Polarimetric radar sea surface target multi-feature intelligent detection method and device

Abstract

The application relates to the technical field of radar detection, and provides a polarized radar sea surface target multi-feature intelligent detection method and device, which comprises the following steps: obtaining sea surface echo signals received by each polarized channel of a polarized radar, wherein the sea surface echo signals comprise echo signals of a to-be-detected unit and echo signals of a reference unit; extracting a plurality of polarization features based on the sea surface echo signals of each polarized channel; obtaining an edge probability density function of each polarization feature of each polarized channel; constructing a joint probability density distribution model based on the edge probability density functions of the polarization features by using a vine copula structure; and calculating a joint probability value of the sea surface echo signals of the to-be-detected unit based on the joint probability density distribution model, and determining that the to-be-detected unit exists a target in the case that the joint probability value is less than a quantile threshold. The application improves the small target recognition capability under the non-Gaussian and non-stationary sea clutter background.

Description

Technical Field

[0001] This invention relates to the field of radar detection technology, and in particular to a method and apparatus for intelligent detection of multiple features of sea surface targets using polarization radar. Background Technology

[0002] Polarimetric radar can acquire the scattering information of a target in multiple polarization channels, and has richer observation dimensions and stronger target detection capabilities than traditional radar. Therefore, it has been widely used in the detection of small targets (targets with low radar cross-section) on the sea surface in recent years.

[0003] Traditional polarimetric radar target detection methods extract several polarimetric features from the echo signal and rely on detection algorithms with fixed statistical assumptions. These algorithms distinguish targets through linear fusion or by setting a fixed threshold. However, due to the strong non-Gaussianity and significant temporal variations of sea clutter signals under complex sea conditions, detection algorithms relying on fixed statistical assumptions struggle to maintain stable performance. Furthermore, while polarimetric feature-based methods can improve the separability between targets and clutter to some extent, the anomaly detectors they employ lack systematicity and flexibility in jointly modeling multiple features, making it difficult to effectively express the potential nonlinear dependencies between features and affecting the reliability of the detection results.

[0004] Therefore, existing polarimetric radar target detection methods struggle to achieve stable and reliable small target detection in complex sea clutter backgrounds that are non-Gaussian and non-stationary. Summary of the Invention

[0005] This invention provides a method and apparatus for intelligent detection of multiple features of sea surface targets using polarimetric radar, which solves the problem that polarimetric radar target detection methods are difficult to achieve stable and reliable small target detection in complex sea clutter backgrounds that are non-Gaussian and non-stationary.

[0006] This invention provides a multi-feature intelligent detection method for sea surface targets using polarimetric radar, comprising the following steps:

[0007] The sea surface echo signals received by each polarization channel of the polarization radar are acquired. The sea surface echo signals include the echo signals of the unit to be detected and the echo signals of the reference unit. The unit to be detected is the currently detected sub-region in the sea surface area detected by the polarization radar, and the reference unit is the sub-region around the unit to be detected.

[0008] Multiple polarization features are extracted from sea surface echo signals of each polarization channel;

[0009] Obtain the marginal probability density function of each polarization feature in each polarization channel;

[0010] A joint probability density distribution model is constructed based on the marginal probability density functions of each polarization feature using the vine copula structure.

[0011] The joint probability value of the sea surface echo signal of the unit to be detected is calculated based on the joint probability density distribution model. If the joint probability value is less than the quantile threshold, it is determined that the unit to be detected contains a target. The quantile threshold is determined based on the joint probability density set of pure clutter samples and a preset false alarm rate.

[0012] According to the present invention, a multi-feature intelligent detection method for sea surface targets using polarization radar is provided, wherein the multiple polarization features include at least three polarization features from three categories: time-domain amplitude fusion features, Doppler spectrum polarization modulation features, and rotation-domain polarization correlation features.

[0013] According to the present invention, a multi-feature intelligent detection method for sea surface targets using polarimetric radar extracts the temporal amplitude fusion features based on the sea surface echo signals from each polarization channel, including:

[0014] For each polarization channel, calculate the amplitude ratio of the sea surface echo signal of the unit under test and the reference unit respectively;

[0015] The temporal amplitude fusion feature is obtained by geometrically averaging the amplitude ratios of each polarization channel.

[0016] According to the present invention, a multi-feature intelligent detection method for sea surface targets using polarization radar is provided, which extracts the Doppler spectral polarization modulation features based on the sea surface echo signals of each polarization channel, including:

[0017] The first Doppler spectrum of the sea surface echo signal for each polarization channel is calculated, and a Doppler spectrum matrix is ​​formed based on the first Doppler spectrum.

