Full polarization synthetic aperture radar image classification method, system and terminal

Abstract

The present invention relates to a full polarization synthetic aperture radar image classification method, comprising: reading in a target coherence matrix T of an image to be classified, performing de-orientation processing to obtain a de-oriented polarization coherence matrix #imgabs0#; performing de-asymmetry processing on the de-oriented polarization coherence matrix #imgabs1# to obtain a de-asymmetry processed polarization coherence matrix T sym ; According to the polarization coherence matrix T after de-asymmetry processing sym , estimate the adaptive estimation parameters of the X-Bragg model, and obtain the optimal parameters δ, Δ; classify the image to be classified according to the attributes according to the optimal parameters δ, Δ obtained above. The present invention also relates to a full polarization synthetic aperture radar image classification system and terminal. The present invention can provide a full polarization synthetic aperture radar image classification with good classification effect.

Description

Technical Field

[0001] The invention relates to a full polarization synthetic aperture radar image classification method, system and terminal. Background Art

[0002] Radar is an important means of detecting ground objects. With the deepening of people's understanding of the target electromagnetic polarization scattering mechanism and the development of radar polarization measurement technology, radar polarization research has gradually attracted attention. Polarization radar obtains the target polarization matrix by measuring the transformation relationship between the incident wave and the scattered wave electric field vector or Stokes vector. For a single target, it is represented by a 2*2 Sinclair scattering matrix, and for distributed targets, it is represented by a 3*3 polarization coherence matrix or polarization covariance matrix. By analyzing and processing these matrices, the physical scattering mechanism and geometric structure information of the target can be inverted, and the target can be classified.

[0003] The extraction of polarization parameters is currently mainly based on polarization decomposition. For distributed targets, the main polarization decompositions include eigendecomposition-based decomposition methods and model-based decomposition methods. Among them, the eigendecomposition-based decomposition method can extract the scattering angle alpha and polarization scattering entropy H, and has been widely used in ground feature attribute analysis and classification. The model-based decomposition decomposes the target into the coherent sum of multiple scattering components such as surface scattering, secondary scattering, volume scattering, and helical scattering, and judges the main scattering mechanism of the target based on the size of the scattering component.

[0004] The X-Bragg scattering model is also called the extended Bragg scattering model, which consists of two parameters, namely the shape parameter δ and the truncation parameter Δ, which are similar in meaning to the alpha angle and the scattering entropy H in the Cloude-Pottier decomposition. Moreover, with the change of parameters δ and Δ, the X-Bragg scattering model can be converted into surface scattering, secondary scattering and volume scattering models. Therefore, the X-Bragg model is a bridge between the characteristic decomposition method and the model-based decomposition method, which is very helpful for in-depth understanding of the two decomposition methods and the relationship between the two types of decomposition. However, there is currently little research on the X-Bragg model, lack of effective parameter estimation methods, and no related applications.

[0005] The existing solutions to the X-Bragg model are incomplete and only consider some characteristics of the polarization coherence matrix. After analysis, the main reasons for the difficulty in solving the X-Bragg model are twofold: (1) The X-Bragg model is an idealized standard model with an orientation angle of 0 and no asymmetric scattering part, but for general targets, its polarization coherence matrix is ​​difficult to satisfy; (2) For general targets, which are composed of a variety of scattering targets and are the sum of multiple scattering mechanisms, there is a certain difference between a single X-Bragg model and the actual polarization coherence matrix, and it is impossible to find two definite parameters that make the X-Bragg model completely consistent with the actual polarization coherence matrix. Summary of the invention

[0006] In view of this, it is necessary to provide a method, system and terminal for classifying full-polarization synthetic aperture radar images.

[0007] The present invention provides a method for classifying full polarization synthetic aperture radar images, which comprises the following steps: a. reading in a target coherence matrix T of an image to be classified, performing de-orientation processing, and obtaining a de-orientated polarization coherence matrix; b. Polarization coherence matrix after deorientation Perform de-asymmetry processing to obtain the polarization coherence matrix T after de-asymmetry processing sym ; c. According to the polarization coherence matrix T after de-asymmetry processing sym , estimate the adaptive estimation parameters of the X-Bragg model and obtain the optimal parameters δ and Δ, where δ is the shape factor and Δ is the truncation parameter; d. classify the image to be classified according to its attributes based on the optimal parameters δ and Δ obtained above.

