types of samples (sea clutter and targets) in the feature space. The impact of the correlation between different features on overall separability is uncertain, leading to the reliance on manual experience to remove redundant features during feature extraction, without improving detection performance. Summary of the Invention

[0004] This invention provides a method, system, electronic device, and storage medium for detecting maritime targets, which addresses the shortcomings of existing technologies where redundant features are mainly removed through manual experience during feature extraction, without improving detection performance.

[0005] This invention provides a method for detecting maritime targets, comprising:

[0006] Acquire radar echo detection data;

[0007] Extract the feature points from the radar echo detection data to form a feature vector;

[0008] The feature vector is matched with the decision region in the feature space to detect whether there is a target in the radar echo detection data. The decision region in the feature space is constructed based on a feature combination strategy generated from sample features.

[0009] According to the present invention, a method for detecting maritime targets, wherein the decision region in the feature space is constructed based on a feature combination strategy generated from sample features, including:

[0010] Features are extracted from radar echo sample data, which includes sea clutter sample data and target sample data, and the features are combined in pairs to obtain multiple feature pairs.

[0011] Based on different feature pair combinations and feature value magnitudes, the distribution of sea clutter sample features and target sample features in the feature space is classified to obtain multiple feature combinations. The horizontal and vertical coordinates of the feature space are used to represent the magnitudes of the two features in the feature pair, respectively.

[0012] Second-order correlation impact analysis was conducted on different feature combinations to obtain feature combination strategies;

[0013] The region of the sea clutter sample feature in the feature space is obtained according to the feature combination strategy and used as the decision region.

[0014] According to a maritime target detection method provided by the present invention, the distribution of sea clutter sample features and target sample features in the feature space is classified based on different feature combination cases and feature value magnitudes to obtain multiple feature combinations, including:

[0015] When both feature values ​​of the target sample feature are less than both feature values ​​of the sea clutter sample feature, or when both feature values ​​of the target sample feature are greater than both feature values ​​of the sea clutter sample feature, a first feature combination is obtained;

[0016] When one feature value of the target sample feature is less than the feature value of the sea clutter sample feature, and at the same time, another feature value of the target sample feature is greater than the feature value of the sea clutter sample feature, a second feature combination is obtained.

[0017] According to the second-order correlation influence analysis of different feature combinations provided by the present invention, a feature combination strategy is obtained, including:

[0018] Calculate the second-order correlation coefficient between the two features in the feature pair, and obtain the second-order correlation based on the second-order correlation coefficient. The second-order correlation includes at least one of positive correlation, negative correlation, and weak correlation.

[0019] Based on the second-order correlation coefficient and the distribution of the target sample features and the sea clutter sample features in the feature space for each feature combination, a feature combination strategy is obtained.

[0020] The present invention provides a method for detecting maritime targets, wherein the feature combination strategy includes:

[0021] When the feature combination is the first feature combination, the target sample feature points and the sea clutter sample feature points are distributed in the lower left and upper right regions of the feature space. Feature combinations with positive second-order correlation are eliminated, and feature combinations with negative second-order correlation are selected. If the feature value range of the target sample feature points and the sea clutter sample feature points in the feature space is less than a preset threshold, feature combinations with weak second-order correlation are selected.

[0022] When the feature combination is the second feature combination, the target sample feature points and the sea clutter sample feature points are distributed in the upper left and lower right regions of the feature space. Feature combinations with negative second-order correlation are eliminated, and feature combinations with positive second-order correlation are selected. If the feature value range of the target sample feature points and the sea clutter sample feature points in the feature space is less than a preset threshold, feature combinations with weak second-order correlation are selected.

[0023] According to a maritime target detection method provided by the present invention, the step of calculating the second-order correlation coefficient of two features in the feature pair and obtaining the second-order correlation based on the second-order correlation coefficient includes:

[0024] Calculate the second-order correlation coefficient:

[0025]

[0026] Wherein, X is the first feature sequence of the sea clutter sample corresponding to one feature in the feature pair, Y is the second feature sequence of the sea clutter sample corresponding to the other feature in the feature pair, Cov(X,Y) represents the covariance between the two feature sequences, and D(X) and D(Y) represent the variances of the two feature sequences, respectively.

