CN112114295A

CN112114295A - Target identification method and system for full-polarization radar

Info

Publication number: CN112114295A
Application number: CN202010742966.8A
Authority: CN
Inventors: 付哲泉; 但波; 刘瑜; 康家方; 谭大宁; 张军涛; 王旭坤; 林迅
Original assignee: Naval Aeronautical University
Current assignee: Naval Aeronautical University
Priority date: 2020-07-29
Filing date: 2020-07-29
Publication date: 2020-12-22
Anticipated expiration: 2040-07-29
Also published as: CN112114295B

Abstract

The invention discloses a method and a system for identifying a target of a fully-polarized radar. The method comprises the following steps: decomposing polarized electromagnetic waves into horizontal polarized electromagnetic waves and vertical polarized electromagnetic waves which are perpendicular to each other, determining a polarized scattering matrix according to the relation between an incident field and a scattering field of the electromagnetic waves, and performing feature extraction on the polarized scattering matrix by respectively adopting a Pauli decomposition method, an H, alpha, A decomposition method and a feature extraction method of structural similarity parameters to obtain a first feature vector, a second feature vector and a third feature vector of target fully polarized HRRP data; and combining the first feature vector, the second feature vector and the third feature vector with a neural network to further realize the identification of the fully-polarized radar target. By adopting the method and the system, the ship target HRRP information of the four polarization channels is used for extracting the target full polarization characteristics, so that more characteristics of the target full polarization HRRP can be reserved, the accuracy of target identification is improved, and the classification effect is improved.

Description

Target identification method and system for full-polarization radar

Technical Field

The invention relates to the technical field of radar target identification, in particular to a method and a system for identifying a fully-polarized radar target.

Background

With the continuous development of radar technology, researchers hope to obtain more effective information of targets for classification and identification of the targets, and radar directors in the traditional system cannot meet the requirement of diversification at the present stage. Polarization domain information in radar target echo signals is another important characteristic used for target identification after characteristics of time domain, frequency domain and the like.

The development of polarization correlation techniques and the application of radar guidance heads in a fully polarized regime have made possible polarization-feature-based target identification. Because the size of the ship target is far larger than the wavelength of the electromagnetic wave of the radar seeker, the ship target needs to be regarded as an ultra-large-size target when the scattering characteristic of the ship on the radar electromagnetic wave is analyzed. The polarization characteristics of echo signals can be changed by vector superposition of electromagnetic waves reflected by different parts of the ship in space, so that the polarization characteristics of ship target echo signals have direct relations with the shape, the posture, the surface coating and the like of the ship. Polarization domain information is one of the indispensable characteristics for completely describing the echo signals of the ship target.

Compared with the traditional single-polarization system radar, the full-polarization system radar has the following advantages:

(1) the radar of the full-polarization system can obtain richer scattering information of the target, and relevant physical characteristics of the target, including shape, material and the like, are estimated according to the information;

(2) the polarization characteristics are extracted by utilizing the full polarization information of the target echo, so that the influence of a target attitude angle and the like on target identification can be reduced, and the classification capability is improved;

(3) the full-polarization radar can utilize different receiving and transmitting polarization states to detect different types of targets more comprehensively and flexibly, and overcomes the limitation of single polarization state of the electromagnetic wave of the single-polarization radar;

(4) the full-polarization system radar can select a proper receiving and transmitting polarization mode according to needs, and the anti-interference and sea clutter resistant capabilities of the radar when a target is detected are improved.

The high-resolution radar technology enables the description of the target information by the echo signals to be more accurate, and the description of the target information by the echo signals is more comprehensive through the polarization domain information. The radar target identification combining the high-resolution technology and the full-polarization information is widely concerned by the learners and becomes an important direction for the development of the radar signal processing and target identification technology.

The High Resolution Range Profile (HRRP) is a vector sum of sub-echoes of a target scattering point obtained by using a broadband signal projected on a radar ray, and the HRRP provides a distribution condition of the target scattering point along a distance direction, and is an important structural feature of a target. The effective separability characteristic of the HRRP is extracted, and the problems of structural similarity between different ships and the amplitude, translation, attitude angle sensitivity and the like caused by high sensitivity of ship buildings to electromagnetic wave scattering can be solved. At present, a target identification research based on the HRRP polarization feature extraction of the fully-polarized broadband system radar mainly comprises the following two aspects:

(1) the HRRP information and the polarization information of the target are directly combined to construct a two-dimensional distance-polarization characteristic, and then different classification identification methods are adopted for identification.

(2) And extracting effective polarization characteristics of the target from the polarization scattering matrix based on a polarization target decomposition theory for target identification.

The two researches mainly focus on extracting features with strong separability from target full-polarization HRRP, and when a part of scholars calculate a coherent matrix, all distance units contained in each element of the coherent matrix are averaged, namely all distance units are used as measurement scales, so that the dimensionality of a feature space and the calculation complexity in a data processing process are reduced to a great extent, but specific features of each distance unit cannot be reserved, and the accuracy of target identification is low.

Disclosure of Invention

The invention aims to provide a method and a system for identifying a target of a fully-polarized radar, which are used for extracting target fully-polarized characteristics by using ship target HRRP information of four polarized channels, can keep more characteristics of the target fully-polarized HRRP, improve the accuracy of target identification and improve the classification effect.

