Far-field system response G matrix measurement method based on adaptive blind source separation

Abstract

The invention provides a far-field system response G matrix measurement method based on adaptive blind source separation. The method comprises the following steps: sequentially and uniformly placing noise sources in a view field at intervals, and respectively measuring to obtain a far-field system response G matrix; after the noise source is placed in the field of view every time, the noise source is measured by the following steps: receiving a microwave radiation signal emitted by the field of view as an observation signal x of a microwave remote sensing image; performing pre-whitening processing on the observation signal x to obtain a processed observation signal, extracting a noise source signal matrix Y from the observation signal, and realizing separation of a noise source and an environment from a signal level; and calculating the actual visibility of the noise source through the obtained signal matrix Y.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of remote sensing, and more particularly relates to a method for measuring a far-field system response G matrix based on adaptive blind source separation, which is suitable for a satellite system carrying an integrated aperture radiometer to carry out sea surface brightness temperature detection and inversion, and belongs to the technical field of satellite microwave ocean remote sensing. BACKGROUND

[0002] The definition of the G matrix first appeared in the process of developing the ESTAR one-dimensional system by Alan B. Tanner et al. They proposed a new inversion imaging algorithm suitable for an integrated aperture radiometer system, i.e., a G matrix-based inversion algorithm. The basic idea of the algorithm is to regard the imaging process of the system as a linear process, i.e.,

[0003] V(u, v) = ∫∫g(ξ, η)T(ξ, η)dξdη Equation (1-1)

[0004] where g(ξ, η) is the generalized system impulse response. The above equation is represented by a simplified linear operator G, i.e.,

[0005] V = GT Equation (1-2)

[0006] where V is the visibility function vector, G is the system impulse response matrix, and T is the brightness temperature.

[0007] In microwave remote sensing, the G matrix (or Green's function matrix) is commonly used to describe the propagation and scattering characteristics of electromagnetic waves. In radar imaging, the G matrix is used to describe the relationship of the return signal from the target object. Optimization of the G matrix can improve the resolution and quality of the image, helping to more accurately reconstruct the shape and features of the ground surface or target object. Therefore, the method for optimizing the measurement of the G matrix is very important for improving the quality of microwave remote sensing images.

[0008] The currently proposed G matrix measurement method is the distribution measurement method, i.e., the system elements such as the directional pattern of each unit antenna, the complex gain of the receiver channel, and the stripe elimination function are measured step by step, and finally combined into the system response G matrix. This method requires very high accuracy in measuring the directional pattern of each unit antenna and the stripe elimination function in actual operation, and the process is complex, the measurement cost is high, and the errors of each single measurement often accumulate, ultimately leading to the damage of the accuracy of the G matrix. SUMMARY

[0009] The technical problem solved by the present application is that, in view of the defects and improvement needs of the prior art, the present application provides a method for measuring a far-field system response G matrix based on adaptive blind source separation, which effectively removes the interference of noise in the environment on the visibility of the noise source.

[0010] The technical scheme of the present application is: a far-field system response G matrix measurement method based on adaptive blind source separation, wherein the noise source is placed in the field of view in sequence and uniformly spaced, and is measured respectively to obtain the far-field system response G matrix;

[0011] After placing the noise source in the field of view each time, the measurement is performed by the following method:

[0012] Receiving the microwave radiation signal emitted by the field of view as the observation signal x of the microwave remote sensing image;

[0013] Performing pre-whitening processing on the observation signal x to obtain the processed observation signal

[0014] Extracting the noise source signal matrix Y from the observation signal , and separating the noise source and the environment from the signal level;

[0015] The actual visibility of the noise source is calculated through the obtained signal matrix Y.