[0027] When the second-order correlation coefficient is positive and greater than a preset first threshold, the feature is positively correlated with the second-order correlation.

[0028] When the second-order correlation coefficient is negative and less than a preset second threshold value, the feature is negatively correlated with the second-order correlation.

[0029] When the second-order correlation coefficient is between the first threshold and the second threshold, the feature is weakly correlated with the second-order correlation.

[0030] According to a maritime target detection method provided by the present invention, the step of matching the feature vector with a decision region in the feature space to detect whether a target exists in the radar echo detection data includes:

[0031] The feature vector is matched with the sea clutter decision region in the feature space. If the feature vector is not in the sea clutter decision region, it is determined that there is a target in the radar echo detection data. If the feature vector is in the sea clutter decision region, it is determined that there is no target in the radar echo detection data.

[0032] The present invention also provides a maritime target detection system, comprising:

[0033] The acquisition module is used to acquire radar echo detection data;

[0034] The extraction module is used to extract feature points from the radar echo detection data to form a feature vector;

[0035] The detection module is used to match the feature vector with the decision region in the feature space to detect whether there is a target in the radar echo detection data. The decision region in the feature space is constructed based on a feature combination strategy generated from sample features.

[0036] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the maritime target detection method described in any of the preceding claims.

[0037] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the maritime target detection method described in any of the preceding claims.

[0038] This invention provides a method, system, electronic device, and storage medium for detecting maritime targets. The method acquires radar echo detection data; extracts feature points from the radar echo detection data to form feature vectors; and matches the feature vectors with decision regions in a feature space to detect the presence of targets in the radar echo detection data. The decision regions in the feature space are constructed based on a feature combination strategy generated from sample features. Because the decision regions are constructed based on the feature combination strategy, matching feature points in the feature sequence with decision regions in the feature space improves the overall separability of the radar echo data under test in the feature space, thereby improving the performance of small maritime target detection. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0040] Figure 1 This is one of the flowcharts of the maritime target detection method provided by the present invention;

[0041] Figure 2 This is the second flowchart of the maritime target detection method provided by the present invention;

[0042] Figure 3 This is the third flowchart of the maritime target detection method provided by the present invention;

[0043] Figure 4 This is a feature distribution map of the first feature combination that is negatively correlated, provided by the present invention;

[0044] Figure 5 This is a feature distribution map of the first feature combination that is positively correlated, provided by the present invention;

[0045] Figure 6 This is a feature distribution map of the positively correlated second feature combination provided by the present invention;

[0046] Figure 7 This is a feature distribution map of the negatively correlated second feature combination provided by the present invention;

[0047] Figure 8 This is a feature distribution map of different combinations of polarization and correlation features provided by the present invention;

[0048] Figure 9 This is a schematic diagram of the structure of the maritime target detection system provided by the present invention;

[0049] Figure 10 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0051] Figure 1A flowchart of the maritime target detection method provided in the embodiments of the present invention is shown below. Figure 1 As shown, the maritime target detection method provided in this embodiment of the invention includes:

[0052] Step 101: Acquire radar echo detection data;

[0053] Step 102: Extract feature points from radar echo detection data to form feature vectors;

[0054] In this embodiment of the invention, the extracted features originate from multiple representation domains of the echo signal, such as the time domain, frequency domain, time-frequency domain, dual-frequency domain, fractional Fourier transform domain, and polarization domain. These features collectively constitute the feature set. The following provides specific extraction methods for several typical features:

[0055] (1) Relative Peak Height (RPH)

[0056] This feature is defined as the ratio of the peak value of the time-domain wave of the unit under test to the average amplitude of the reference unit. It is used to reflect the difference in the proportion of the peak values ​​of the target and sea clutter signals, as well as the degree of peak fluctuation. The expression is as follows:

[0057]

[0058] In this scenario, if the radar receives multiple consecutive pulses within a range cell, these pulses constitute a target cell. Simultaneously, the observation vectors of P reference cells surrounding the target cell are obtained, thus constructing a target detection problem. m ,x p These represent the echo signals of the element under test and the reference element, respectively, where P represents the number of reference elements, and MAX(x) represents the echo signal of the element under test and the reference element, respectively. m The denominator represents the peak value of the echo from the unit under test, and the denominator represents the average amplitude of the reference unit.

[0059] (2) Time domain entropy mean (TEM)

[0060] This feature is defined as the average value of the time-domain information entropy, used to reflect the degree of disorder in the waveforms of the target echo and sea clutter signals. Because the echo fluctuations differ, the degree of disorder in the echo sequences may vary. The TEM feature calculation method is as follows.

[0061] Let the time-domain echo signal x of length L be represented as x = [x1, x2, x3, x4, ..., x L Define a rectangular window function of length W, slide it through the time-domain echo signal x, and obtain L+W-1 signals of length W:

[0062] s i =[xi ,...,x L+W-1 ], i = 1, 2, ..., L + W - 1

[0063] Calculate signal s i The temporal entropy value is as follows:

[0064]

[0065] Where, p f (n) represents the normalized magnitude, as follows:

[0066]

[0067] The mean value of the time-domain entropy is calculated as follows:

[0068]

[0069] (3) Relative Doppler Peak Height (RDPH)

[0070] This feature is defined as the ratio of the Doppler peak value of the target cell to the average Doppler peak value of the reference cell. It reflects the energy magnitude of different frequency components of the signal, allowing comparison of the difference in peak abrupt changes between the target and sea clutter in the Doppler domain. The RDPH feature is calculated as follows:

[0071]

[0072] in,

[0073]

[0074] γ = [-δ1,δ2]∪[δ2,δ1]

[0075] In the formula, DAS represents the Doppler amplitude spectrum, f d Indicates the Doppler frequency point. The Doppler frequency that maximizes the DAS amplitude is δ1, which is determined by the average Doppler bandwidth of the sea clutter, and δ2 is determined by the guard interval of the Doppler peak. For example, δ1 = 50Hz and δ2 = 5Hz can be taken. #γ represents the number of Doppler elements in the interval γ, and P represents the number of reference elements.

[0076] (4) Relative Vector Entropy (RVE)

[0077] This feature is defined as the ratio of the information entropy of the unit under test to the information entropy of the reference unit. It reflects the degree of disorder in the signal waveform. Compared with sea clutter, the target and sea clutter have different echo energy intensities and fluctuations, so the degree of disorder in the echo sequence may differ. The RVE feature is calculated as follows:

[0078]

[0079] Wherein, VE(x) m ) represents the Doppler vector entropy, as follows:

[0080]

[0081]

[0082] In the formula, This represents the normalized Doppler spectrum.

[0083] Step 103: Match the feature vector with the decision region in the feature space to detect whether there is a target in the radar echo detection data. The decision region in the feature space is constructed based on the feature combination strategy generated from the sample features.

[0084] Currently, mainstream feature extraction methods fuse features based on second-order correlation to remove redundancy. That is, if two features have a strong second-order correlation, they are considered redundant, and a redundant feature needs to be removed when constructing a high-dimensional feature space. However, when constructing two-dimensional or even high-dimensional feature spaces from one-dimensional feature spaces, the core of target detection lies in separating two types of samples (sea clutter and targets) in the feature space. The impact of the correlation between different features on overall separability is uncertain, leading to the reliance on manual experience to remove redundant features during feature extraction, without improving detection performance.

[0085] This invention provides a method for detecting maritime targets. The method involves acquiring radar echo detection data; extracting feature points from the radar echo detection data to form feature vectors; and matching these feature vectors with decision regions in a feature space to detect the presence of targets in the radar echo detection data. The decision regions in the feature space are constructed based on a feature combination strategy generated from sample features. Because the decision regions are constructed based on the feature combination strategy, matching feature points in the feature sequence with decision regions in the feature space improves the overall separability of the radar echo data under test in the feature space, thereby enhancing the performance of small maritime target detection.

[0086] Based on any of the above embodiments, such as Figure 2 As shown, the method for constructing the decision region in the feature space in the above steps includes:

[0087] Step 201: Extract features from radar echo sample data. Radar echo sample data includes sea clutter sample data and target sample data. Combine the features in pairs to obtain multiple feature pairs.

[0088] Step 202: Based on the different combinations of features and the magnitude of feature values, classify the distribution of sea clutter sample features and target sample features in the feature space to obtain multiple feature combinations. The horizontal and vertical coordinates of the feature space are used to represent the magnitudes of the two features in the feature pair, respectively.