[0018] Based on the preset virtual state vectors of multiple receiving polarization states, the virtual state vectors of multiple transmitting polarization states, and the Doppler spectrum matrix, multiple second Doppler spectra under transmitting and receiving polarization are calculated.

[0019] The composite Doppler spectrum is obtained by weighted summation of multiple second Doppler spectra and their corresponding weights.

[0020] At least one of the spectral peak value, energy concentration, and spectral entropy is extracted from the composite Doppler spectrum as the polarization modulation feature of the Doppler spectrum.

[0021] According to the present invention, a multi-feature intelligent detection method for sea surface targets using polarization radar is provided, wherein the weights are determined based on the normalized second Doppler spectrum polarization modulation.

[0022] According to the present invention, a multi-feature intelligent detection method for sea surface targets using polarization radar extracts the rotation domain polarization-related features based on the sea surface echo signals of each polarization channel, including:

[0023] The polarization scattering matrix is ​​determined based on the sea surface echo signal of each polarization channel;

[0024] The polarization scattering matrix is ​​rotated by a preset angular step size to obtain a rotation matrix, and each rotation matrix corresponds to a rotation angle.

[0025] For each rotation angle, calculate the cross-correlation coefficient between any two polarization channels;

[0026] Extract at least one of the maximum, minimum, and range of the cross-correlation coefficient with respect to the rotation angle as the rotation domain polarization correlation feature.

[0027] According to the present invention, a multi-feature intelligent detection method for sea surface targets using polarimetric radar is provided, which utilizes a vinecopula structure to construct a joint probability density distribution model based on the marginal probability density functions of each polarimetric feature, including:

[0028] Integrating the marginal probability density function of each polarization feature yields the corresponding cumulative distribution function, which is used to convert the corresponding polarization feature sample into a standard uniform variable.

[0029] Calculate the joint copula density based on the standard uniform variables corresponding to each polarization characteristic;

[0030] The marginal probability density functions of each polarization feature are multiplied by the joint copula density to construct a joint probability density distribution model that reflects the statistical dependence among multiple polarization features.

[0031] This invention also provides a polarimetric radar intelligent detection device for multiple features of sea surface targets, comprising the following modules:

[0032] The echo signal acquisition module is used to acquire the sea surface echo signals received by each polarization channel of the polarization radar. The sea surface echo signals include the echo signals of the unit to be detected and the echo signals of the reference unit. The unit to be detected is the currently detected sub-region in the sea surface area detected by the polarization radar, and the reference unit is the sub-region surrounding the unit to be detected.

[0033] The multi-polarization feature extraction module is used to extract multiple polarization features based on the sea surface echo signals of each polarization channel.

[0034] The probability density function acquisition module is used to obtain the edge probability density function of each polarization feature in each polarization channel.

[0035] The joint model building module is used to construct a joint probability density distribution model based on the marginal probability density function of each polarization feature using the vine copula structure.

[0036] The joint probability calculation module is used to calculate the joint probability value of the sea surface echo signal of the unit to be detected based on the joint probability density distribution model. If the joint probability value is less than the quantile threshold, it is determined that the unit to be detected contains a target. The quantile threshold is determined based on the joint probability density set of pure clutter samples and a preset false alarm rate.

[0037] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the program, it implements the polarization radar intelligent detection method for multiple features of sea surface targets as described above.

[0038] 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 intelligent detection method for multiple features of polarization radar sea surface targets as described above.

[0039] The present invention provides a multi-feature intelligent detection method and device for sea surface targets using polarimetric radar. This method extracts multiple polarimetric features from the sea surface echo signals of each polarization channel; obtains the edge probability density function of each polarization feature in each polarization channel, preserving the nonparametric characteristics of the corresponding sea surface echo signal, thus retaining sufficient original information for subsequent joint modeling. Furthermore, it utilizes a vine copula structure to construct a joint probability density distribution model based on the edge probability density functions of each polarization feature, effectively capturing the nonlinear correlation and conditional dependence between polarization features, improving the accuracy of the statistical interaction description of multiple features. Simultaneously, in the detection stage, an anomaly metric is constructed on a pure clutter sample set using the joint probability density distribution model, and the quantization quantile threshold of the false alarm rate is achieved by combining the quantile principle, making the detection results more stable and reliable. Moreover, the copula-based anomaly detection framework is suitable for high-dimensional, complex dependent feature scenarios, improving the ability to identify small sea surface targets in complex non-Gaussian and non-stationary sea clutter backgrounds. Attached Figure Description

[0040] 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.

[0041] Figure 1 This is a flowchart illustrating the intelligent detection method for multiple features of sea surface targets using polarization radar provided by the present invention.

[0042] Figure 2This is a comparison chart of the receiver operating characteristic (ROC) curves of the polarization radar intelligent detection method for multi-feature sea surface targets provided by this invention and other methods in the prior art.