[0016] Preferably, the noise source is placed in the field of view in sequence and uniformly spaced by the following method:

[0017] Fixing the noise source in the far field, fixing the Y-shaped array on the turntable, the array is horizontal to the ground, and the output center of the noise source is directly opposite the center of the Y-shaped array, that is, the intersection of the three arms of the Y-shaped array;

[0018] Setting a three-dimensional rectangular coordinate system, the origin is set at the center of the Y-shaped array, the x-axis extends in the horizontal plane along the direction of one arm of the array; the y-axis is in the horizontal plane and is at a 90° angle with the x-axis, and faces the side; the z-axis is perpendicular to the ground and points upward;

[0019] Rotating the turntable to rotate the array around the z-axis to change the azimuth angle, so that the azimuth angle is sequentially increased from 0° to 360°; the elevation angle is changed by pitching up and down, so that the elevation angle is sequentially increased from 0° to 90°; then the whole turntable is rotated by 180° to measure the part from -90° to 0°; through the above operation, that is, the method of rotating the array while fixing the noise source, the noise-containing observation signal at each position in the field of view can be obtained.

[0020] Preferably, the processed observation signal

[0021] The row vectors of the observation signal x are averaged to obtain the averaged observation signal X;

[0022] The autocorrelation matrix R of the averaged observation signal X is constructed After performing eigenvalue decomposition on the autocorrelation matrix R , the diagonal matrix composed of the largest n eigenvalues is denoted as ΛS =diag{λ1≥λ2…≥λ n}, corresponding to n feature vectors V S ;

[0023] Calculate the autocorrelation matrix The average of the remaining mn small eigenvalues ​​is used as an estimate of the variance of the additive white noise.

[0024] according to Calculate the observed signal after pre-whitening.

[0025] m represents the autocorrelation matrix. The number of eigenvalues, where n is a preset positive integer.

[0026] Preferably, the acquisition of the observed signal is achieved in the following manner. Extract the noise source signal matrix Y:

[0027] For the observed signal Perform a time delay and calculate the autocorrelation matrix after the delay.

[0028] The autocorrelation matrix after delay is Perform eigenvalue decomposition;

[0029] according to Estimate the mixing matrix W; express The matrix formed by the eigenvectors of ;

[0030] according to Calculate the noise source signal matrix Y. represent The transpose of a matrix.

[0031] A far-field system response G matrix measurement system based on adaptive blind source separation includes a noise source, a turntable, and a measurement unit;

[0032] The noise source is fixed in the far field, and the Y-shaped array is fixed on the turntable. The array is horizontal with respect to the ground, and the output center of the noise source is directly opposite the center of the Y-shaped array, that is, the intersection of the three arms of the Y-shaped array.

[0033] By rotating the turntable, the noise sources are effectively moved at equivalent angles, thus placing them sequentially and evenly spaced within the field of view. After each rotation to its designated position, the measurement unit performs the measurement, and the far-field system response matrix G is obtained based on the final measurement results. The measurement unit's measurements include: receiving the microwave radiation signal emitted from the field of view as the observation signal x for the microwave remote sensing image; and performing pre-whitening processing on the observation signal x to obtain the processed observation signal. From the observed signal The noise source signal matrix Y is extracted to separate the noise source from the environment at the signal level; the actual visibility of the noise source is calculated using the obtained signal matrix Y.

[0034] Preferably, a three-dimensional rectangular coordinate system is set, with the origin set at the center of the Y-shaped array, the x-axis in the horizontal plane extending along one arm of the array; the y-axis in the horizontal plane, forming a 90° angle with the x-axis and pointing to the side; and the z-axis perpendicular to the ground and pointing upwards.

[0035] Rotating the turntable causes the array to rotate around the z-axis to change the azimuth angle, increasing it sequentially from 0° to 360°; tilting up and down changes the elevation angle, increasing it sequentially from 0° to 90°; then rotating the turntable as a whole by 180° measures the elevation angle from -90° to 0°; through the above operations, i.e., fixing the noise source while rotating the array, the noise source angle is equivalently moved, thus achieving the goal of placing the noise sources sequentially and evenly at intervals within the field of view.

[0036] Preferably, the measurement unit obtains the processed observation signal in the following manner.

[0037] The row vector of the observed signal x is averaged to obtain the averaged observed signal X;

[0038] Construct the autocorrelation matrix of the mean-normalized observed signal X For autocorrelation matrix After eigenvalue decomposition, the diagonal matrix formed by the n largest eigenvalues ​​is denoted as Λ. S =diag{λ1≥λ2…≥λ n}, corresponding to n feature vectors V S ;

[0039] Calculate the autocorrelation matrix The average of the remaining mn small eigenvalues ​​is used as an estimate of the variance of the additive white noise.