In order to achieve the purpose, the invention provides the following scheme:

a fully polarized radar target identification method comprises the following steps:

acquiring polarized electromagnetic waves, and decomposing the polarized electromagnetic waves into horizontal polarized electromagnetic waves and vertical polarized electromagnetic waves which are vertical to each other;

determining a polarization scattering matrix according to the relation between the incident field of the horizontally polarized electromagnetic wave and the scattering field of the horizontally polarized electromagnetic wave and according to the relation between the incident field of the vertically polarized electromagnetic wave and the scattering field of the vertically polarized electromagnetic wave; elements of the polarization scattering matrix are respectively a horizontal co-polarization item, a vertical co-polarization item and two cross-polarization items corresponding to HRRP data of four polarization channels;

extracting the characteristics of the polarization scattering matrix by adopting a Pauli decomposition method to obtain a first characteristic vector of target full polarization HRRP data; the first feature vector reflects the proportion of the scattering type in the target echo;

performing feature extraction on the polarization scattering matrix by adopting an H, alpha, A decomposition method to obtain a second feature vector of the target fully-polarized HRRP data; the second feature vector reflects a probability of occurrence of a scattering mechanism;

performing feature extraction on the polarization scattering matrix by adopting a feature extraction method of structural similarity parameters to obtain a third feature vector of the target fully-polarized HRRP data; the third feature vector reflects similarity of scattering features between the two targets;

and training a neural network model by taking the first feature vector, the second feature vector and the third feature vector of the target fully-polarized HRRP data as input and the radar target class as output to obtain a trained neural network model, and identifying a fully-polarized radar target according to the trained neural network model.

Optionally, the determining a polarization scattering matrix according to a relationship between the incident field of the horizontally polarized electromagnetic wave and the scattered field of the horizontally polarized electromagnetic wave and according to a relationship between the incident field of the vertically polarized electromagnetic wave and the scattered field of the vertically polarized electromagnetic wave specifically includes:

determining a polarization scattering matrix according to the following formula:

wherein S is a polarization scattering matrix,

is the incident field of the horizontally polarized electromagnetic wave,

is the incident field of a vertically polarized electromagnetic wave,

is the scattered field of the horizontally polarized electromagnetic wave,

scattered field, S, being vertically polarized electromagnetic waves_HHBeing horizontally co-polarized terms, S_VVIs a vertically co-polarized term, S_HVIs the first cross-polarization term, S_VHIs the second cross-polarization term.

Optionally, the performing feature extraction on the polarization scattering matrix by using a Pauli decomposition method to obtain a first feature vector of the target fully-polarized HRRP data specifically includes:

pauli base pair is adopted for theThe polarization scattering matrix is expressed to obtain S ═ aX₀+bX₁+cX₂Determining a, b and c as a first feature vector; wherein, X₀、X₁And X₂The first term, the second term and the third term of Pauli base are respectively, and a, b and c are respectively the proportion of three scattering types in the target echo;

acquiring HRRP data of four polarization channels; HRRP data of the four polarization channels are respectively s_HH(n)、s_HV(n)、s_VH(n) and s_VV(N), wherein N is 0,1,2, and N-1, N is the number of radar range units of the HRRP data, and s is the number of radar range units of the HRRP data_HH(n) is the horizontal co-polarization term of the nth radar range unit, s_HV(n) is the first cross-polarization term, s, of the nth radar range unit_VH(n) is the second cross-polarization term, s, of the nth radar range unit_VV(n) is the vertical co-polarization term of the nth radar range unit;

determining the first feature vector according to the HRRP data of the four polarized channels by adopting the following formula:

wherein, a (n), b (n) and c (n) are the first feature vectors of the nth radar range unit.

Optionally, the performing feature extraction on the polarization scattering matrix by using an H, α, a decomposition method to obtain a second feature vector of the target fully-polarized HRRP data specifically includes:

determining a polarization scattering vector of each radar range unit according to the first feature vector; the polarization scattering vector k (n) of the nth radar range bin is as follows:

determining a coherent matrix of each radar range unit according to the polarization scattering vector; the coherence matrix t (n) for the nth radar range bin is as follows:

T(n)＝k(n)·k^H(n)

performing eigenvalue decomposition on the coherent matrix to obtain

And will be_j(n) determining a jth eigenvalue of a coherence matrix for an nth radar range cell; wherein, mu_j(n) is lambda_j(n) corresponding feature vectors;

calculating the probability of each scattering mechanism according to the characteristic values; probability P of occurrence of jth scattering mechanism of nth radar range unit_j(n) the following:

respectively calculating a scattering entropy, a scattering angle and an anisotropy according to the probability of occurrence of a scattering mechanism, and determining the scattering entropy, the scattering angle and the anisotropy as second feature vectors; the scattering angle comprises a dominant scattering angle, and the dominant scattering angle is a scattering angle corresponding to the maximum value in the characteristic values of the coherence matrix;

wherein the content of the first and second substances,

wherein H (n) is the scattering entropy of the nth radar range unit, alpha (n) is the scattering angle of the nth radar range unit, A (n) is the anisotropy of the nth radar range unit, and alpha_j(n) j characteristic value pair of n radar range unitThe corresponding scattering angle.

Optionally, the performing feature extraction on the polarization scattering matrix by using a feature extraction method of the structural similarity parameter to obtain a third feature vector of the target full-polarization HRRP data specifically includes:

respectively determining a polarization scattering vector of a target and a polarization scattering vector of a standard body; the standard body comprises a flat plate, a dihedral angle, a horizontal dipole, a cylinder, a left helix and a right helix;

calculating similarity parameters of the target and each standard body according to the polarization scattering vector of the target and the polarization scattering vector of the standard body, and determining the similarity parameters of the target and each standard body as third feature vectors; the formula for calculating the similarity parameter is as follows:

in the formula, r (S)₁,S₂) Is a similarity parameter between polarization scattering matrices, S₁Is a polarization scattering matrix of the target, S₂Is a polarization scattering matrix of a standard volume, k₁Is the polarization scattering vector, k, of the target₂Is the polarization scattering vector of the standard volume.