[0008] Preferably, the step a specifically comprises:

[0009] Assume that the input polarization target coherence matrix is ​​labeled as T and has the following form:

[0010]

[0011] Among them, A0, B0, B, C, D, E, F, G, H are Huynen parameters, all real numbers;

[0012] Orientation Angle Calculated by the following formula:

[0013]

[0014] Then the polarization coherence matrix after deorientation is It is expressed as:

[0015]

[0016] in, is the rotation operator;

[0017]

[0018] Preferably, the step b specifically includes:

[0019] Using the formula

[0020]

[0021] Among them, T sym is the symmetry component, T asym is the asymmetric component, and removing the asymmetry is to obtain the symmetric component, T sym and T asym Respectively expressed as:

[0022]

[0023] in:

[0024]

[0025]

[0026]

[0027] T sym This is the polarization coherence matrix after de-asymmetry processing.

[0028] Preferably, the step c specifically comprises:

[0029] The X-Bragg model is a truncated uniform distribution model, expressed as:

[0030]

[0031]

[0032]

[0033]

[0034] Among them, δ is the shape factor, Δ is the truncation parameter, δ and Δ are the parameters of the X-Bragg model adaptive estimation, according to T sym To estimate the parameters δ and Δ, the steps are as follows:

[0035] (1) Setting the initial value

[0036]

[0037]

[0038] (2) Using δ0 and Δ0 as initial inputs, Newton's method is applied to obtain the optimal parameters δ and Δ, where the optimal judgment criterion is:

[0039]

[0040] where Tr represents the trace of the polarization coherence matrix.

[0041] Preferably, the step d specifically comprises:

[0042] Based on the obtained optimal parameters δ and Δ, the images to be classified are classified according to their attributes, and the classification criteria are:

[0043] (1) 0≤δ≤0.8, Δ≤45, classified as surface scattering under low entropy;

[0044] (2) 0.8<δ<1.2, Δ≤45, classified as dipole scattering under low entropy;

[0045] (3) 1.2≤δ, Δ≤45, classified as even scattering under low entropy;

[0046] (4) 0≤δ≤0.8, Δ>45, classified as odd scattering under high entropy;

[0047] (5) 0.8<δ<1.2, Δ>45, classified as even scattering under high entropy;

[0048] (6)1.2≤δ, Δ>45, classified as dipole scattering under high entropy.

[0049] The present invention provides a full polarization synthetic aperture radar image classification system, which comprises a de-orientation processing module, a de-asymmetry processing module, a parameter estimation module, and a classification module, wherein: the de-orientation processing module is used to read in a target coherence matrix T of an image to be classified, perform de-orientation processing, and obtain a de-orientated polarization coherence matrix The de-asymmetry processing module is used to process the polarization coherence matrix after de-orientation. Perform de-asymmetry processing to obtain the polarization coherence matrix T after de-asymmetry processing sym The parameter estimation module is used to obtain the polarization coherence matrix T after the de-asymmetry processing. sym , the adaptive estimation parameters of the X-Bragg model are estimated to obtain the optimal parameters δ and Δ, where δ is the shape factor and Δ is the truncation parameter; the classification module is used to classify the image to be classified according to the attributes based on the optimal parameters δ and Δ obtained above.

[0050] Preferably, the de-orientation processing module is specifically used for:

[0051] Assume that the input polarization target coherence matrix is ​​labeled as T and has the following form:

[0052]

[0053] Among them, A0, B0, B, C, D, E, F, G, H are Huynen parameters, all real numbers;

[0054] Orientation Angle Calculated by the following formula:

[0055]

[0056] Then the polarization coherence matrix after deorientation is It is expressed as:

[0057]

[0058] in, is the rotation operator;

[0059]

[0060] Preferably, the de-asymmetry processing module is specifically used for:

[0061] Using the formula

[0062]

[0063] Among them, T sym is the symmetry component, T asym is the asymmetric component, and removing the asymmetry is to obtain the symmetric component, T sym and T asym Respectively expressed as:

[0064]

[0065] in:

[0066]

[0067]

[0068]

[0069] T sym This is the polarization coherence matrix after de-asymmetry processing.