[0043] Figure 3 This is a schematic diagram of the structure of the intelligent detection device for multiple features of polarization radar targets on the sea surface provided by the present invention.

[0044] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0045] 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.

[0046] The polarimetric radar intelligent detection method for multiple features of sea surface targets according to embodiments of the present invention, such as... Figure 1 As shown, the procedure includes steps S110 to S150.

[0047] Step S110: Acquire the sea surface echo signals received by each polarization channel of the polarization radar. The sea surface echo signals include the echo signals of the unit to be detected and the echo signals of the reference unit. The unit to be detected is the currently detected sub-region in the sea surface area detected by the polarization radar, and the reference unit is the sub-region surrounding the unit to be detected.

[0048] Specifically, when polarimetric radar detects targets on the sea surface, it divides the detection area into multiple grid-like sub-regions and performs target detection in each sub-region. The currently detected sub-region is called the target cell, and the surrounding sub-regions are called reference cells. Of course, the target cell may or may not contain a target. Based on this, the sea surface target detection problem can be formalized as a binary hypothesis testing process, assuming... Indicates the unit to be detected in the first... n A quick snapshot of the sea surface echo signal Indicates the first n The clutter signal in the sea surface echo signal captured in a quick shot Indicates the first n If a potential target echo signal is detected in a quick snapshot of the sea surface echo signal, the detection task can be described as follows:

[0049] (1).

[0050] in, K Indicates the number of reference units. N Indicates the number of snapshots. Indicates the unit to be detected in the first... n Take a quick shot of the polarization channel. p Sea surface echo signal, Indicates the first k The reference unit in the first n Take a quick shot of the polarization channel. p Sea surface echo signal, Indicates the unit to be detected in the first... n Take a quick shot of the polarization channel. p The target echo signal, Indicates the unit to be detected in the first... n Take a quick shot of the polarization channel. p Clutter signals, Indicates the first k The reference unit in the first n Take a quick shot of the polarization channel. p Clutter signals, H 0 indicates that there is no target in the unit to be detected. H 1 indicates that the unit to be detected has a target. For example, the polarization channel... p It may include a horizontal transmit horizontal receive (HH) polarization channel, a horizontal transmit vertical receive (HV) polarization channel, a vertical transmit horizontal receive (VH) polarization channel, and a vertical transmit vertical receive (VV) polarization channel.

[0051] Step S120: Extract multiple polarization features based on the sea surface echo signal of each polarization channel, that is, extract multiple polarization features for each polarization channel.

[0052] Step S130: Obtain the edge probability density function of each polarization feature in each polarization channel. The edge probability density function of the polarization feature retains the nonparametric characteristics of the corresponding sea surface echo signal and also fully preserves the original information of the sea surface echo signal.

[0053] Step S140: Construct a joint probability density distribution model based on the marginal probability density functions of each polarization feature using the vine copula structure. In this step, a joint probability density distribution model is established based on the marginal probability density functions of each polarization feature. This joint probability density distribution model reflects the statistical dependencies between multiple polarization features and can effectively capture the nonlinear correlations and conditional dependencies between various polarization features, thereby improving the accuracy of describing the statistical interactions of multiple features.

[0054] Step S150: Calculate the joint probability value of the sea surface echo signal of the unit to be detected based on the joint probability density distribution model. If the joint probability value is less than the quantile threshold, determine that the unit to be detected contains a target. The quantile threshold is determined based on the joint probability density set of pure clutter samples and a preset false alarm rate.

[0055] Specifically, based on the joint probability density distribution model, the joint probability density set is calculated on the pure clutter sample set. According to the set false alarm rate Calculate quantile threshold Th :

[0056] (2).

[0057] in, Quantile function, false alarm rate The value range is 10 -3 ~10 -1 For example, the index is taken at 21 points at even intervals between -3 and -1.

[0058] The joint probability value of the unit to be detected is calculated using the joint probability density distribution model. ,exist If the target is present in the detected unit, it is determined that the target echo signal is present in the sea surface echo signal of each polarization channel of the polarization radar obtained in step S110. Otherwise, the target is not present in the detected unit, that is, the target echo signal is present in the sea surface echo signal of each polarization channel of the polarization radar obtained in step S110.