The invention also provides a full-polarization radar target identification system, which comprises:

the electromagnetic wave decomposition module is used for acquiring polarized electromagnetic waves and decomposing the polarized electromagnetic waves into horizontal polarized electromagnetic waves and vertical polarized electromagnetic waves which are vertical to each other;

the polarized scattering matrix determining module is used for determining a polarized scattering matrix according to the relation between the incident field of the horizontally polarized electromagnetic wave and the scattering field of the horizontally polarized electromagnetic wave and according to the relation between the incident field of the vertically polarized electromagnetic wave and the scattering field of the vertically polarized electromagnetic wave; elements of the polarization scattering matrix are respectively a horizontal co-polarization item, a vertical co-polarization item and two cross-polarization items corresponding to HRRP data of four polarization channels;

the first characteristic vector determining module is used for extracting characteristics of the polarization scattering matrix by adopting a Pauli decomposition method to obtain a first characteristic vector of target full-polarization HRRP data; the first feature vector reflects the proportion of the scattering type in the target echo;

the second characteristic vector determining module is used for extracting characteristics of the polarization scattering matrix by adopting an H, alpha and A decomposition method to obtain a second characteristic vector of the target full-polarization HRRP data; the second feature vector reflects a probability of occurrence of a scattering mechanism;

the third characteristic vector determining module is used for extracting the characteristics of the polarization scattering matrix by adopting a characteristic extraction method of the structural similarity parameters to obtain a third characteristic vector of the target full-polarization HRRP data; the third feature vector reflects similarity of scattering features between the two targets;

and the fully-polarized radar target identification module is used for training a neural network model by taking the first feature vector, the second feature vector and the third feature vector of the target fully-polarized HRRP data as input and taking the radar target category as output to obtain a trained neural network model, and carrying out fully-polarized radar target identification according to the trained neural network model.

Optionally, the polarized scattering matrix determining module specifically includes:

a polarized scattering matrix determination unit for determining a polarized scattering matrix according to the following formula:

wherein S is a polarization scattering matrix,

is the incident field of the horizontally polarized electromagnetic wave,

is the incident field of a vertically polarized electromagnetic wave,

is the scattered field of the horizontally polarized electromagnetic wave,

Optionally, the first feature vector determining module specifically includes:

a polarization scattering matrix processing unit for representing the polarization scattering matrix by Pauli base to obtain S ═ aX₀+bX₁+cX₂Determining a, b and c as a first feature vector; wherein, X₀、X₁And X₂The first term, the second term and the third term of Pauli base are respectively, and a, b and c are respectively the proportion of three scattering types in the target echo;

the HRRP data acquisition units of the four polarization channels are used for acquiring the HRRP data of the four polarization channels; HRRP data of the four polarization channels are respectively s_HH(n)、s_HV(n)、s_VH(n) and s_VV(N), wherein N is 0,1,2, and N-1, N is the number of radar range units of the HRRP data, and s is the number of radar range units of the HRRP data_HH(n) is the horizontal co-polarization term of the nth radar range unit, s_HV(n) is the first cross-polarization term, s, of the nth radar range unit_VH(n) is the second cross-polarization term, s, of the nth radar range unit_VV(n) is the vertical co-polarization term of the nth radar range unit;

a first feature vector determining unit, configured to determine the first feature vector according to the HRRP data of the four polarization channels by using the following formula:

Optionally, the second feature vector determining module specifically includes:

the polarized scattering vector determining unit is used for determining the polarized scattering vector of each radar range unit according to the first feature vector; the polarization scattering vector k (n) of the nth radar range bin is as follows:

a coherent matrix determining unit, configured to determine a coherent matrix of each radar range unit according to the polarization scattering vector; the coherence matrix t (n) for the nth radar range bin is as follows:

T(n)＝k(n)·k^H(n)

an eigenvalue decomposition unit for performing eigenvalue decomposition on the coherent matrix to obtain

a probability calculation unit for calculating the probability of each scattering mechanism according to the characteristic values; probability P of occurrence of jth scattering mechanism of nth radar range unit_j(n) the following:

a second feature vector determination unit, configured to calculate a scattering entropy, a scattering angle, and an anisotropy degree according to a probability of occurrence of a scattering mechanism, and determine the scattering entropy, the scattering angle, and the anisotropy degree as a second feature vector; the scattering angle comprises a dominant scattering angle, and the dominant scattering angle is a scattering angle corresponding to the maximum value in the characteristic values of the coherence matrix;

wherein the content of the first and second substances,

wherein H (n) is the scattering entropy of the nth radar range unit, alpha (n) is the scattering angle of the nth radar range unit, A (n) is the anisotropy of the nth radar range unit, and alpha_jAnd (n) is a scattering angle corresponding to the jth eigenvalue of the nth radar range unit.

Optionally, the third feature vector determining module specifically includes:

the polarized scattering vector determining unit of the target and the polarized scattering vector determining unit of the standard body are used for respectively determining the polarized scattering vector of the target and the polarized scattering vector of the standard body; the standard body comprises a flat plate, a dihedral angle, a horizontal dipole, a cylinder, a left helix and a right helix;

the third feature vector determining unit is used for calculating similarity parameters of the target and each standard body according to the polarized scattering vector of the target and the polarized scattering vector of the standard body, and determining the similarity parameters of the target and each standard body as third feature vectors; the formula for calculating the similarity parameter is as follows:

Compared with the prior art, the invention has the beneficial effects that:

the invention provides a fully-polarized radar target identification method and a system, which are characterized in that polarized electromagnetic waves are decomposed into horizontal polarized electromagnetic waves and vertical polarized electromagnetic waves which are perpendicular to each other, a polarized scattering matrix is determined according to the relation between an incident field and a scattering field of the electromagnetic waves, and the polarized scattering matrix is subjected to characteristic extraction by respectively adopting a Pauli decomposition method, an H, alpha, A decomposition method and a structural similarity parameter characteristic extraction method to obtain a first characteristic vector, a second characteristic vector and a third characteristic vector of target fully-polarized HRRP data; the first feature vector, the second feature vector and the third feature vector are target full polarization features, and the target full polarization features are combined with a neural network to further realize identification of the full polarization radar target. The ship target HRRP information of the four polarization channels is used for extracting the target full polarization characteristics, and a single distance unit is selected as a measurement scale, so that more characteristics of the target full polarization HRRP can be reserved, the target identification accuracy is improved, and the classification effect is improved.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.