[0040] according to Calculate the observed signal after pre-whitening.

[0041] m represents the autocorrelation matrix. The number of eigenvalues, where n is a preset positive integer.

[0042] Preferably, the measurement unit realizes the measurement from the observed signal in the following manner. Extract the noise source signal matrix Y:

[0043] For the observed signal Performing time delay processing and calculating the time-delayed autocorrelation matrix

[0044] The time-delayed autocorrelation matrix is Performing eigenvalue decomposition;

[0045] According to Estimating the mixing matrix W; The matrix composed of the eigenvectors of The matrix composed of the eigenvectors of

[0046] According to Calculating the noise source signal matrix Y, The transpose matrix of the matrix The transpose matrix of the matrix.

[0047] A computer readable storage medium, the computer readable storage medium stores a computer program, the computer program is executed by the processor to realize the steps of the adaptive blind source separation based far field system response G matrix measurement method.

[0048] A computer software product, comprising: a processor and a storage device;

[0049] The storage device is used for storing one or more programs,

[0050] When the one or more programs are executed by one or more processors, so that the one or more processors implement the adaptive blind source separation based far field system response G matrix measurement method.

[0051] The beneficial effects of the present application compared with the prior art are:

[0052] (1) The adaptive blind source separation based far field system response G matrix measurement method provided by the present application, for the received observation signal, first reduces the correlation between each component in the signal through pre-whitening processing, and then extracts the strong noise source signal, so as to realize the separation of noise source and environment from the signal level, so as to obtain more accurate visibility information, and then by placing the noise point source in the field of view range C uniformly spaced different directions and respectively measuring, the significantly optimized system impulse response G matrix can be obtained.

[0053] (2) The present application provides a high-precision system response G matrix, which can improve the brightness temperature and salinity inversion precision, can improve the on-orbit application effect of China's salinity satellite, can provide a good solution for subsequent sea surface radiation brightness temperature detection, and can lay a solid technical foundation for subsequent model project. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1A flow chart of a method for measuring a far-field system response G matrix based on adaptive blind source separation is provided for an embodiment of the present application.

[0055] Figure 2 A schematic diagram of a method for fixing a point source rotating array is provided.

[0056] Figure 3 A simulated background noise image is provided for an embodiment of the present application.

[0057] Figure 4 A simulated strong noise point source brightness temperature image is provided for an embodiment of the present application.

[0058] Figure 5 A raw brightness temperature scene image after adding a strong noise point source is provided for an embodiment of the present application.

[0059] Figure 6 A strong noise point source brightness temperature image separated is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0060] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. In addition, the technical features involved in the embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0061] In the present application, the terms "first", "second", etc. (if any) in the present application and the accompanying drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence.

[0062] The present application provides a method for measuring a system response G matrix using an external noise source by adaptive blind source separation. Since the signals received by the antenna array contain information of the artificially placed noise source and the ambient background noise, the background information and the noise source information are mixed together to form blind sources at this time. The function of the blind source separation method is to separate the required noise source information and the unwanted background noise interference information. This method can greatly reduce the error accumulation caused by step-by-step measurement compared with the traditional on-orbit satellite G matrix measurement method, and the measured G matrix is more accurate, thereby obtaining a clearer satellite image.

[0063] In order to solve the technical problem that the system response G matrix measured is not accurate enough due to the inaccurate output visibility measurement of the noise source in the existing measurement system, the application provides a method for measuring the far-field system response G matrix based on adaptive blind source separation, and the overall idea is that the noise source is directly separated from the signal level to exclude the influence of other radio frequency interference and improve the accuracy of parameter estimation, and the visibility of different azimuth noise sources is obtained by fixing the point source scanning array, so that the measured system response G matrix is effectively optimized.

[0064] The following is an embodiment.

[0065] Embodiment 1

[0066] A method for measuring the far-field system response G matrix based on adaptive blind source separation, as shown in Figure 1 , comprising:

[0067] (S1) placing a noise source in the field of view range under the open field condition;

[0068] In this embodiment, the noise source is placed in an open outdoor environment to provide a measurement environment with less reflection and meet the far-field condition of the integrated aperture microwave radiometer.