[0059] The polarimetric radar-based intelligent detection method for multi-feature sea surface targets in this embodiment extracts multiple polarimetric features from the sea surface echo signals of each polarization channel; obtains the edge probability density function of each polarization feature in each polarization channel, preserving the nonparametric characteristics of the corresponding sea surface echo signal, thus retaining sufficient original information for subsequent joint modeling; and utilizes the vinecopula structure to construct a joint probability density distribution model based on the edge probability density functions of each polarization feature, effectively capturing the nonlinear correlation and conditional dependence between each polarization feature, improving the accuracy of the statistical interaction description of multiple features. Simultaneously, in the detection stage, an anomaly metric is constructed on the pure clutter sample set using the joint probability density distribution model, and the quantization quantile threshold of the false alarm rate is achieved by combining the quantile principle, making the detection results more stable and reliable. Moreover, the anomaly detection framework based on copula is suitable for high-dimensional, complex dependent feature scenarios, improving the ability to identify small sea surface targets in complex non-Gaussian, non-stationary sea clutter backgrounds.

[0060] In some embodiments, the plurality of polarization features includes at least three polarization features selected from three categories: time-domain amplitude fusion features, Doppler spectrum polarization modulation features, and rotation-domain polarization correlation features. These three categories of features all contain rich physical meaning, describing the characteristics of sea surface echo signals and target echo signals in the time domain, frequency domain, and polarization rotation domain, respectively. The features have low correlation but are complementary, providing a certain degree of differentiation between sea surface echo signals and target echo signals, which is beneficial for subsequent detection of sea surface targets. Preferably, selecting one polarization feature from each category can further improve the ability to identify small sea surface targets against complex non-Gaussian and non-stationary sea clutter backgrounds.

[0061] In some embodiments, when the polarization feature is a time-domain amplitude fusion feature, step S120, extracting the time-domain amplitude fusion feature based on the sea surface echo signal of each polarization channel, includes:

[0062] Step S121a: For each polarization channel, calculate the amplitude ratio of the sea surface echo signal of the unit under test and the reference unit respectively. The specific calculation formula is as follows:

[0063] (3);

[0064] (4).

[0065] in, .

[0066] A p Polarization channels of the unit under test in all snapshots p The average amplitude of the sea surface echo signal. For K reference units, polarization channels are used in all snapshots. p The average amplitude of the sea surface echo signal, where | represents the modulus operation.

[0067] Step S122a: Obtain the temporal amplitude fusion feature by geometrically averaging the amplitude ratios corresponding to each polarization channel. Specifically, the formula for geometric mean fusion is as follows:

[0068] (5).

[0069] In some embodiments, when the polarization feature is a Doppler spectral polarization modulation feature, step S120, extracting the Doppler spectral polarization modulation feature based on the sea surface echo signal of each polarization channel, includes:

[0070] Step S121b: Calculate the first Doppler spectrum of the sea surface echo signal for each polarization channel, and form a Doppler spectrum matrix based on the first Doppler spectrum.

[0071] Specifically, the first Doppler spectrum is obtained by performing a Fourier transform on the sea surface echo signals of each polarization channel. The transform formula is as follows:

[0072] (6).

[0073] in, f Indicates frequency, n Indicates a quick snapshot. T express The pulse repetition frequency. The first Doppler spectrum corresponding to each polarization channel. Forming the Doppler spectral matrix The first Doppler spectra corresponding to the four polarization channels form a 2×2 Doppler spectrum matrix. ,Right now:

[0074] .

[0075] in, This represents the first Doppler spectrum corresponding to polarization channel HH. This represents the first Doppler spectrum corresponding to the polarization channel HV. This represents the first Doppler spectrum corresponding to the polarization channel VH. This represents the first Doppler spectrum corresponding to the polarization channel VV.

[0076] Step S122b: Based on the preset virtual state vectors of multiple receiving polarization states, the virtual state vectors of multiple transmitting polarization states, and the Doppler spectrum matrix, calculate multiple second Doppler spectra under transmitting and receiving polarization. .

[0077] (7).

[0078] in, v t and v r The virtual state vectors representing the emission polarization states are respectively and the virtual state vector of the receiving polarization state , This indicates that the matrix is ​​transposed using the conjugate operation. , , and Both represent polarization ellipse descriptors. , , and Each takes multiple different values, resulting in multiple virtual state vectors for the transmit polarization state and multiple virtual state vectors for the receive polarization state, for example: , , and By taking five different values ​​for each, 25 different transmit and receive polarization states are obtained. Combining these with formula (7), the second Doppler spectrum under 625 different transmit and receive polarization states can be calculated. S 1(f)~ S 625 (f).

[0079] Step S123b: Obtain the composite Doppler spectrum by weighted summation of multiple second Doppler spectra and their corresponding weights. The specific weighted summation formula is as follows:

[0080] (8).

[0081] in, L This indicates the total number of the second Doppler spectrum. Based on different and Calculated according to the above formula (7), This indicates the weight corresponding to the second Doppler spectrum.

[0082] Step S124b: Extract at least one of the spectral peak, energy concentration and spectral entropy from the composite Doppler spectrum as the polarization modulation feature of the Doppler spectrum.