Fig. 1 is a flowchart of a target identification method for a fully-polarized radar according to an embodiment of the present invention;

fig. 2 is a structural diagram of a target identification system of a fully-polarized radar according to a first embodiment of the present invention;

FIG. 3 is a diagram of a residual error structure according to a second embodiment of the present invention;

FIG. 4 is a diagram of a second embodiment of a rolling module;

FIG. 5 is a block diagram of a neural network model according to a second embodiment of the present invention;

fig. 6 is a flowchart of target identification based on polarization feature extraction in the second embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.

Example one

Fig. 1 is a flowchart of a method for identifying a target of a fully-polarized radar in an embodiment of the present invention, and as shown in fig. 1, the method for identifying a target of a fully-polarized radar includes:

step 101: the polarized electromagnetic wave is acquired and decomposed into a horizontally polarized electromagnetic wave and a vertically polarized electromagnetic wave that are perpendicular to each other.

Step 102: determining a polarization scattering matrix according to the relationship between the incident field of the horizontally polarized electromagnetic wave and the scattering field of the horizontally polarized electromagnetic wave and according to the relationship between the incident field of the vertically polarized electromagnetic wave and the scattering field of the vertically polarized electromagnetic wave; the elements of the polarization scattering matrix are respectively a horizontal co-polarization term, a vertical co-polarization term and two cross-polarization terms corresponding to the HRRP data of the four polarization channels.

Step 102, specifically comprising:

wherein S is a polarization scattering matrix,

is the incident field of the horizontally polarized electromagnetic wave,

is the incident field of a vertically polarized electromagnetic wave,

is the scattered field of the horizontally polarized electromagnetic wave,

Step 103: extracting the characteristics of the polarization scattering matrix by adopting a Pauli decomposition method to obtain a first characteristic vector of target full polarization HRRP data; the first feature vector reflects the fraction of scatter types in the target echo.

Step 103, specifically comprising:

expressing the polarization scattering matrix by Pauli base to obtain S ═ aX₀+bX₁+cX₂Determining a, b and c as a first feature vector; wherein, X₀、X₁And X₂The first term, the second term and the third term of Pauli base are respectively, a, b and c are respectively the ratio of three scattering types in the target echoWeighing;

acquiring HRRP data of four polarization channels; HRRP data of four polarization channels are respectively s_HH(n)、s_HV(n)、s_VH(n) and s_VV(N), wherein N is 0,1,2, and N-1, N is the number of radar range units of the HRRP data, and s is the number of radar range units of the HRRP data_HH(n) is the horizontal co-polarization term of the nth radar range unit, s_HV(n) is the first cross-polarization term, s, of the nth radar range unit_VH(n) is the second cross-polarization term, s, of the nth radar range unit_VV(n) is the vertical co-polarization term of the nth radar range unit;

determining a first feature vector from the HRRP data for the four polarization channels using the following equation:

Step 104: extracting the characteristics of the polarization scattering matrix by adopting an H, alpha and A decomposition method to obtain a second characteristic vector of the target full polarization HRRP data; the second feature vector reflects the probability of the occurrence of scattering mechanisms.

Step 104, specifically comprising:

T(n)＝k(n)·k^H(n)

carrying out eigenvalue decomposition on the coherent matrix to obtain

respectively calculating a scattering entropy, a scattering angle and an anisotropy according to the probability of the occurrence of a scattering mechanism, and determining the scattering entropy, the scattering angle and the anisotropy as second feature vectors; the scattering angle comprises a dominant scattering angle, and the dominant scattering angle is a scattering angle corresponding to the maximum value in the characteristic values of the coherent matrix;

wherein the content of the first and second substances,

Step 105: extracting the characteristics of the polarization scattering matrix by adopting a characteristic extraction method of the structural similarity parameters to obtain a third characteristic vector of the target full polarization HRRP data; the third feature vector reflects the similarity of scattering features between the two targets.

Step 105, specifically comprising:

Step 106: and training the neural network model by taking the first feature vector, the second feature vector and the third feature vector of the target fully-polarized HRRP data as input and the radar target class as output to obtain a trained neural network model, and identifying the fully-polarized radar target according to the trained neural network model.

Fig. 2 is a structural diagram of a fully polarized radar target identification system in an embodiment of the present invention. As shown in fig. 2, a fully polarized radar target identification system includes:

an electromagnetic wave decomposition module 201 for acquiring the polarized electromagnetic wave and decomposing the polarized electromagnetic wave into a horizontal polarized electromagnetic wave and a vertical polarized electromagnetic wave perpendicular to each other.

A polarized scattering matrix determining module 202, configured to determine a polarized scattering matrix according to a relationship between an incident field of the horizontally polarized electromagnetic wave and a scattering field of the horizontally polarized electromagnetic wave, and according to a relationship between an incident field of the vertically polarized electromagnetic wave and a scattering field of the vertically polarized electromagnetic wave; the elements of the polarization scattering matrix are respectively a horizontal co-polarization term, a vertical co-polarization term and two cross-polarization terms corresponding to the HRRP data of the four polarization channels.

The polarized scattering matrix determining module 202 specifically includes:

wherein S is a polarization scattering matrix,

is the incident field of the horizontally polarized electromagnetic wave,

is the incident field of a vertically polarized electromagnetic wave,

is the scattered field of the horizontally polarized electromagnetic wave,

The first feature vector determining module 203 is configured to perform feature extraction on the polarization scattering matrix by using a Pauli decomposition method to obtain a first feature vector of the target fully-polarized HRRP data; the first feature vector reflects the fraction of scatter types in the target echo.