[0070] Preferably, the parameter estimation module is specifically used for:

[0071] The X-Bragg model is a truncated uniform distribution model, expressed as:

[0072]

[0073]

[0074]

[0075]

[0076] Among them, δ is the shape factor, Δ is the truncation parameter, δ and Δ are the parameters of the X-Bragg model adaptive estimation, according to T sym Estimate the parameters δ and Δ, including:

[0077] (1) Setting the initial value

[0078]

[0079]

[0080] (2) Using δ0 and Δ0 as initial inputs, Newton's method is applied to obtain the optimal parameters δ and Δ, where the optimal judgment criterion is:

[0081]

[0082] where Tr represents the trace of the polarization coherence matrix.

[0083] Preferably, the classification module is specifically used for:

[0084] Based on the obtained optimal parameters δ and Δ, the images to be classified are classified according to their attributes, and the classification criteria are:

[0085] (1) 0≤δ≤0.8, Δ≤45, classified as surface scattering under low entropy;

[0086] (2) 0.8<δ<1.2, Δ≤45, classified as dipole scattering under low entropy;

[0087] (3) 1.2≤δ, Δ≤45, classified as even scattering under low entropy;

[0088] (4) 0≤δ≤0.8, Δ>45, classified as odd scattering under high entropy;

[0089] (5) 0.8<δ<1.2, Δ>45, classified as even scattering under high entropy;

[0090] (6)1.2≤δ, Δ>45, classified as dipole scattering under high entropy.

[0091] The present invention provides a terminal, comprising a processor and a memory coupled to the processor, wherein the memory stores program instructions for implementing the full polarization synthetic aperture radar image classification method; and the processor is used to execute the program instructions stored in the memory to perform full polarization synthetic aperture radar image classification.

[0092] The beneficial effects of this application include:

[0093] (1) First, the polarization coherence matrix is ​​de-oriented and de-asymmetric, and the remaining symmetric scattering components are compared with the X-Bragg model to maintain consistency in form;

[0094] (2) The adaptive parameter solution is achieved through nonlinear programming. The Wishart distance used as the optimal discrimination criterion has been widely used in the classification and clustering of polarimetric SAR data, with good theoretical and application effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] Figure 1 This is a flow chart of the full polarization synthetic aperture radar image classification method according to an embodiment of the present application;

[0096] Figure 2 This is a hardware architecture diagram of the full polarization synthetic aperture radar image classification system of an embodiment of the present application;

[0097] Figure 3 This is a schematic diagram of the terminal structure of an embodiment of the present application. DETAILED DESCRIPTION

[0098] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0099] See also Figure 1 , which is a flowchart of a preferred embodiment of the full polarization synthetic aperture radar image classification method of the present invention.

[0100] Step S1, read the target coherence matrix T of the image to be classified, perform de-orientation processing, and obtain the polarization coherence matrix after de-orientation Specifically:

[0101] Assume that the input polarization target coherence matrix is ​​labeled as T and has the following form:

[0102]

[0103] Among them, A0, B0, B, C, D, E, F, G, and H are all Huynen parameters, and are all real numbers.

[0104] Orientation Angle Calculated by the following formula:

[0105]

[0106] Then the polarization coherence matrix after deorientation is It is expressed as:

[0107]

[0108] in, is the rotation operator.

[0109]

[0110] in, Respectively represent the matrix after deorientation The element in the i-th row and j-th column of the matrix.

[0111] Step S2, the polarization coherence matrix after deorientation Perform de-asymmetry processing to obtain the polarization coherence matrix T after de-asymmetry processing sym Specifically:

[0112] Using the formula

[0113]

[0114] Among them, T sym is the symmetry component, T asym is the asymmetric component, and removing the asymmetry is to obtain the symmetric component, T sym and T asym Respectively expressed as:

[0115]

[0116] in:

[0117]

[0118]

[0119]

[0120] T sym This is the polarization coherence matrix after de-asymmetry processing.