[0083] Specifically, the spectral peak value PDM = max( S mod (f)).

[0084] The formula for calculating Energy Concentration Ratio (PDR) is as follows:

[0085] .

[0086] Spectral entropy PDE = entrophy( S mod (f)).

[0087] The specific formula for calculating the spectral entropy function entrophy() is as follows. First, let S... mod (f) The amplitude range of each frequency point is divided into several intervals:

[0088] .

[0089] The proportion P of the number of frequency points falling within each amplitude range to the total number of frequency points. j Based on the ratio P jCalculate the spectral entropy based on the entropy function:

[0090] .

[0091] In some embodiments, the weights are determined based on the normalized second Doppler spectrum polarization modulation. Specifically, for each second Doppler spectrum... Normalization was performed to obtain the normalized second Doppler spectrum. For the second Doppler spectrum Weights are obtained by polarization modulation. :

[0092] (9).

[0093] The weights are determined based on the polarization modulation of the normalized second Doppler spectrum, so that components with aligned frequencies and relatively stable amplitudes under multiple polarization combinations are retained (target features), while components with frequency drift and violent fluctuations are averaged out during fusion (clutter features). This is beneficial for amplifying the difference between the Doppler spectra of the target and clutter, and is more conducive to the subsequent identification of small targets on the sea surface.

[0094] In some embodiments, when the polarization feature is a rotational domain polarization-related feature, step S120 involves extracting the rotational domain polarization-related feature based on the sea surface echo signal of each polarization channel, including:

[0095] Step S121c: Determine the polarization scattering matrix based on the sea surface echo signals of each polarization channel. For example, the sea surface echo signals corresponding to the four polarization channels form a 2×2 polarization scattering matrix. S as follows:

[0096] (10).

[0097] in, This represents the sea surface echo signal corresponding to the polarization channel HH. This represents the sea surface echo signal corresponding to the polarization channel HV. This represents the sea surface echo signal corresponding to the polarization channel VH. This represents the sea surface echo signal corresponding to polarization channel VV. It should be noted that a polarization scattering matrix is ​​formed for the sea surface echo signals corresponding to the four polarization channels of both the detected unit and the reference unit. S Therefore, formula (10) does not distinguish between the unit to be detected and the reference unit.

[0098] Step S122c: Rotate the polarization scattering matrix according to a preset angular step size to obtain a rotation matrix. S ( θ Each rotation matrix corresponds to a rotation angle. θSpecifically, the polarization basis rotation formula is as follows:

[0099] (11);

[0100] (12).

[0101] in, express The transpose of the matrix, S ( θ )and S They have the same shape, both being 2×2 matrices, and their matrix elements are about... θ The function.

[0102] Step S123c: For each rotation angle, calculate the cross-correlation coefficient between any two polarization channels. Taking the HH and HV polarization channels as an example, the formula for calculating the cross-correlation coefficient between the HH and HV polarization channels is as follows:

[0103] (13).

[0104] in, Represents N snapshots The expected value is obtained by averaging the N snapshots according to the formula (10) above. Each element of matrix S changes with the snapshot n.

[0105] Step S124c: Extract at least one of the maximum, minimum, and range of the cross-correlation coefficient with respect to the rotation angle as the rotation domain polarization correlation feature.

[0106] In some embodiments, step S130, obtaining the marginal probability density function of each polarization feature of each polarization channel specifically includes: using the kernel density estimation (KDE) algorithm to obtain the marginal probability density function of each polarization feature of each polarization channel. The formula for the kernel density estimation (KDE) algorithm is as follows:

[0107] (14).

[0108] in, Indicates the first i The marginal probability density function of each polarization feature z The independent variable representing the marginal probability density function, M Indicates the number of random samples. Indicates the first i The first polarization characteristic m A random sample, h Here, K represents the bandwidth, and K() represents the Gaussian kernel function. It should be noted that: It is the polarization feature extracted from the sample in the pure clutter sample set.

[0109] In some embodiments, step S140, constructing a joint probability density distribution model based on the marginal probability density function of each polarization feature using the vine copula structure, includes:

[0110] Step S141: Integrate the marginal probability density function of each polarization feature to obtain the corresponding cumulative distribution function. The cumulative distribution function is used to convert the corresponding polarization feature samples. Transform into a standard uniform variable ,Right now , Indicates the first i A standard uniform variable with polarization characteristics.

[0111] Step S142: Calculate the joint copula density based on the standard uniform variables corresponding to each polarization feature. Taking one polarization feature from each of the above three types of polarization features as an example, the joint copula density of the three polarization features is... The specific calculation method is as follows:

[0112] (15);

[0113] (16);

[0114] (17).