The first feature vector determining module 203 specifically includes:

polarized scattering matrix processing sheetAn element for representing the polarization scattering matrix by Pauli basis to obtain S ═ aX₀+bX₁+cX₂Determining a, b and c as a first feature vector; wherein, X₀、X₁And X₂The first term, the second term and the third term of Pauli base are respectively, and a, b and c are respectively the proportion of three scattering types in the target echo;

the HRRP data acquisition units of the four polarization channels are used for acquiring the HRRP data of the four polarization channels; HRRP data of four polarization channels are respectively s_HH(n)、s_HV(n)、s_VH(n) and s_VV(N), wherein N is 0,1,2, and N-1, N is the number of radar range units of the HRRP data, and s is the number of radar range units of the HRRP data_HH(n) is the horizontal co-polarization term of the nth radar range unit, s_HV(n) is the first cross-polarization term, s, of the nth radar range unit_VH(n) is the second cross-polarization term, s, of the nth radar range unit_VV(n) is the vertical co-polarization term of the nth radar range unit;

a first feature vector determining unit, configured to determine a first feature vector according to HRRP data of four polarization channels by using the following formula:

The second feature vector determining module 204 is configured to perform feature extraction on the polarization scattering matrix by using an H, α, a decomposition method to obtain a second feature vector of the target fully-polarized HRRP data; the second feature vector reflects the probability of the occurrence of scattering mechanisms.

The second feature vector determining module 204 specifically includes:

the polarization scattering vector determining unit is used for determining the polarization scattering vector of each radar range unit according to the first feature vector; the polarization scattering vector k (n) of the nth radar range bin is as follows:

the coherent matrix determining unit is used for determining a coherent matrix of each radar range unit according to the polarization scattering vector; the coherence matrix t (n) for the nth radar range bin is as follows:

T(n)＝k(n)·k^H(n)

the second feature vector determining unit is used for respectively calculating the scattering entropy, the scattering angle and the anisotropy according to the probability of the occurrence of the scattering mechanism and determining the scattering entropy, the scattering angle and the anisotropy as second feature vectors; the scattering angle comprises a dominant scattering angle, and the dominant scattering angle is a scattering angle corresponding to the maximum value in the characteristic values of the coherent matrix;

wherein the content of the first and second substances,

A third feature vector determining module 205, configured to perform feature extraction on the polarization scattering matrix by using a feature extraction method of the structural similarity parameter, so as to obtain a third feature vector of the target fully-polarized HRRP data; the third feature vector reflects the similarity of scattering features between the two targets.

The third feature vector determination module specifically includes:

the third feature vector determining unit is used for calculating similarity parameters of the target and each standard body according to the polarization scattering vector of the target and the polarization scattering vector of the standard body, and determining the similarity parameters of the target and each standard body as third feature vectors; the formula for calculating the similarity parameter is as follows:

And the fully-polarized radar target identification module 206 is configured to train the neural network model by taking the first feature vector, the second feature vector and the third feature vector of the target fully-polarized HRRP data as inputs and taking the radar target category as an output, obtain a trained neural network model, and perform fully-polarized radar target identification according to the trained neural network model.

Example two

The polarization of an electromagnetic wave is defined as the way in which the orientation of the electric field vector E in space of the electromagnetic wave varies with time. For planar electromagnetic waves, electromagnetic waves e of arbitrary polarization_HVCan be described by decomposing into two mutually perpendicular linear polarization components, namely:

in the formula (1), H and V of lower subscripts represent a horizontal polarization component and a vertical polarization component; a is_HAnd

respectively representing the amplitude and phase in the horizontally polarized component of the planar electromagnetic wave; a is_VAnd

it represents the amplitude and phase, respectively, in the vertically polarized component of the planar electromagnetic wave.

The polarization scattering matrix S represents the incident field E of the electromagnetic wave at the targetⁱAnd the scattered field E^sThe relationship between them can be expressed as:

E^s＝SEⁱ (2)

when the radar seeker detects a target, the distance between the radar and the target is long, so that the electromagnetic wave at the target can be regarded as a plane wave. Under the horizontal polarization and vertical polarization bases, the polarization scattering matrix of the target can be represented by a 2 × 2 matrix, and then equation (2) becomes:

as can be seen from equation (3), when the radar guide head of the broadband full polarization system detects a target, HRRP information in 4 transmit-receive polarization states, i.e., HH, HV, VH, and VV, can be obtained. In the polarization scattering matrix S, S_HHAnd S_VVBeing co-polar terms, S_HVAnd S_VHIs a cross-polarization term. For the target of the linear scatterer, it can be known from the reciprocal theorem that the single-station scattering matrix is symmetric, i.e. S_HV＝S_VH. (compared with the HRRP obtained by a single-polarization radar, the fully-polarized HRRP can obtain richer scattering information of the target, and the scattering characteristic difference between different targets is reflected in a vector form to the maximum extent.)

Feature extraction is performed by a polarization scattering matrix S, which comprises a Pauli decomposition base and a H, alpha, A, alpha base₁The physical significance of the extracted features is clear and can reflect the target characteristics from different angles based on 3 aspects of the structural similarity parameter.

1) Polarization scattering matrix characteristic extraction based on Pauli decomposition

Polarization scattering matrix feature extraction based on Pauli decomposition is a coherent decomposition method, and the total backscattering of a target is represented by weighted summation of several typical scattering mechanisms. Other common coherent decomposition methods are the method based on Krogager decomposition and the method based on Cameron decomposition, which are more intuitive than others.

Under the linear orthogonal basis of vertical and horizontal polarization, the Pauli basis can be expressed as:

in the case of satisfying the reciprocity theorem (S)_HV＝S_VH) In this case, Pauli base can be simplified to X₀、X₁、 X₂The three terms, i.e., the polarization scattering matrix, are expressed as:

combining formula (4) and formula (5) gives:

a. b and c are complex numbers respectively representing the proportion of the three scattering types in the target echo, | a²、 |b|²And | c |)²The energies of the three scattering types contained in the target echo are respectively represented.