[0121] Step S3, according to the polarization coherence matrix T after de-asymmetry processing sym , estimate the adaptive estimation parameters of the X-Bragg model and obtain the optimal parameters δ and Δ. Specifically include:

[0122] The X-Bragg model is a truncated uniform distribution model, expressed as:

[0123]

[0124]

[0125]

[0126]

[0127] Among them, δ is the shape factor, Δ is the truncation parameter, and δ and Δ are the parameters of the adaptive estimation of the X-Bragg model. sym To estimate the parameters δ and Δ, the steps are as follows:

[0128] (1) Setting the initial value

[0129]

[0130]

[0131] (2) Using δ0 and Δ0 as initial inputs, Newton's method is applied to obtain the optimal parameters δ and Δ, where the optimal judgment criterion is:

[0132]

[0133] where Tr represents the trace of the polarization coherence matrix.

[0134] Step S4, classify the images to be classified by attributes according to the optimal parameters δ and Δ obtained above. Specifically:

[0135] Based on the obtained optimal parameters δ and Δ, the images to be classified are classified according to their attributes, and the classification criteria are:

[0136] (1) 0≤δ≤0.8, Δ≤45, classified as surface scattering under low entropy;

[0137] (2) 0.8<δ<1.2, Δ≤45, classified as dipole scattering under low entropy;

[0138] (3) 1.2≤δ, Δ≤45, classified as even scattering under low entropy;

[0139] (4) 0≤δ≤0.8, Δ>45, classified as odd scattering under high entropy;

[0140] (5) 0.8<δ<1.2, Δ>45, classified as even scattering under high entropy;

[0141] (6)1.2≤δ, Δ>45, classified as dipole scattering under high entropy.

[0142] See also Figure 2, which is a hardware architecture diagram of the full polarization synthetic aperture radar image classification system 10 of the present invention. The system comprises: a de-orientation processing module 101, a de-asymmetry processing module 102, a parameter estimation module 103, and a classification module 104.

[0143] The de-orientation processing module 101 is used to read the target coherence matrix T of the image to be classified, perform de-orientation processing, and obtain the polarization coherence matrix after de-orientation. in:

[0144] Assume that the input polarization target coherence matrix is ​​labeled as T and has the following form:

[0145]

[0146] Among them, A0, B0, B, C, D, E, F, G, and H are all Huynen parameters, and are all real numbers.

[0147] Orientation Angle Calculated by the following formula:

[0148]

[0149] Then the polarization coherence matrix after deorientation is It is expressed as:

[0150]

[0151] in, is the rotation operator.

[0152]

[0153] in, Respectively represent the matrix after deorientation The element in the i-th row and j-th column of the matrix.

[0154] The de-asymmetry processing module 102 is used to process the polarization coherence matrix after de-orientation. Perform de-asymmetry processing to obtain the polarization coherence matrix T after de-asymmetry processing sym Specifically:

[0155] Using the formula

[0156]

[0157] Among them, T sym is the symmetry component, T asym is the asymmetric component, and removing the asymmetry is to obtain the symmetric component, T sym and T asym Respectively expressed as:

[0158]

[0159] in:

[0160]

[0161]

[0162]

[0163] T sym This is the polarization coherence matrix after de-asymmetry processing.

[0164] The parameter estimation module 103 is used to obtain the polarization coherence matrix T after the de-asymmetry processing. sym , estimate the adaptive estimation parameters of the X-Bragg model and obtain the optimal parameters δ and Δ. Specifically include:

[0165] The X-Bragg model is a truncated uniform distribution model, expressed as:

[0166]

[0167]

[0168]

[0169]

[0170] Among them, δ is the shape factor, Δ is the truncation parameter, and δ and Δ are the parameters of the adaptive estimation of the X-Bragg model. sym To estimate the parameters δ and Δ, the steps are as follows:

[0171] (1) Setting the initial value

[0172]

[0173]

[0174] (2) Using δ0 and Δ0 as initial inputs, Newton's method is applied to obtain the optimal parameters δ and Δ, where the optimal judgment criterion is:

[0175]

[0176] where Tr represents the trace of the polarization coherence matrix.

[0177] The classification module 104 is used to classify the images to be classified according to their attributes based on the optimal parameters δ and Δ obtained above. Specifically:

[0178] The classification module 104 classifies the image to be classified according to its attributes based on the obtained optimal parameters δ and Δ. The classification criteria are:

[0179] (1) 0≤δ≤0.8, Δ≤45, classified as surface scattering under low entropy;

[0180] (2) 0.8<δ<1.2, Δ≤45, classified as dipole scattering under low entropy;

[0181] (3) 1.2≤δ, Δ≤45, classified as even scattering under low entropy;

[0182] (4) 0≤δ≤0.8, Δ>45, classified as odd scattering under high entropy;

[0183] (5) 0.8<δ<1.2, Δ>45, classified as even scattering under high entropy;

[0184] (6)1.2≤δ, Δ>45, classified as dipole scattering under high entropy.

[0185] See also Figure 3 , is a schematic diagram of the terminal structure of an embodiment of the present application. The terminal 50 includes a processor 51 and a memory 52 coupled to the processor 51.

[0186] The memory 52 stores program instructions for implementing the above-mentioned full-polarization synthetic aperture radar image classification method.

[0187] The processor 51 is used to execute the program instructions stored in the memory 52 to perform full polarization synthetic aperture radar image classification.

[0188] The processor 51 may also be referred to as a CPU (Central Processing Unit). The processor 51 may be an integrated circuit chip having the ability to process signals. The processor 51 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0189] This application establishes an adaptive parameter estimation method for the X-Bragg model, and the "adaptive" is reflected in two aspects: (1) for the polarization coherence matrix, the purpose of de-orientation and de-asymmetric scattering is achieved through adaptive orientation angle calculation and asymmetric scatterer calculation, so that the remaining symmetric scattering components are consistent with the X-Bragg model in form; (2) through the nonlinear programming solution, the local optimal solution of the two parameters is adaptively obtained, so that the remaining symmetric scattering components are closest to the obtained local optimal X-Bragg model. And on the basis of solving the X-Bragg model, a full polarization synthetic aperture radar image classification with good classification effect is provided.

[0190] Although the present invention has been described with reference to the current preferred embodiments, those skilled in the art should understand that the above-mentioned preferred embodiments are only used to illustrate the present invention and are not used to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for classifying fully polarized synthetic aperture radar images, characterized in that: The method comprises the following steps: a. Read the target coherence matrix T of the image to be classified, perform de-orientation processing, and obtain the polarization coherence matrix after de-orientation b. Polarization coherence matrix after deorientation Perform de-asymmetry processing to obtain the polarization coherence matrix T after de-asymmetry processing sym ; c. According to the polarization coherence matrix T after de-asymmetry processing sym , estimate the adaptive estimation parameters of the X-Bragg model and obtain the optimal parameters δ and Δ, where δ is the shape factor and Δ is the truncation parameter; d. According to the optimal parameters δ and Δ obtained above, the images to be classified are classified according to their attributes; wherein: The step a specifically comprises: Assume that the input polarization target coherence matrix is ​​labeled as T and has the following form: Among them, A0, B0, B, C, D, E, F, G, H are Huynen parameters, all real numbers; Orientation Angle Calculated by the following formula: Then the polarization coherence matrix after deorientation is It is expressed as: in, is the rotation operator; in, Respectively represent the matrix after deorientation The element in the i-th row and j-th column of the matrix; The step b specifically comprises: Using the formula Among them, T sym is the symmetry component, T asym is the asymmetric component, and removing the asymmetry is to obtain the symmetric component, T sym and T asym Respectively expressed as: in: T sym That is the polarization coherence matrix after de-asymmetry processing; The step c specifically comprises: The X-Bragg model is a truncated uniform distribution model, expressed as: Among them, δ is the shape factor, Δ is the truncation parameter, δ and Δ are the parameters of the X-Bragg model adaptive estimation, according to T sym To estimate the parameters δ and Δ, the steps are as follows: (1) Setting the initial value (2) Using δ0 and Δ0 as initial inputs, Newton's method is applied to obtain the optimal parameters δ and Δ, where the optimal judgment criterion is: where Tr represents the trace of the polarization coherence matrix.