[0069] (S2) receiving the microwave radiation signal emitted by the field of view as the observation signal x of the microwave remote sensing image;

[0070] The observation signal x received in this embodiment is the original visibility obtained by calculation.

[0071] It is easy to understand that the observation signal x is specifically an MxN matrix, where M represents the number of observed antennas, and N represents the number of sampling points. After calculating the autocorrelation matrix of the observation signal, the corresponding visibility can be obtained, and the corresponding brightness temperature image, i.e. the microwave remote sensing image, can be obtained by inverting the visibility.

[0072] (S3) pre-whitening the observation signal x to obtain the observation signal

[0073] Pre-whitening refers to operations such as mean removal and whitening on the original signal, which linearly transforms the observation data vector to have a mean of zero and a variance of 1, and removes the correlation between observations. The correlation of the original observation signal x is unknown, and the step (S2) of this embodiment pre-whitens the observation signal x, which can reduce the correlation between components and make the signal of the noise source more easily separated. In addition, after pre-whitening, the covariance matrix of the signal is an identity matrix, which makes the calculation more simple.

[0074] In a preferred example of the present application, step (S3) specifically comprises:

[0075] (S31) mean value processing is performed on the row vectors of the observation signal x to obtain a mean value processed observation signal X;

[0076] For the kth (1≤k≤M) row vector x(k,:) in the observation signal x, the mean value processing can be represented by the following formula:

[0077] X(k,:) = x(k,:) - mean(x(k,:)) Formula (1-6)

[0078] wherein mean(·) represents mean value; X(k,:) represents the mean value processed x(k,:), and is also the kth row vector in the observation signal X;

[0079] (S32) constructing the autocorrelation matrix of the mean value processed observation signal X Autocorrelation matrix The expression of the autocorrelation matrix is specifically as follows:

[0080]

[0081] wherein the superscript "T" represents transposition;

[0082] (S33) performing eigenvalue decomposition on the autocorrelation matrix After the eigenvalue decomposition, the diagonal matrix composed of the largest n eigenvalues is denoted as Λ S = diag{λ1≥λ2…≥λ n}, and the corresponding n eigenvectors are V S = [v1,v2,…v n ];

[0083] Performing eigenvalue decomposition on the autocorrelation matrix can be represented as follows:

[0084]

[0085] wherein U X represents the matrix composed of the eigenvectors of , Λ X represents the diagonal matrix composed of the eigenvalues of , V N represents the matrix composed of the m-n eigenvectors corresponding to the m-n small eigenvalues of , and Λ N represents the diagonal matrix composed of the m-n small eigenvalues of ;

[0086] (S34) calculating the autocorrelation matrix The average of the remaining mn small eigenvalues ​​is used as an estimate of the variance of the additive white noise.

[0087] The specific expression is as follows:

[0088]

[0089] (S35) according to Calculate the observed signal after pre-whitening.

[0090] in, m represents the autocorrelation matrix. The number of eigenvalues, where n is a preset positive integer and n≤m.

[0091] (S4) From the observed signal Extract the noise source signal matrix Y from it;

[0092] Step (S4) of this embodiment involves observing the signal after pre-whitening processing. The noise signal matrix Y is extracted from it, which provides the basis for subsequent blind source separation.

[0093] In a preferred embodiment of the present invention, step (S4) specifically includes:

[0094] (S41) Observation signal Perform a time delay and calculate the autocorrelation matrix after the delay. for:

[0095]

[0096] Where N represents the number of sampling points, and p represents the delay time. This represents the observed signal at time t after pre-whitening processing. express The observed signal after a delay of p sampling points; through delay processing, the delayed signal can better satisfy the independence assumption and reduce correlation; in practical applications, the delay number can be set according to the specific observation scenario. Optionally, in this embodiment, p = 10 is set.

[0097] (S42) The autocorrelation matrix after the delay is: Eigenvalue decomposition is performed, and the decomposition formula is as follows:

[0098]

[0099] in, Indicate The matrix formed by the eigenvectors of , express a matrix composed of the eigenvectors of the matrix

[0100] (S43) according to estimate the mixing matrix W;

[0101] (S44) according to calculate the noise source signal matrix Y.