[0115] in, The copula density represents the combined density of polarization feature 1 and polarization feature 2. The joint copula density represents polarization feature 2 and polarization feature 3. This represents the joint copula density of polarization features 1 and 3 under the condition of polarization feature 2.

[0116] Step S143: Multiply the marginal probability density function of each polarization feature with the joint copula density to construct a joint probability density distribution model that reflects the statistical dependence among multiple polarization features. Taking the above three polarization features as an example, the joint probability density distribution model is expressed by the following formula:

[0117] (18).

[0118] Based on the above joint probability density distribution model, in step S150: calculate the joint probability value of the sea surface echo signal of the unit to be detected based on the joint probability density distribution model, that is, substitute the three polarization features corresponding to the sea surface echo signal of the unit to be detected into the above formula (18) to replace the independent variables z1, z2 and z3 of the edge probability density function to obtain the joint probability value of the three polarization features.

[0119] like Figure 2 As shown, the detection performance of existing anomaly detectors and the method proposed in this invention is compared on the IPIX 1993 dataset. Compared with existing anomaly detectors, the copula modeling framework proposed in this invention can more accurately capture the statistical correlation between multiple features, significantly improving the detection probability of small targets while maintaining a controllable false alarm rate. Specifically, as... Figure 2 As shown, the ROC (receiver operating characteristic curve) curve ( Figure 2 The curve represented by the red triangle in the middle is the ROC curve corresponding to the method of this invention. This shows that the method of this invention achieves a certain ROC curve within 10... -3 Up to 10 -1 It exhibits higher detection performance across the false alarm rate range, verifying its robustness and effectiveness in complex sea clutter environments.

[0120] The following describes the intelligent detection device for multiple features of sea surface targets using polarimetric radar, which is described below. The intelligent detection device for multiple features of sea surface targets using polarimetric radar described above can be referred to in correspondence with each other.

[0121] The polarization radar intelligent detection device for multi-feature sea surface targets according to embodiments of the present invention, such as... Figure 3 As shown, it includes:

[0122] The echo signal acquisition module 310 is used to acquire the sea surface echo signals received by each polarization channel of the polarization radar. The sea surface echo signals include the echo signals of the unit to be detected and the echo signals of the reference unit. The unit to be detected is the currently detected sub-region in the sea surface area detected by the polarization radar, and the reference unit is the sub-region around the unit to be detected.

[0123] The multi-polarization feature extraction module 320 is used to extract multiple polarization features based on the sea surface echo signal of each polarization channel.

[0124] The probability density function acquisition module 330 is used to acquire the edge probability density function of each polarization feature of each polarization channel.

[0125] The joint model building module 340 is used to construct a joint probability density distribution model based on the marginal probability density function of each polarization feature using the vine copula structure.

[0126] The joint probability calculation module 350 is used to calculate the joint probability value of the sea surface echo signal of the unit to be detected based on the joint probability density distribution model. If the joint probability value is less than the quantile threshold, it is determined that the unit to be detected contains a target. The quantile threshold is determined based on the joint probability density set of pure clutter samples and a preset false alarm rate.

[0127] In some embodiments, the plurality of polarization features include at least three polarization features from three categories: time-domain amplitude fusion features, Doppler spectrum polarization modulation features, and rotation-domain polarization correlation features.

[0128] In some embodiments, the multipolar feature extraction module 320 is specifically used for:

[0129] For each polarization channel, calculate the amplitude ratio of the sea surface echo signal of the unit under test and the reference unit respectively;

[0130] The temporal amplitude fusion feature is obtained by geometrically averaging the amplitude ratios of each polarization channel.

[0131] In some embodiments, the multipolar feature extraction module 320 is specifically used for:

[0132] The first Doppler spectrum of the sea surface echo signal for each polarization channel is calculated, and a Doppler spectrum matrix is ​​formed based on the first Doppler spectrum.

[0133] Based on the preset virtual state vectors of multiple receiving polarization states, the virtual state vectors of multiple transmitting polarization states, and the Doppler spectrum matrix, multiple second Doppler spectra under transmitting and receiving polarization are calculated.

[0134] The composite Doppler spectrum is obtained by weighted summation of multiple second Doppler spectra and their corresponding weights.

[0135] At least one of the spectral peak value, energy concentration, and spectral entropy is extracted from the composite Doppler spectrum as the polarization modulation feature of the Doppler spectrum.

[0136] In some embodiments, the weights are determined based on the normalized second Doppler spectral polarization modulation.

[0137] In some embodiments, the multipolar feature extraction module 320 is specifically used for:

[0138] The polarization scattering matrix is ​​determined based on the sea surface echo signal of each polarization channel;

[0139] The polarization scattering matrix is ​​rotated by a preset angular step size to obtain a rotation matrix, and each rotation matrix corresponds to a rotation angle.