2. The method according to claim 1, characterized in that The step d specifically comprises: Based on the obtained optimal parameters δ and Δ, the images to be classified are classified according to their attributes, and the classification criteria are: (1) 0≤δ≤0.8, Δ≤45, classified as surface scattering under low entropy; (2) 0.8<δ<1.2, Δ≤45, classified as dipole scattering under low entropy; (3) 1.2≤δ, Δ≤45, classified as even scattering under low entropy; (4) 0≤δ≤0.8, Δ>45, classified as odd scattering under high entropy; (5) 0.8<δ<1.2, Δ>45, classified as even scattering under high entropy; (6)1.2≤δ, Δ>45, classified as dipole scattering under high entropy.

3. A full polarization synthetic aperture radar image classification system, characterized in that: The system comprises a de-orientation processing module, a de-asymmetry processing module, a parameter estimation module, and a classification module, wherein: The de-orientation processing module is used to read the target coherence matrix T of the image to be classified, perform de-orientation processing, and obtain the polarization coherence matrix after de-orientation. The de-asymmetry processing module is used to process the polarization coherence matrix after de-orientation. Perform de-asymmetry processing to obtain the polarization coherence matrix T after de-asymmetry processing sym ; The parameter estimation module is used to obtain the polarization coherence matrix T after de-asymmetry processing. sym , estimate the adaptive estimation parameters of the X-Bragg model and obtain the optimal parameters δ and Δ, where δ is the shape factor and Δ is the truncation parameter; The classification module is used to classify the images to be classified according to their attributes based on the optimal parameters δ and Δ obtained above; wherein: The de-orientation processing module is specifically used for: Assume that the input polarization target coherence matrix is ​​labeled as T and has the following form: Among them, A0, B0, B, C, D, E, F, G, H are Huynen parameters, all real numbers; Orientation Angle Calculated by the following formula: Then the polarization coherence matrix after deorientation is It is expressed as: in, is the rotation operator; in, Respectively represent the matrix after deorientation The element in the i-th row and j-th column of the matrix; The de-asymmetry processing module is specifically used for: Using the formula Among them, T sym is the symmetry component, T asym is the asymmetric component, and removing the asymmetry is to obtain the symmetric component, T sym and T asym Respectively expressed as: in: T sym That is the polarization coherence matrix after de-asymmetry processing; The parameter estimation module is specifically used for: The X-Bragg model is a truncated uniform distribution model, expressed as: Among them, δ is the shape factor, Δ is the truncation parameter, δ and Δ are the parameters of the X-Bragg model adaptive estimation, according to T sym Estimate the parameters δ and Δ, including: (1) Setting the initial value (2) Using δ0 and Δ0 as initial inputs, Newton's method is applied to obtain the optimal parameters δ and Δ, where the optimal judgment criterion is: where Tr represents the trace of the polarization coherence matrix.

4. The system according to claim 3, characterized in that The classification module is specifically used for: Based on the obtained optimal parameters δ and Δ, the images to be classified are classified according to their attributes, and the classification criteria are: (1) 0≤δ≤0.8, Δ≤45, classified as surface scattering under low entropy; (2) 0.8<δ<1.2, Δ≤45, classified as dipole scattering under low entropy; (3) 1.2≤δ, Δ≤45, classified as even scattering under low entropy; (4) 0≤δ≤0.8, Δ>45, classified as odd scattering under high entropy; (5) 0.8<δ<1.2, Δ>45, classified as even scattering under high entropy; (6)1.2≤δ, Δ>45, classified as dipole scattering under high entropy.

5. A terminal, characterized in that: The terminal includes a processor and a memory coupled to the processor, wherein: The memory stores program instructions for implementing the full polarization synthetic aperture radar image classification method according to any one of claims 1-2; The processor is used to execute the program instructions stored in the memory to perform full polarization synthetic aperture radar image classification.

Images