[0102] (S5) calculate the actual visibility of the noise source through the obtained signal matrix Y;

[0103] Based on the RFI signal matrix Y obtained in step (S4), the step (S5) of the embodiment realizes the separation of the noise source from the signal level and further calculates the more accurate visibility of the noise source. Step (S5) specifically includes:

[0104] calculate the covariance matrix of the observation signal of the separated noise source to obtain the corresponding signal visibility;

[0105] It is easy to understand that the measurement result of the synthetic aperture radiometer is obtained by analyzing the correlation between the outputs of the two antenna units, which is called the visibility function; the definition of the covariance matrix is also the output of the receiving antenna correlation, so the elements of the covariance matrix correspond to the elements of the visibility function, as shown in the following formula:

[0106]

[0107] Where R represents the covariance matrix, V represents the corresponding visibility, y i represents the output of the i-th antenna, (u ij , v ij ) is the baseline, equal to the distance between the two antennas i, j in the xy plane normalized by the wavelength; therefore, after the covariance matrix is calculated, the corresponding signal visibility can be obtained accordingly;

[0108] (S6) by placing the noise source in the field of view in turn and uniformly spaced to measure respectively, that is, the optimized system response G matrix can be obtained.

[0109] Based on the output visibility of a position obtained in steps (S1) to (S5), the step (S6) of the embodiment repeats the above steps and further obtains the optimized system response G matrix. Step (S6) specifically includes:

[0110] (S61) set a three-dimensional rectangular coordinate system, with the origin set at the center of the array, i.e. the intersection of the three arms of the Y-shaped array; the x-axis extends in the horizontal plane along the direction of one arm of the array; the y-axis is in the horizontal plane and forms a 90° angle with the x-axis, pointing to the side; the z-axis is perpendicular to the ground and points upward.

[0111] (S62) As shown in Figure 2 Fig. 6, the left and right rotating turntable is used to rotate the array around the z axis to change the azimuth angle, and the azimuth angle is sequentially increased from 0° to 360°; the up and down tilting is used to change the elevation angle, and the elevation angle is sequentially increased from 0° (the array is horizontal to the ground) to 90° (the array is vertical to the sky); then the turntable plane is rotated by 180 degrees along the z axis to measure the part from -90 degrees to 0 degrees of the elevation angle. Through the above operation, the observation signal x containing noise at (ξ, η) = (a, b) is obtained, that is, x(ξ = a, η = b);

[0112] (S63) The method of rotating the array by fixing the point source is used to equivalently move the point source angle accurately, and each matrix element g m (ξ c ,η c ) of the system response G matrix is obtained.

[0113] g m (ξ c ,η c ) = v m |T(ξ,η) = δ(ξ-ξ c ,η-η c ) Equation (1-4)

[0114] where v m is the signal response measured by the mth correlator, that is, a pair of antennas in the antenna array; δ(ξ-ξ c ,η-η c ) is a two-dimensional Dirac delta function, indicating that the fixed point source is located at (ξ c ,η c ); (ξ c ,η c ) is the coordinate of the horizontal angle and the vertical angle where the point source (the position of the fixed point source is the position in the field of view after equivalence) is located.

[0115] At this time, the output visibility v m of the mth correlator represents the matrix element g m (ξ c ,η c ). In equation (1-5), the subscript 1 and the like represent the first position, and the Cth position.

[0116] Therefore, by rotating the two-dimensional array to measure the signal response of the noise source, we can gradually construct the G matrix of the entire two-dimensional array:

[0117]

[0118] In order to present the effect of the method, a noise source in one direction is selected and the brightness temperature obtained by the visibility inversion after optimization is compared with the brightness temperature obtained by the original visibility without blind source separation in the embodiment, so as to reflect the accuracy of the method. The following test based on simulation data further explains the beneficial effects that can be achieved by the embodiment. Specifically, the related parameters of the SMOS (Soil Moisture and Ocean Salinity, soil moisture and ocean salinity) satellite radiometer system are simulated, Figure 3 is the simulated background noise; Figure 4 is the simulated noise source brightness temperature image; Figure 5 is the noise source in Figure 3 is the original brightness temperature image obtained after adding the noise source in Figure 4 .