Pauli bases are a complete set of orthogonal bases and have the advantage of simple form. Pauli decomposition can be carried out on HRRP data with noise, namely the Pauli decomposition has certain anti-interference capability. Meanwhile, Pauli decomposition only corresponds to two scattering types, namely odd scattering and even scattering, and cannot describe more complex scattering characteristics of the target. Assuming that HRRP data of each polarization channel is s_HH(n)、s_HV(n)、s_VH(n)、s_VV(N), wherein N is 0,1, 2. When a single distance unit is selected as the metric, polarization scattering characteristics a, b, c based on Pauli decomposition can be obtained by equation (9).

Through the data processing, 3 feature vectors of a, b and c of the target fully-polarized HRRP data can be obtained.

2) Based on H, alpha, A, alpha₁Decomposed polarization scattering matrix feature extraction

Representing a, b, c features as a vector, and coefficients

Extracting, and putting corresponding positions in other contents to form vectors, so that a polarization scattering vector can be obtained:

the coherence matrix can also be defined as follows:

T＝k·k^H (11)

in the formula (11), T is a semi-positive Hermit matrix, and T can be obtained by performing characteristic decomposition on T:

in formula (12), T_i＝λ_iμ_i·μ_i ^H(i ═ 1,2,3) respectively correspond to a steady-state target; lambda [ alpha ]_iIs the i-th eigenvalue of the coherence matrix T, and₁≥λ₂≥λ₃≥0；μ_ias a characteristic value λ_iThe corresponding feature vector, expressed as:

in the formula (13), α_iAnd beta_iRespectively representing the scattering mechanism and the orientation angle of the target;

denotes S_HH+S_VVThe phase of (d);_idenotes S_HH+S_VVAnd S_HH-S_VVA phase difference therebetween; gamma ray_iThen represents S_HH+S_VVAnd S_HVThe phase difference between them.

According to the probability of occurrence of each scattering mechanism

The following 3 characteristic parameters can be defined:

h is scattering entropy which describes the proportion of different scattering mechanisms in the total scattering process and also describes the randomness of target scattering, and the value interval is [0,1 ]. When the value of H is increased from 0 to 1, the scattering of the target changes from a fully polarized state to a fully unpolarized state. Alpha is a scattering angle, describes the degree of freedom inside the target, and simultaneously reflects the main scattering mechanism of the target, and the value interval is [0 degrees, 90 degrees ]. When the value of alpha is increased from 0 degrees to 90 degrees, the main scattering mechanism of the target is changed from isotropic surface scattering into dihedral angle scattering. A is the degree of anisotropy, which describes the degree of anisotropy of the scattering of the target.

α₁The scattering angle corresponding to the maximum value in the eigenvalues of the coherence matrix is called the dominant scattering angle, and it represents the dominant scattering mechanism of the target echo signal. Thus, it can be retained as an important feature.

When a single distance unit is selected as a measurement scale, H, alpha, A, alpha of the HRRP data of the radar target of the broadband full-polarization system₁The feature extraction process can be summarized as shown in table 1:

TABLE 1H, α, A, α₁Feature extraction process

The H, alpha, A, alpha of the target full polarization HRRP data can be obtained through the data processing₁There are 4 feature vectors in total.

3) Polarized scattering matrix feature extraction based on structural similarity parameters

The polarization scattering matrix characteristic extraction based on the structural similarity parameter reflects the similarity of scattering characteristics between two targets, and the parameter is independent of the attitude angle of the targets and the power of echo signals.

From the above, the polarization scattering vector is obtained

Similarity parameters between different scattering matrices can be obtained from the polarization scattering vectors, and the similarity parameter r is defined as follows:

in the formula (17), the vector k₁And k₂Respectively a polarization scattering matrix S₁And S₂The polarized scattering vector of (1).

Representing the sum of the squares of the elemental modes of the polarization scattering vector.

The similarity parameters of the target and the standard body can be obtained by combining the formula (17) and the polarization scattering matrix of the common standard body. The polarization scattering matrix and Pauli-based polarization scattering vector for the common standard are shown in Table 2.

TABLE 2 polarization Scattering matrix for common standards and polarization Scattering vector under Pauli basis

And (3) bringing the polarization scattering vector of each standard body under Pauli basis into a formula (17), so as to obtain the structural similarity parameters of the target and different standard bodies, wherein the structural similarity parameters are respectively as follows:

(1) structural similarity parameters of target and plate:

(2) structural similarity parameters of the target to dihedral:

(3) structural similarity parameters of the target and horizontal dipoles:

(4) structural similarity parameters of the target and cylinder:

(5) structural similarity parameters of target and left helix:

(6) structural similarity parameters of target and right helix:

selecting a single distance unit as a measurement scale, calculating the structural similarity parameter of the target for the polarization scattering matrix of each distance unit in turn, and obtaining the HRRP data s of each polarization channel_HH(n)、 s_HV(n)、s_VH(n)、s_VV(n) the structural similarity parameters of the target and the plate are:

the structural similarity parameters between the target and other standards can be obtained by referring to the formulas (19 to 23). R of target full polarization HRRP data can be obtained through the data processing_plane、r_dihedral、r_dipole、r_cylinder、r_lhelix、 r_rhelixThere are 6 feature vectors.

By performing feature extraction in the above 3 ways on the ship target fully polarized HRRP data, each ship target can finally obtain 13 feature vectors, as shown in table 3.

TABLE 3 naval vessel target feature vector set

Table 1Ship targetfeature vector set

Pauli decomposition, H, alpha, A, alpha are carried out on the polarized scattering matrix₁The physical meanings of the obtained feature vectors are clear by decomposing and extracting the features of 3 aspects of the structural similarity parameters, and the target characteristics can be reflected from different angles.