[0140] For each rotation angle, calculate the cross-correlation coefficient between any two polarization channels;

[0141] Extract at least one of the maximum, minimum, and range of the cross-correlation coefficient with respect to the rotation angle as the rotation domain polarization correlation feature.

[0142] In some embodiments, the joint model construction module 340 includes:

[0143] The function integration module is used to integrate the marginal probability density function of each polarization feature to obtain the corresponding cumulative distribution function. The cumulative distribution function is used to convert the corresponding polarization feature sample into a standard uniform variable.

[0144] The joint copula density calculation module is used to calculate the joint copula density based on the standard uniform variables corresponding to each polarization feature.

[0145] The quadrature module is used to quadrature the marginal probability density functions of each polarization feature with the joint copula density to construct a joint probability density distribution model that reflects the statistical dependencies between multiple polarization features.

[0146] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a polarization radar intelligent detection method for multi-feature sea surface targets, which includes:

[0147] The system acquires sea surface echo signals received by each polarization channel of the polarization radar. The sea surface echo signals include echo signals of the unit to be detected and echo signals of the reference unit. The unit to be detected is the currently detected sub-region in the sea surface area detected by the polarization radar, and the reference unit is the sub-region surrounding the unit to be detected.

[0148] Multiple polarization features are extracted from sea surface echo signals of each polarization channel.

[0149] Obtain the marginal probability density function of each polarization feature in each polarization channel.

[0150] A joint probability density distribution model is constructed using the vine copula structure based on the marginal probability density functions of each polarization feature.

[0151] The joint probability value of the sea surface echo signal of the unit to be detected is calculated based on the joint probability density distribution model. If the joint probability value is less than the quantile threshold, it is determined that the unit to be detected contains a target. The quantile threshold is determined based on the joint probability density set of pure clutter samples and a preset false alarm rate.

[0152] Furthermore, the logical instructions in the aforementioned memory 430 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, 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 described in 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.

[0153] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the polarization radar intelligent detection method for multiple features of sea surface targets provided by the above methods, the method including:

[0154] The system acquires sea surface echo signals received by each polarization channel of the polarization radar. The sea surface echo signals include echo signals of the unit to be detected and echo signals of the reference unit. The unit to be detected is the currently detected sub-region in the sea surface area detected by the polarization radar, and the reference unit is the sub-region surrounding the unit to be detected.

[0155] Multiple polarization features are extracted from sea surface echo signals of each polarization channel.

[0156] Obtain the marginal probability density function of each polarization feature in each polarization channel.

[0157] A joint probability density distribution model is constructed using the vine copula structure based on the marginal probability density functions of each polarization feature.

[0158] The joint probability value of the sea surface echo signal of the unit to be detected is calculated based on the joint probability density distribution model. If the joint probability value is less than the quantile threshold, it is determined that the unit to be detected contains a target. The quantile threshold is determined based on the joint probability density set of pure clutter samples and a preset false alarm rate.

[0159] In another aspect, 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 intelligent detection method for multiple features of polarization radar sea surface targets provided by the methods described above, the method comprising:

[0160] The system acquires sea surface echo signals received by each polarization channel of the polarization radar. The sea surface echo signals include echo signals of the unit to be detected and echo signals of the reference unit. The unit to be detected is the currently detected sub-region in the sea surface area detected by the polarization radar, and the reference unit is the sub-region surrounding the unit to be detected.

[0161] Multiple polarization features are extracted from sea surface echo signals of each polarization channel.

[0162] Obtain the marginal probability density function of each polarization feature in each polarization channel.

[0163] A joint probability density distribution model is constructed using the vine copula structure based on the marginal probability density functions of each polarization feature.

[0164] The joint probability value of the sea surface echo signal of the unit to be detected is calculated based on the joint probability density distribution model. If the joint probability value is less than the quantile threshold, it is determined that the unit to be detected contains a target. The quantile threshold is determined based on the joint probability density set of pure clutter samples and a preset false alarm rate.

[0165] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. 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.

[0166] 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., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0167] 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 multi-feature intelligent detection method for sea surface targets using polarimetric radar, characterized in that, include: The sea surface echo signals received by each polarization channel of the polarization radar are acquired. The sea surface echo signals include the echo signals of the unit to be detected and the echo signals of the reference unit. The unit to be detected is the currently detected sub-region in the sea surface area detected by the polarization radar, and the reference unit is the sub-region around the unit to be detected. Multiple polarization features are extracted from sea surface echo signals of each polarization channel; Obtain the marginal probability density function of each polarization feature in each polarization channel; A joint probability density distribution model is constructed based on the marginal probability density functions of each polarization feature using the vine copula structure. The joint probability value of the sea surface echo signal of the unit to be detected is calculated based on the joint probability density distribution model. If the joint probability value is less than the quantile threshold, it is determined that the unit to be detected contains a target. The quantile threshold is determined based on the joint probability density set of pure clutter samples and a preset false alarm rate.

2. The intelligent detection method for multiple features of sea surface targets using polarimetric radar according to claim 1, characterized in that, The multiple polarization features include at least three polarization features from three categories: time-domain amplitude fusion features, Doppler spectrum polarization modulation features, and rotation-domain polarization correlation features.

3. The intelligent detection method for multiple features of sea surface targets using polarimetric radar according to claim 2, characterized in that, The temporal amplitude fusion features are extracted based on the sea surface echo signals from each polarization channel, including: For each polarization channel, calculate the amplitude ratio of the sea surface echo signal of the unit under test and the reference unit respectively; The temporal amplitude fusion feature is obtained by geometrically averaging the amplitude ratios of each polarization channel.

4. The intelligent detection method for multiple features of sea surface targets using polarimetric radar according to claim 2, characterized in that, The Doppler spectral polarization modulation features are extracted based on the sea surface echo signals from each polarization channel, including: The first Doppler spectrum of the sea surface echo signal for each polarization channel is calculated, and a Doppler spectrum matrix is ​​formed based on the first Doppler spectrum. Based on the preset virtual state vectors of multiple receiving polarization states, the virtual state vectors of multiple transmitting polarization states, and the Doppler spectrum matrix, multiple second Doppler spectra under transmitting and receiving polarization are calculated. The composite Doppler spectrum is obtained by weighted summation of multiple second Doppler spectra and their corresponding weights. At least one of the spectral peak value, energy concentration, and spectral entropy is extracted from the composite Doppler spectrum as the polarization modulation feature of the Doppler spectrum.

5. The intelligent detection method for multiple features of sea surface targets using polarimetric radar according to claim 4, characterized in that, The weights are determined based on the normalized second Doppler spectrum polarization modulation.

6. The intelligent detection method for multiple features of sea surface targets using polarimetric radar according to claim 2, characterized in that, The rotational domain polarization correlation features are extracted based on the sea surface echo signals from each polarization channel, including: The polarization scattering matrix is ​​determined based on the sea surface echo signal of each polarization channel; The polarization scattering matrix is ​​rotated by a preset angular step size to obtain a rotation matrix, and each rotation matrix corresponds to a rotation angle. For each rotation angle, calculate the cross-correlation coefficient between any two polarization channels; Extract at least one of the maximum, minimum, and range of the cross-correlation coefficient with respect to the rotation angle as the rotation domain polarization correlation feature.

7. The intelligent detection method for multiple features of sea surface targets using polarimetric radar according to any one of claims 1 to 6, characterized in that, A joint probability density distribution model is constructed using the vine copula structure based on the marginal probability density functions of each polarization feature, including: Integrating the marginal probability density function of each polarization feature yields the corresponding cumulative distribution function, which is used to convert the corresponding polarization feature sample into a standard uniform variable. Calculate the joint copula density based on the standard uniform variables corresponding to each polarization characteristic; The marginal probability density functions of each polarization feature are multiplied by the joint copula density to construct a joint probability density distribution model that reflects the statistical dependence among multiple polarization features.

8. A polarimetric radar-based intelligent detection device for multiple features of sea surface targets, characterized in that, include: The echo signal acquisition module is used to acquire the sea surface echo signals received by each polarization channel of the polarization radar. The sea surface echo signals include the echo signals of the unit to be detected and the echo signals of the reference unit. The unit to be detected is the currently detected sub-region in the sea surface area detected by the polarization radar, and the reference unit is the sub-region around the unit to be detected. The multi-polarization feature extraction module is used to extract multiple polarization features based on the sea surface echo signal of each polarization channel; The probability density function acquisition module is used to obtain the edge probability density function of each polarization feature in each polarization channel; The joint model building module is used to construct a joint probability density distribution model based on the marginal probability density function of each polarization feature using the vine copula structure. The joint probability calculation module is used to calculate the joint probability value of the sea surface echo signal of the unit to be detected based on the joint probability density distribution model. If the joint probability value is less than the quantile threshold, it is determined that the unit to be detected contains a target. The quantile threshold is determined based on the joint probability density set of pure clutter samples and a preset false alarm rate.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the intelligent detection method for multiple features of polarized radar sea surface targets as described in any one of claims 1 to 7.

10. 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 intelligent detection method for multiple features of polarized radar sea surface targets as described in any one of claims 1 to 7.

Images