[0119] The separated strong noise source brightness temperature image obtained after processing by the method for measuring the far-field system response G matrix based on adaptive blind source separation provided in the embodiment is as shown in Figure 6 .

[0120] Based on the simulation results shown in Figures 3 to 6 , it can be known that the method for measuring the far-field system response G matrix based on adaptive blind source separation provided in the embodiment can effectively improve the accuracy of the visibility elements in the G matrix and obtain a more optimized system response G matrix.

[0121] The application further provides a far-field system response G matrix measurement system based on adaptive blind source separation, comprising a noise source, a turntable and a measurement unit.

[0122] The noise source is fixed in the far field, and the Y-shaped array is fixed on the turntable, with the array being horizontal to the ground and the output center of the noise source being directly opposite to the center of the Y-shaped array, that is, the intersection of the three arms of the Y-shaped array.

[0123] The noise source angle is equivalent to moving by rotating the turntable, that is, the noise source is placed in the field of view in turn and uniformly spaced, and the measurement is completed by the measurement unit after each time being turned to the position, and the far-field system response G matrix is obtained according to the final measurement result; the measurement of the measurement unit comprises: receiving the microwave radiation signal emitted by the field of view as the observation signal x of the microwave remote sensing image; performing pre-whitening processing on the observation signal x to obtain the processed observation signal extracting the noise source signal matrix Y from the observation signal , so as to separate the noise source and the environment from the signal level; the actual visibility of the noise source is calculated by the obtained signal matrix Y.

[0124] The specific function implementation of the measurement unit in the system can be referred to the introduction in the method, and will not be described here.

[0125] The application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the adaptive blind source separation based far-field system response G matrix measurement method.

[0126] The application further provides a computer software product, which comprises a processor and a storage device.

[0127] The storage device is used for storing one or more programs.

[0128] When the one or more programs are executed by one or more processors, the one or more processors implement the adaptive blind source separation based far-field system response G matrix measurement method.

[0129] The application has been disclosed above with reference to the preferred embodiments, but is not intended to limit the application, and any person skilled in the art can make possible changes and modifications to the technical solutions of the application by using the disclosed methods and technical contents without departing from the spirit and scope of the application. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the application without departing from the technical solutions of the application all belong to the protection scope of the technical solutions of the application.

Claims

1. A method for measuring the far-field system response G matrix based on adaptive blind source separation, characterized in that: The noise source is placed in the field of view in turn and evenly spaced, and the far-field system response G matrix is obtained by measuring respectively; After placing the noise source in the field of view each time, the measurement is performed in the following manner: The microwave radiation signal emitted by the field of view is received as the observation signal x of the microwave remote sensing image; Pre-whiten the observation signal x to obtain a processed observation signal extracting a noise source signal matrix Y from the observation signals from the signal level, the separation of the noise source and the environment is realized; The actual visibility of the noise source is calculated based on the obtained signal matrix Y.

2. The method of claim 1, wherein: The noise source is placed in the field of view in turn and evenly spaced in the following manner: The noise source is fixed in the far field, and the Y-shaped array is fixed on the turntable, with the array being horizontal to the ground and the output center of the noise source being opposite to the center of the Y-shaped array, i.e. the intersection of the three arms of the Y-shaped array; A three-dimensional rectangular coordinate system is set, with the origin being set at the center of the Y-shaped array, the x-axis being in the horizontal plane and extending along one arm of the array, the y-axis being in the horizontal plane and being at a 90° angle with the x-axis and facing the side, and the z-axis being perpendicular to the ground and pointing upward; The turntable is rotated to rotate the array around the z-axis to change the azimuth angle, so that the azimuth angle is sequentially increased from 0° to 360°; the elevation angle is changed by pitching up and down, so that the elevation angle is sequentially increased from 0° to 90°; then the turntable is rotated by 180° as a whole to measure the part from -90° to 0°; the observation signal containing noise at each position in the field of view can be obtained by the above operation, i.e. the method of rotating the array while fixing the noise source, which is equivalent to moving the angle of the noise source.

3. The method of claim 1, wherein: The processed observation signal is obtained by The row vectors of the observation signal x are averaged to obtain the averaged observation signal X; Construct the autocorrelation matrix of the meaned observation signal X Perform eigenvalue decomposition on the autocorrelation matrix After eigenvalue decomposition, the diagonal matrix composed of the largest n eigenvalues is denoted as Λ S = diag{λ1≥λ2…≥λ n n} and the corresponding n eigenvectors are denoted as V S ; Computing autocorrelation matrix The average of the remaining m-n small eigenvalues as an estimate of the additive white noise variance According to Computing the observed signal after pre-whitening m represents the number of eigenvalues of the autocorrelation matrix n is a predetermined positive integer.

4. The method of claim 3, wherein: The extraction of the noise source signal matrix Y from the observation signals is achieved by the following way: to the observed signal delayed and a self-correlation matrix is calculated The autocorrelation matrix after time delay is perform eigenvalue decomposition; According to estimating a mixing matrix W; denotes a matrix of eigenvectors of According to computing a noise source signal matrix Y, representing the transpose of a matrix.

5. An adaptive blind source separation based far field system response G matrix measurement system, characterized by: The noise source, the turntable and the measurement unit are included; The noise source is fixed in the far field, and the Y-shaped array is fixed on the turntable, with the array being horizontal to the ground and the output center of the noise source being opposite to the center of the Y-shaped array, i.e. the intersection of the three arms of the Y-shaped array; By rotating the turntable to move the noise source angle equivalently, i.e. to realize the noise source in the field of view is placed in turn and uniformly spaced, after each rotation to the position, the measurement unit completes the measurement, and the far field system response G matrix is obtained according to the final measurement result; the measurement of the measurement unit includes: receiving the microwave radiation signal emitted by the field of view as the observation signal x of the microwave remote sensing image; the observation signal x is pre-whitened to obtain the processed observation signal The noise source signal matrix Y is extracted from the observation signal , and the separation of the noise source and the environment is realized from the signal level; the actual visibility of the noise source is calculated through the obtained signal matrix Y.

6. The system of claim 5, wherein: A three-dimensional rectangular coordinate system is set, with the origin being set at the center of the Y-shaped array, the x-axis being in the horizontal plane and extending along one arm of the array, the y-axis being in the horizontal plane and being at a 90° angle with the x-axis and facing the side, and the z-axis being perpendicular to the ground and pointing upward; The turntable is rotated to rotate the array around the z-axis to change the azimuth angle, so that the azimuth angle is sequentially increased from 0° to 360°; the elevation angle is changed by pitching up and down, so that the elevation angle is sequentially increased from 0° to 90°; then the turntable is rotated by 180° as a whole to measure the part from -90° to 0°; the observation signal containing noise at each position in the field of view can be obtained by the above operation, i.e. the method of rotating the array while fixing the noise source, which is equivalent to moving the angle of the noise source.

7. The system of claim 1, wherein: The measurement unit obtains the processed observation signal by the following manner The row vectors of the observation signal x are averaged to obtain the averaged observation signal X; Construct the autocorrelation matrix of the meaned observation signal X Perform eigenvalue decomposition on the autocorrelation matrix After eigenvalue decomposition, the diagonal matrix composed of the largest n eigenvalues is denoted as Λ S = diag{λ1≥λ2…≥λ n n}, and the corresponding n eigenvectors are denoted as V S ; Computing autocorrelation matrix The average of the remaining m-n small eigenvalues as an estimate of the additive white noise variance According to Computing the observed signal after pre-whitening m represents the number of eigenvalues of the autocorrelation matrix n is a predetermined positive integer.

8. The system of claim 7, wherein: The measurement unit implements the extraction of the noise source signal matrix Y from the observation signals by the following way: to the observed signal delayed and a self-correlation matrix is calculated The autocorrelation matrix after time delay is perform eigenvalue decomposition; According to estimating a mixing matrix W; denotes a matrix of eigenvectors of According to computing a noise source signal matrix Y, representing the transpose of a matrix.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. The computer program is executed by the processor to realize the steps of the far-field system response G matrix measurement method based on adaptive blind source separation according to any one of claims 1-4.

10. A computer software product, characterized by It includes: a processor and a storage device; a storage device for storing one or more programs, when the one or more programs are executed by one or more processors, the one or more processors implement the far-field system response G matrix measurement method based on adaptive blind source separation according to any one of claims 1-4.

Images