The depth of the neural network is important, and the deep convolutional neural network can extract and fuse features of different layers for end-to-end target identification. However, the network layer number is deepened, which causes the problem of saturated recognition accuracy, and the residual structure is generally introduced to overcome the problem, and the structure is shown in fig. 3.

The residual block in the residual structure is composed of convolution layers, and the number of convolution layers in the figure is 2. The output of the residual structure is the sum of the input features and the output of the last convolutional layer, represented by equation (25)

x_l+1＝F(x_l)+x_l (25)

In the formula, x_l、x_l+1Respectively representing input and output characteristics of the residual error structure of the l layer; f (x)_l) Representing a mapping of the residual block.

Studies have shown that F (x) is mapped by fitting_l) Instead of the required mapping F (x)_l)+x_lThe problem of saturated accuracy of deep network identification can be effectively relieved. In extreme cases, if the network has extracted the optimal features required for classification, the residual structure only needs to perform the identity mapping of the jump connection to ensure the highest identification accuracy. For neural networks, zeroing the residual block is more efficient than fitting an identity map using a multi-layer neural network.

In order to facilitate understanding of the neural network classifier used in the present invention, a detailed structure of the fusion convolution module and the improved residual structure neural network proposed for vessel target single polarization channel identification is briefly introduced here, as shown in fig. 4. In fig. 4, mx 1 × N indicates one-dimensional data with an input feature of mx 1, the number of feature layers is N, and s is the moving step size of the convolution kernel. The convolution module is set to be a highly modularized network structure, and the expandability is strong. The characteristics extracted by the upper network are used as the input of the network of the upper network, and the input passes through 2 branches. In the left branch, firstly, features between different layers are fused by using a convolution kernel of 1 × 1, the fused features are equally divided into a plurality of branches on the layer number, each branch has 3-layer features, each branch is respectively subjected to feature extraction by using a convolution kernel of 3 × 1, the step length is 2, the feature layer number is unchanged, and the dimensionality is halved. And then splicing the branch features, determining the size of x according to the complexity of a classification task (the structure is similar to an inclusion structure, but the size and the number of convolution kernels of each branch in the inclusion are customized step by step, uniformly selecting a small-scale convolution kernel of 3 multiplied by 1 to reduce the structural design difficulty, and simultaneously ensuring the identification effect.) the spliced features use the convolution kernel of 1 multiplied by 1 to perform feature fusion and increase the number of feature layers, and the features are divided into two parts according to the layers to prepare for the feature fusion of the two subsequent branches. And the right branch directly uses a convolution kernel of 1 multiplied by 1 to carry out feature fusion on input, increases the number of feature layers, simultaneously divides the features into two parts according to the number of the feature layers, and carries out addition and splicing operation on the features corresponding to the left branch.

The output of the convolution module is reduced by half in characteristic dimension and doubled in layer number relative to the input. The effect of the right branch is similar to that of a residual error network, each layer of the network module can acquire information from a loss function and an original input signal, characteristics and gradients can be more effectively transmitted, the utilization rate of shallow characteristics is improved, and the problems of gradient loss and identification rate saturation which are possibly generated along with continuous deepening of the network are solved.

The loss function is used to measure the difference between the predicted value and the true value, and is generally denoted by L (y _, y), where y _ denotes the predicted value and y denotes the true value. (for multi-class convolutional neural networks, Softmax Loss (SL) is commonly used as a Loss function, but from a clustering perspective, SL-extracted features can be larger in intra-class distance than in inter-class distance

For the functional form of Softmax, features can be promoted to be closer by enhancing boundary constraint between different target categories and increasing the distance between the categories. The calculation formula of the Softmax loss function is as follows:

where x is the input to the last fully connected layer,

represents the ith deep feature and belongs to the yi category, and d represents the dimension of the feature;

weight matrix representing the last full connection layer

Column j of (1);

representing the weighting result of the ith sample target feature; m represents the number of samples in each batch in the network training process; n represents the number of categories of objects.

The loss function fuses boundary constraint and center clustering, and the specific expression is as follows:

wherein s is used to control the cosine distance between features, which embodies the similarity of features; mu is used for controlling the distance between the characteristic edges;

denotes the y th_iThe center of the features of each of the object classes,

continuously updated as the depth characteristics of each batch of data change. L is_AMSAnd psi (theta) is introduced to be cos theta-mu to restrict the inter-class distance, and the recognition effect is improved by increasing the inter-class distance of the features. L is_CA class center is constructed for each target class feature, and target features far from the class center are punished. The intra-class distance of the target features is more compact, and the purposes of reducing the intra-class distance and increasing the inter-class distance are achievedThe effect of the detachment. The update formula for the category center is as follows:

when y is_iFor class j targets, the identification is correct, where (-) equals 1, otherwise (-) equals 0. In a joint loss function L_AMSCUnder the constraints, the training and learning process of the model can be summarized as shown in table 4:

TABLE 4 training and learning Process of the model

The model block diagram of the invention is shown in fig. 5, and mainly comprises an initial convolution layer, a plurality of convolution modules with the same topology structure connected in sequence and two final full-connection layers. The dimensionality of the last full-connection layer is 2, so that the characteristics extracted by the model can be visualized conveniently, and the clustering effect of the characteristics can be analyzed.

The numbers in parentheses indicate the data dimension of the HRRP sample after passing through the layer, as in fig. 4. And the output data dimensionality of the continuous convolution modules and the first full-connection layer is determined according to the number of the convolution modules. The result of the final output layer is one-dimensional data corresponding to the target class, here the number of target classes 3. The initial convolution layer in the model selects a one-dimensional convolution kernel with the dimension of 7 multiplied by 1, and the convolution kernel with relatively large dimension is selected at the first layer of the network, so that the extraction of the corresponding characteristics such as contour, texture and the like in the target HRRP data is facilitated. And carrying out batch normalization and Relu activation on the extracted features after each convolution operation in the model.