[0089] In this embodiment of the invention, the distribution of sea clutter sample features and target sample features in the feature space is classified according to different feature combination cases and feature value magnitudes, resulting in multiple feature combinations, including:

[0090] The first feature combination is obtained when both feature values ​​of the target sample feature are less than the two feature values ​​of the sea clutter sample feature, or when both feature values ​​of the target sample feature are greater than the two feature values ​​of the sea clutter sample feature.

[0091] When one feature value of the target sample feature is less than the feature value of the sea clutter sample feature, and at the same time, another feature value of the target sample feature is greater than the feature value of the sea clutter sample feature, a second feature combination is obtained.

[0092] In this embodiment of the invention, the number of features in a feature combination is not limited. For example, if three features are to be combined, the feature types of the three combinations are determined sequentially by combining each feature in pairs, using a two-dimensional feature combination type. Taking two-dimensional features as an example, for different feature combinations, since the relative magnitudes of the target and clutter feature values ​​differ, based on this difference, the two-dimensional feature combinations are divided into four different combination types. The specific combination methods are as follows:

[0093] The first combination method: the eigenvalues ​​of the two features of the target sample are both smaller than the eigenvalues ​​of the two features of the sea clutter sample;

[0094] The second combination method is: the feature value of feature 1 of the target sample is less than the feature value of feature 1 of the sea clutter sample, and at the same time, the feature value of feature 2 of the target sample is greater than the feature value of feature 2 of the sea clutter sample.

[0095] The third combination method: the feature values ​​of both features of the target sample are greater than the feature values ​​of both features of the sea clutter sample;

[0096] The fourth combination method: the feature value of feature 1 of the target sample is greater than the feature value of feature 1 of the sea clutter sample, while the feature value of feature 2 of the target sample is less than the feature value of feature 2 of the sea clutter sample.

[0097] Among them, the third combination is equivalent to the first combination, and the first and third combinations constitute the first feature combination; the fourth combination is equivalent to the second combination, and the second and fourth combinations constitute the second feature combination.

[0098] For the first feature combination, in the two-dimensional feature space, the target echo feature points and sea clutter feature points are located in the lower left and upper right corners of the two-dimensional coordinate system, respectively.

[0099] For the second feature combination, in the two-dimensional feature space, the target echo feature points and sea clutter feature points are located in the lower right and upper left corners of the two-dimensional coordinate system, respectively.

[0100] Step 203: Perform second-order correlation impact analysis on different feature combinations to obtain feature combination strategies;

[0101] Step 204: Obtain the region of sea clutter sample features in the feature space according to the feature combination strategy, and use it as the decision region.

[0102] In this embodiment of the invention, a feature combination strategy is formed by the correlation effect of feature combinations. Under the three conditions of positive correlation, negative correlation and weak correlation, feature combination strategies are formed under two combination types according to the form of two-dimensional characteristics in the feature space. Two-dimensional features are selected according to the feature combination strategy, and the selected two-dimensional features are used to construct the feature space to complete the target detection.

[0103] Based on any of the above embodiments, such as Figure 3 As shown, the above steps involve performing second-order correlation impact analysis on different feature combinations to obtain feature combination strategies, including:

[0104] Step 301: Calculate the second-order correlation coefficient between the two features in the feature pair, and obtain the second-order correlation based on the second-order correlation coefficient. The second-order correlation includes at least one of positive correlation, negative correlation and weak correlation.

[0105] In this embodiment of the invention, using historically observed sea clutter data, the second-order correlation of the acquired feature sequences is calculated iteratively for different feature combinations in a pairwise manner. The first feature sequence is denoted as X, and the second feature sequence as Y. For example, X can represent any one of the feature sequences such as relative peak height, mean temporal entropy value, relative Doppler peak height, and relative Doppler vector entropy, while Y can represent a different feature sequence than X. In this embodiment of the invention, no restrictions are placed on the features.

[0106] Calculate the second-order correlation coefficient between the two features in a feature pair, and obtain the second-order correlation based on the second-order correlation coefficient, including:

[0107] Calculate the second-order correlation coefficient:

[0108]

[0109] Where X is the first feature sequence of the sea clutter sample corresponding to one feature in the feature pair, Y is the second feature sequence of the sea clutter sample corresponding to the other feature in the feature pair, Cov(X,Y) represents the covariance between the two feature sequences, and D(X) and D(Y) represent the variances of the two feature sequences, respectively; Cov(X,Y) is calculated as follows:

[0110] Cov(X,Y)=E[(XE(X))(YE(Y))]

[0111] E(X) and E(Y) represent the mathematical expectations of the two feature sequences, respectively, and are calculated as follows:

[0112]

[0113]

[0114] The calculation methods for D(X) and D(Y) are as follows:

[0115]

[0116]

[0117] Where X(i) and Y(i) represent the i-th feature in the two feature sequences, i = 1, 2, ..., N, and N represents the number of features in each feature sequence.

[0118] By iterating through the data, the second-order correlation coefficients of any pairwise combination of features can be obtained.

[0119] In this embodiment of the invention, taking two-dimensional features as an example, in two-dimensional feature combination detection, when the correlation between features is strong, the two features can be linearly represented by each other. At this time, the feature points tend to converge towards the linear regression line. The stronger the correlation, the greater the degree of convergence towards the regression line; conversely, the weaker the correlation, the smaller the degree of convergence. Therefore, under the four combination methods, the distribution pattern of two-dimensional features in the feature space will exhibit the following regular characteristics:

[0120] When two-dimensional features are positively correlated, if one feature value increases, the other feature value will also increase. The distribution of feature points is an approximately elliptical feature convex hull region with a positive slope on the major axis, and the convex hull region changes along the major axis.

[0121] When two-dimensional features are negatively correlated, if one feature value increases, the other feature value decreases. The distribution of feature points is an approximately elliptical feature convex hull region with a negative slope of the major axis, and the convex hull region changes along the major axis.

[0122] When the correlation between features is weak, or even when the correlation coefficient is close to 0, the region where feature points are concentrated is approximately circular, irregular in shape, and has a low degree of convergence along a certain straight line.

[0123] In this embodiment of the invention, since there is a certain linear dependence between features, a second-order correlation analysis method is used to statistically calculate the correlation between different features to obtain a correlation coefficient matrix. By calculating the second-order correlation of different feature combinations, three types of feature correlation relationships, namely positive correlation, negative correlation and weak correlation, are obtained, thereby forming the correlation influence law under two feature combination types.

[0124] When the second-order correlation coefficient is positive and greater than the preset first threshold, the second-order correlation of the feature pair is positive.

[0125] When the second-order correlation coefficient is negative and less than the preset second threshold, the feature pair is negatively correlated in the second order.

[0126] When the second-order correlation coefficient is between the first and second thresholds, the feature pair is weakly correlated.

[0127] In this embodiment of the invention, the first threshold value is 0.368 and the second threshold value is -0.368. It should be noted that this application does not impose specific restrictions on the first and second threshold values, and those skilled in the art can choose according to their needs.

[0128] Step 302: Based on the second-order correlation coefficient and the distribution of target sample features and sea clutter sample features in the feature space for each feature combination, obtain the feature combination strategy.

[0129] In this embodiment of the invention, the feature combination strategy includes:

[0130] When the feature combination is the first feature combination, the target sample feature points and the sea clutter sample feature points are distributed in the lower left and upper right regions of the feature space. Feature combinations with positive second-order correlation are eliminated, and feature combinations with negative second-order correlation are selected. If the feature value range of the target sample feature points and the sea clutter sample feature points in the feature space is less than the preset threshold, feature combinations with weak second-order correlation are selected.

[0131] In this embodiment of the invention, when the feature combination is the first feature combination, the target echo and sea clutter feature points are distributed in the lower left and upper right regions of the coordinate system. To reduce the overlap of the feature distribution regions between the target echo and sea clutter, negatively correlated features should be selected for combination and application, so that the changing trends of the target echo and sea clutter feature regions are parallel to each other, thereby reducing the overlap of feature distribution regions. Figure 4As shown; if a positively correlated feature combination is selected, the feature regions of the two types of samples show the same trend. When the long axis of the feature region is long, the feature regions are prone to overlap, such as... Figure 5 As shown.

[0132] When the feature combination is the second feature combination, the target sample feature points and the sea clutter sample feature points are distributed in the upper left and lower right regions of the feature space. Feature combinations with negative second-order correlation are eliminated, and feature combinations with positive second-order correlation are selected. If the feature value range of the target sample feature points and the sea clutter sample feature points in the feature space is less than the preset threshold, feature combinations with weak second-order correlation are selected.

[0133] In this embodiment of the invention, when the feature combination is the second feature combination, the target echo and sea clutter feature points are distributed in the upper left and lower right regions of the coordinate system. To reduce the overlapping area of ​​feature points, positively correlated features can be selected for fusion application, making the changing trends of the feature regions of the two types of samples parallel, thereby reducing overlap. Figure 6 As shown; if a combination of negatively correlated features is selected, the feature regions of the two types of samples change in the same direction. When the range of change is large, overlap is likely to occur, such as... Figure 7 As shown.

[0134] Feature combination strategies are formed by considering the correlation between feature combinations. Under the three conditions of positive correlation, negative correlation, and weak correlation, feature combination strategies are formed under two combination types based on the form of two-dimensional characteristics in the feature space. Two-dimensional features are selected according to the feature combination strategy, and the selected two-dimensional features are used to construct the feature space to complete target detection.

[0135] Based on any of the above embodiments, the step of matching the feature vector with the decision region in the feature space to detect whether a target exists in the radar echo detection data includes:

[0136] The feature vector is matched with the sea clutter decision region in the feature space. If the feature vector is not within the sea clutter decision region, it is determined that there is a target in the radar echo detection data. If the feature vector is within the sea clutter decision region, it is determined that there is no target in the radar echo detection data.

[0137] In this embodiment of the invention, the sea clutter decision region can be obtained by either the convex hull decision region obtained by the convex hull algorithm or the concave hull decision region obtained by the concave hull algorithm.

[0138] This invention provides a method for detecting maritime targets. The method acquires radar echo detection data; extracts feature points from the radar echo detection data to form feature vectors; and matches these feature vectors with decision regions in the feature space to detect the presence of targets in the radar echo detection data. The decision regions in the feature space are constructed based on a feature combination strategy generated from sample features. Since the decision regions are constructed based on the feature combination strategy, matching feature points in the feature sequence with decision regions in the feature space improves the overall separability of the radar echo data under test in the feature space, thereby improving the detection performance of small maritime targets. The feature extraction method provided by this invention, through a feature combination strategy, selects features with second-order correlation based on the feature combination type to reduce the overlap between sea clutter and targets in the feature space, improving overall separability and thus enhancing the detection performance of maritime targets, especially small maritime targets. The main steps include:

[0139] Step 1: Extract features from radar echo data to form a feature sequence. The extracted features come from multiple representation domains of the echo signal, such as time domain, frequency domain, time-frequency domain, dual-frequency domain, and polarization domain.

[0140] Step 2: Classify the distribution of feature sequences of the two types of samples (i.e., sea clutter and targets) in the feature space. Taking two-dimensional features as an example, for different feature combinations, since the relative magnitudes of target and clutter feature values ​​differ, based on this difference, the two-dimensional feature combinations are divided into four different combination types.

[0141] Step 3: Second-order correlation analysis of different feature combinations. Since there may be a certain linear dependence between features, a second-order correlation analysis method is used to statistically calculate the correlation between different features, obtaining a correlation coefficient matrix. Based on the three cases of positive correlation, negative correlation, and weak correlation, the correlation influence patterns under four combination types are formed.

[0142] Step 4: Formulate feature combination strategies based on correlation effects. Under the three conditions of positive correlation, negative correlation, and weak correlation, feature combination strategies are formulated for four combination types based on the form of two-dimensional characteristics in the feature space, and a target detection method applying two-dimensional features is constructed.

[0143] Experimental data shows that this invention helps improve the performance of maritime target detection, specifically as follows: The proposed method is verified using measured data. Using channel buoys as small maritime targets to be detected, the performance of the proposed method is verified under both HH and VV polarization modes.

[0144] The proposed method was validated using measured data. Using channel buoys as the small maritime targets to be detected, the performance of the proposed method was verified under both HH and VV polarization modes.

[0145] The performance of the proposed method is analyzed using echo data collected under sea state 4 conditions. The polarization modes include HH and VV, and the maritime target is a channel buoy, a typical small maritime target. For example, four features—RDPH, RVE, TEM, and RPH—are extracted and combined to obtain the distribution of feature points in the feature space, as shown below. Figure 8 As shown, the RDPH-RVE combination exhibits a negative correlation, while the TEM-RPH combination shows a positive correlation. Both of these combinations are first-feature combinations. According to the feature combination strategy of this application, when the feature combination type is a first-feature combination, two-dimensional features exhibiting negative correlation should be selected for combination application. In other words, the RDPH-RVE combination conforms to the combination strategy, and selecting these two feature combinations when constructing a two-dimensional feature detection method is more conducive to achieving separability. For comparison, the performance of the detection method when applying the TEM-RPH feature combination was also analyzed. Under HH polarization conditions, the detection probability was calculated using the convex hull algorithm. The detection probability of the RVE and RDPH feature combination was 49.54%, and that of the TEM and RPH feature combination was 22.85%. Under VV polarization conditions, the detection probability of the RVE and RDPH feature combination was 97.96%, and that of the TEM and RPH feature combination was 75.79%. Therefore, the feature detection method based on the feature combination strategy proposed in this application, verified by experimental data, can effectively improve the detection performance of small targets at sea.

[0146] The maritime target detection system provided by the present invention is described below. The maritime target detection described below and the maritime target detection method described above can be referred to in correspondence.

[0147] Figure 9 This is a schematic diagram of a maritime target detection system provided in an embodiment of the present invention, such as... Figure 9 As shown, the maritime target detection system provided in this embodiment of the invention includes:

[0148] Acquisition module 901 is used to acquire radar echo detection data;

[0149] Extraction module 902 is used to extract feature points from radar echo detection data to form feature vectors;

[0150] The detection module 903 is used to match the feature vector with the decision region in the feature space to detect whether there is a target in the radar echo detection data. The decision region in the feature space is constructed based on the feature combination strategy generated according to the sample features.

[0151] The maritime target detection system provided in this embodiment of the invention improves the overall separability of radar echo data under test in the feature space through the detection module, thereby improving the detection performance of small maritime targets.

[0152] Figure 10 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 10 As shown, the electronic device may include a processor 1010, a communications interface 1020, a memory 1030, and a communication bus 1040. The processor 1010, communications interface 1020, and memory 1030 communicate with each other via the communication bus 1040. The processor 1010 can call logical instructions in the memory 1030 to execute a maritime target detection method. This method includes: acquiring radar echo detection data; extracting feature points from the radar echo detection data to form a feature vector; and matching the feature vector with a decision region in the feature space to detect whether a target exists in the radar echo detection data. The decision region in the feature space is constructed based on a feature combination strategy generated from sample features.

[0153] Furthermore, the logical instructions in the aforementioned memory 1030 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0154] On the other hand, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the maritime target detection method provided by the above methods. The method includes: acquiring radar echo detection data; extracting feature points from the radar echo detection data to form a feature vector; and matching the feature vector with a decision region in the feature space to detect whether a target exists in the radar echo detection data. The decision region in the feature space is constructed based on a feature combination strategy generated according to sample features.

[0155] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0156] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting maritime targets, characterized in that, include: Acquire radar echo detection data; Extract the feature points from the radar echo detection data to form a feature vector; The feature vector is matched with the decision region in the feature space to detect whether there is a target in the radar echo detection data. The decision region in the feature space is constructed based on the feature combination strategy generated according to the sample features. The decision region in the feature space is constructed based on a feature combination strategy generated from sample features, including: extracting features from radar echo sample data, which includes sea clutter sample data and target sample data, and combining the features in pairs to obtain multiple feature pairs; classifying the distribution of sea clutter sample features and target sample features in the feature space according to different feature pair combinations and feature value magnitudes to obtain multiple feature combinations, wherein the horizontal and vertical coordinates of the feature space are used to represent the magnitudes of the two features in the feature pairs; performing second-order correlation influence analysis on different feature combinations to obtain a feature combination strategy; and obtaining the region of the sea clutter sample features in the feature space according to the feature combination strategy, which serves as the decision region.

2. The method for detecting maritime targets according to claim 1, characterized in that, The distribution of sea clutter sample features and target sample features in the feature space is classified according to different feature combinations and feature value magnitudes, resulting in multiple feature combinations, including: When both feature values ​​of the target sample feature are less than both feature values ​​of the sea clutter sample feature, or when both feature values ​​of the target sample feature are greater than both feature values ​​of the sea clutter sample feature, a first feature combination is obtained. When one feature value of the target sample feature is less than the feature value of the sea clutter sample feature, and at the same time, another feature value of the target sample feature is greater than the feature value of the sea clutter sample feature, a second feature combination is obtained.

3. The method for detecting maritime targets according to claim 2, characterized in that, The second-order correlation impact analysis of different feature combinations yields feature combination strategies, including: Calculate the second-order correlation coefficient between the two features in the feature pair, and obtain the second-order correlation based on the second-order correlation coefficient. The second-order correlation includes at least one of positive correlation, negative correlation, and weak correlation. Based on the second-order correlation coefficient and the distribution of the target sample features and the sea clutter sample features in the feature space for each feature combination, a feature combination strategy is obtained.

4. The method for detecting maritime targets according to claim 3, characterized in that, The feature combination strategy includes: When the feature combination is the first feature combination, the target sample feature points and the sea clutter sample feature points are distributed in the lower left and upper right regions of the feature space. Feature combinations with positive second-order correlation are eliminated, and feature combinations with negative second-order correlation are selected. If the feature value range of the target sample feature points and the sea clutter sample feature points in the feature space is less than a preset threshold, feature combinations with weak second-order correlation are selected. When the feature combination is the second feature combination, the target sample feature points and the sea clutter sample feature points are distributed in the upper left and lower right regions of the feature space. Feature combinations with negative second-order correlation are eliminated, and feature combinations with positive second-order correlation are selected. If the feature value range of the target sample feature points and the sea clutter sample feature points in the feature space is less than a preset threshold, feature combinations with weak second-order correlation are selected.

5. The method for detecting maritime targets according to claim 3, characterized in that, The calculation of the second-order correlation coefficient between the two features in the feature pair, and the determination of the second-order correlation based on the second-order correlation coefficient, includes: Calculate the second-order correlation coefficient: Wherein, X is the first feature sequence of the sea clutter sample corresponding to one feature in the feature pair, and Y is the second feature sequence of the sea clutter sample corresponding to the other feature in the feature pair. This represents the covariance between two feature sequences. and These represent the variances of the two feature sequences, respectively. When the second-order correlation coefficient is positive and greater than a preset first threshold, the feature is positively correlated with the second-order correlation. When the second-order correlation coefficient is negative and less than a preset second threshold value, the feature is negatively correlated with the second-order correlation. When the second-order correlation coefficient is between the first threshold and the second threshold, the feature is weakly correlated with the second-order correlation.

6. The method for detecting maritime targets according to claim 1, characterized in that, The step of matching the feature vector with a decision region in the feature space to detect whether a target exists in the radar echo detection data includes: The feature vector is matched with the sea clutter decision region in the feature space. If the feature vector is not in the sea clutter decision region, it is determined that there is a target in the radar echo detection data. If the feature vector is in the sea clutter decision region, it is determined that there is no target in the radar echo detection data.

7. A maritime target detection system, characterized in that, include: The acquisition module is used to acquire radar echo detection data; The extraction module is used to extract feature points from the radar echo detection data to form a feature vector; The detection module is used to match the feature vector with the decision region in the feature space to detect whether there is a target in the radar echo detection data. The decision region in the feature space is constructed based on the feature combination strategy generated according to the sample features. The decision region in the feature space is constructed based on a feature combination strategy generated from sample features, including: extracting features from radar echo sample data, which includes sea clutter sample data and target sample data, and combining the features in pairs to obtain multiple feature pairs; classifying the distribution of sea clutter sample features and target sample features in the feature space according to different feature pair combinations and feature value magnitudes to obtain multiple feature combinations, wherein the horizontal and vertical coordinates of the feature space are used to represent the magnitudes of the two features in the feature pairs; performing second-order correlation influence analysis on different feature combinations to obtain a feature combination strategy; and obtaining the region of the sea clutter sample features in the feature space according to the feature combination strategy, which serves as the decision region.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the maritime target detection method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the maritime target detection method as described in any one of claims 1 to 6.

Images