The model parameters are kept unchanged when the ship target single polarization HRRP data are aimed at, but because the ship target full polarization HRRP data are processed simultaneously, after the full polarization HRRP data are subjected to feature extraction, the HRRP of each data set is changed from one-dimensional data to two-dimensional data. Therefore, when ship target complete polarization HRRP data is processed, the structure of the neural network model needs to be correspondingly adjusted. The convolution kernels in the network structure are expanded into two-dimensional convolution kernels by one-dimensional convolution kernels, the size of the convolution kernels of the initial convolution layer is changed from 7 x 1 to 7 x 3, and the size of the convolution kernels in the convolution module is changed from 3 x 1 to 3 x 3. The step size of the convolution kernel is changed from one dimension to two dimensions correspondingly, the step size of the convolution kernel is set to be (2,1), namely the step size of the convolution kernel moving in the distance unit direction is 2, and the step size of the convolution kernel moving among data channels is 1. The convolution kernel of size 1 × 1 in the convolution block corresponds to a step size set to (1, 1). The pooling block in the pooling layer also corresponds to a change from one-dimensional 3 × 1 to two-dimensional 3 × 3, and the step size is changed to s ═ 2, 1. In summary, the flow chart of the method of the present invention is shown in fig. 6.

The invention provides a high-efficiency extensible neural network by improving a residual error structure aiming at ship target fully polarized HRRP data, the structure is further extended on the basis of single-polarized HRRP data processing, a convolution kernel in the network structure is extended from a one-dimensional convolution kernel to a two-dimensional convolution kernel, and the convolution kernel size of an initial convolution kernel and the convolution kernel size in a convolution module are correspondingly extended. The problems that gradient explosion and disappearance can occur in the back propagation process of the convolutional neural network during training and the network recognition rate is saturated along with the increase of the network depth can be solved. Meanwhile, center clustering and feature boundary constraint are fused, a new combined loss function is provided, and the identification accuracy is further improved. Through the design of a modular structure, the model can be efficiently expanded when aiming at single-polarization or full-polarization target HRRP data so as to adapt to classification tasks with different difficulties. The method aims at the problem that when separability features are extracted from target full-polarization HRRP, the specific features of all distance units cannot be reserved by using all distance units as measurement scales. According to the method, a single distance unit is selected as a measurement scale when the HRRP information of the ship targets of the four polarization channels is comprehensively utilized, and the specific characteristics of each distance unit can be reserved to the maximum extent.

The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In summary, this summary should not be construed to limit the present invention.

Claims

1. A full polarization radar target identification method is characterized by comprising the following steps:

2. The method according to claim 1, wherein the determining a polarization scattering matrix according to a relationship between an incident field of the horizontally polarized electromagnetic wave and a scattered field of the horizontally polarized electromagnetic wave and according to a relationship between an incident field of the vertically polarized electromagnetic wave and a scattered field of the vertically polarized electromagnetic wave specifically comprises:

wherein S is a polarization scattering matrix,

is the incident field of the horizontally polarized electromagnetic wave,

is the incident field of a vertically polarized electromagnetic wave,

is the scattered field of the horizontally polarized electromagnetic wave,

3. The method for identifying the fully polarimetric radar target according to claim 2, wherein the performing feature extraction on the polarimetric scattering matrix by using a Pauli decomposition method to obtain a first feature vector of target fully polarimetric HRRP data specifically comprises:

expressing the polarization scattering matrix by Pauli base to obtain S ═ aX₀+bX₁+cX₂Determining a, b and c as a first feature vector; wherein, X₀、X₁And X₂The first term, the second term and the third term of Pauli base are respectively, and a, b and c are respectively the proportion of three scattering types in the target echo;

4. The method for identifying the fully polarimetric radar target according to claim 3, wherein the performing feature extraction on the polarimetric scattering matrix by using an H, α, A decomposition method to obtain a second feature vector of target fully polarimetric HRRP data specifically comprises:

T(n)＝k(n)·k^H(n)

performing eigenvalue decomposition on the coherent matrix to obtain

wherein the content of the first and second substances,

wherein H (n) is the scattering entropy of the nth radar range unit, alpha (n) is the scattering angle of the nth radar range unit, A (n) is the anisotropy of the nth radar range unit, and alpha_j(n) is the nth radar rangeThe scatter angle corresponding to the jth eigenvalue of the cell.

5. The method for identifying the fully polarimetric radar target according to claim 4, wherein the performing feature extraction on the polarization scattering matrix by using the feature extraction method of the structural similarity parameter to obtain a third feature vector of the target fully polarimetric HRRP data specifically comprises:

6. A fully-polarized radar target identification system, comprising:

7. The system according to claim 6, wherein the polarization scattering matrix determination module specifically includes:

wherein S is a polarization scattering matrix,

is the incident field of the horizontally polarized electromagnetic wave,

is the incident field of a vertically polarized electromagnetic wave,

is the scattered field of the horizontally polarized electromagnetic wave,

8. The fully-polarized radar target recognition system of claim 7, wherein the first feature vector determination module specifically comprises:

the HRRP data acquisition units of the four polarization channels are used for acquiring the HRRP data of the four polarization channels; HRRP data of the four polarization channels are respectively s_HH(n)、s_HV(n)、s_VH(n) and s_VV(N), wherein N is 0,1,2, and N-1, N is the number of radar range units of the HRRP data, and s is the number of radar range units of the HRRP data_HH(n) is the horizontal co-polarization of the nth radar range unitItem, s_HV(n) is the first cross-polarization term, s, of the nth radar range unit_VH(n) is the second cross-polarization term, s, of the nth radar range unit_VV(n) is the vertical co-polarization term of the nth radar range unit;

9. The fully-polarized radar target recognition system of claim 8, wherein the second feature vector determination module specifically comprises:

T(n)＝k(n)·k^H(n)

wherein the content of the first and second substances,

10. The fully-polarized radar target recognition system of claim 9, wherein the third feature vector determination module specifically comprises: