阅读说明：本技术 多通道合成孔径雷达rpca幅相联合目标检测方法与装置 (Multichannel synthetic aperture radar RPCA amplitude-phase combined target detection method and device ) 是由 黄岩 魏晨阳 傅东宁 徐刚 蔡龙珠 翟建锋 于志强 于 2020-07-22 设计创作，主要内容包括：针对强杂波背景下的多通道合成孔径雷达系统,本发明结合鲁棒主成分分析(PRCA)方法和幅相联合检测方法提出了一种多通道合成孔径雷达RPCA幅相联合目标检测方法与装置。该方法首先将各个通道的聚焦图像列向量化后堆叠得到矩阵X,然后利用改进的RPCA方法将矩阵X分解得到低秩矩阵L和稀疏矩阵S,最后引入相位信息进行幅相联合检测。本发明综合考虑了具体的杂波和运动目标信息,有效改善运动目标检测性能；通过两步检测方法,充分结合RPCA和幅相联合检测方法的优点,降低低信杂比条件下的虚警率；并且本发明方法采用的矩阵分解技术,可以极大地降低运算复杂度,提高算法效率。(The invention provides a multichannel synthetic aperture radar RPCA amplitude-phase combined target detection method and a multichannel synthetic aperture radar RPCA amplitude-phase combined target detection device, which are provided by combining a robust principal component analysis (PRCA) method and an amplitude-phase combined detection method aiming at a multichannel synthetic aperture radar system under a strong clutter background. The method comprises the steps of firstly vectorizing and stacking focused image columns of all channels to obtain a matrix X, then decomposing the matrix X by using an improved RPCA method to obtain a low-rank matrix L and a sparse matrix S, and finally introducing phase information to carry out amplitude-phase joint detection. The invention comprehensively considers the specific clutter and the moving target information, and effectively improves the detection performance of the moving target; by adopting a two-step detection method, the advantages of the RPCA and amplitude phase combined detection method are fully combined, and the false alarm rate under the condition of low signal-to-noise ratio is reduced; and the matrix decomposition technology adopted by the method can greatly reduce the operation complexity and improve the algorithm efficiency.)

1. The method for detecting the amplitude-phase combined target of the multichannel synthetic aperture radar RPCA is characterized by comprising the following steps:

(1) vectorizing and stacking the focused images of all channels to obtain a matrix X, wherein each column of the matrix X represents one channel;

(2) decomposing the matrix X by using an improved RPCA method to obtain a low-rank matrix L and a sparse matrix S; where the matrix decomposition problem is expressed as:

s.t.card(S)≤p.

wherein, L ═ uv^H，

(3) after obtaining the sparse matrix S, taking any channel of the sparse matrix S, and introducing phase information to carry out amplitude-phase joint detection, wherein a detector is represented as: zeta_new＝S₁Claim (1-cos θ), wherein,representing the interferometric phase vector between the SAR channels,indicates a real number field, <' > indicates a dot product operation, S₁Indicates taking a column in S, ζ_newRepresenting the final result of target detection.

2. The method for detecting the RPCA (Radar Cross-correlation analysis) amplitude-phase combined target of the multi-channel synthetic aperture radar as claimed in claim 1, wherein the method for solving the matrix decomposition problem in the step (2) is as follows:

changing the matrix decomposition problem into a first sub-problemAnd the second sub-problem

Aiming at the first subproblem, in each iteration process, the fixed sparse matrix S is unchanged, and in the kth iteration, the updating formulas of u and v are respectively:

u_(k+1)＝l

v_(k+1)＝(X-S_(k))^Hl

wherein l is (X-S)_(k))v_(k)Singular value decomposition left eigenvector;

and aiming at the second subproblem, solving by a hard threshold algorithm,whereinThe representation is projected to the omega space, which takes the largest p elements in the matrix.

3. The method for detecting the amplitude-phase joint target of the multi-channel synthetic aperture Radar (RPCA) according to claim 1, wherein the threshold value p is reduced in step (2) to prevent false alarm.

4. Multichannel synthetic aperture radar RPCA amplitude and phase unites target detection device, its characterized in that includes:

the channel stacking module is used for stacking the focused images of the channels after vectorization to obtain a matrix X, and each column of the matrix X represents one channel;

the matrix decomposition module is used for decomposing the matrix X by utilizing an improved RPCA method to obtain a low-rank matrix L and a sparse matrix S; where the matrix decomposition problem is expressed as:

s.t.card(S)≤p.

wherein, L ═ uv^H，card (S) denotes the cardinality of the sparse matrix S, p denotes a cardinality threshold, | · | | computationally |_FRepresents the Frobenius norm,

and the joint detection module is used for taking any channel of the sparse matrix S after obtaining the sparse matrix S and introducing phase information to carry out amplitude-phase joint detection, and the detector is expressed as: zeta_new＝S₁Claim (1-cos θ), wherein,representing the interferometric phase vector between the SAR channels,indicates a real number field, <' > indicates a dot product operation, S₁Indicates taking a column in S, ζ_newRepresenting the final result of target detection.

5. A multi-channel synthetic aperture radar RPCA amplitude-phase joint target detection apparatus comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the computer program, when loaded into the processor, implements a multi-channel synthetic aperture radar RPCA amplitude-phase joint target detection method according to any one of claims 1 to 3.

Technical Field

The invention relates to a method and a device for detecting an RPCA amplitude-phase combined target of a multi-channel synthetic aperture radar, belonging to the technical field of radar imaging.

Background

Synthetic Aperture Radars (SAR) have received wide attention in both the civilian and military fields because of their all-weather operational capabilities throughout the day. Ground Moving Target Indication (GMTI) is one of the most important tasks of the SAR system for remote sensing, such as vehicle detection and terrorist monitoring in traffic monitoring. Typically, weak targets are likely to be submerged in a strongly cluttered background, which is difficult to detect in a simple manner. Based on this, some conventional methods based on multi-channel SAR (MC-SAR) systems, such as a bias phase center antenna (DPCA) method, a space-time adaptive processing (STAP) method, and an along-track interference (ATI) method, are proposed, which utilize the extra degree of freedom (DOF) provided by multi-channels to suppress the strong clutter background. The above algorithms all have their own advantages and disadvantages, the offset phase-centric antenna approach is easy to implement, but its performance will drop dramatically if the antenna baseline does not match the DPCA conditions; the space-time adaptive processing method has excellent performance in suppressing interference, but needs to accurately estimate the covariance matrix of the clutter, and if a target signal is mixed into a training sample of the clutter, the suppression performance is seriously degraded, thereby resulting in a lower output signal-to-noise ratio. The interference method along the flight path is influenced by motion errors, so that signals among channels are unbalanced, interference phase terms among the channels are greatly influenced, and weak moving targets cannot be detected.

In recent years, a Robust Principal Component Analysis (RPCA) type low rank recovery method has become very popular in the field of signal processing because it can separate different constituent components from a contaminated set of correlation databases. The existing literature has proved that the moving target is sparse in image dimension, and the moving speed of the moving target can cause the difference of echo signals among channels; for a broadband synthetic aperture radar system, echoes in a strong clutter region have a certain low-rank structure among multiple channels. Based on this, common RPCA low-rank recovery class methods such as an enhanced lagrange multiplier (ALM) method and a GoDec algorithm can separate a low-rank matrix from a sparse matrix by processing a basic model of RPCA. Unlike DPCA and STAP based approaches, it does not need to consider DPCA conditions and selection of training samples. Although the standard RPCA method has the advantages of suppressing clutter as above, the value of the hyper-parameter in the RPCA basic model is not appropriate, and thus the target detection process may have a high false alarm rate (PFA) under a low signal-to-clutter ratio condition. In addition, compared with the offset phase center antenna, the space-time adaptive processing and the along-track interference method, the robust principal component analysis method is low in efficiency and needs a large amount of time for iteration to converge.

Disclosure of Invention

The purpose of the invention is as follows: the invention aims to provide a method and a device for detecting an RPCA (synthetic aperture radar) amplitude-phase combined target, so as to reduce the possibility of false alarm rate (PFA) and reduce the complexity of operation.

The technical scheme is as follows: in order to achieve the purpose, the invention adopts the following technical scheme:

the method for detecting the amplitude-phase combined target of the multichannel synthetic aperture radar RPCA comprises the following steps:

(1) vectorizing and stacking the focused images of all channels to obtain a matrix X, wherein each column of the matrix X represents one channel;

(2) decomposing the matrix X by using an improved RPCA method to obtain a low-rank matrix L and a sparse matrix S; where the matrix decomposition problem is expressed as:

s.t.card(S)≤p.

wherein, L ═ uv^H，card (S) denotes the cardinality of the sparse matrix S, p denotes a cardinality threshold, | · | | computationally |_FRepresents the Frobenius norm,representing a complex number field, wherein M represents the number of SAR channels, and N represents the number of pixels of a single-channel SAR focused image;

(3) obtaining a sparse matrix S, namely obtaining a target detection preliminary result of all channels, taking any one channel of the target detection preliminary result, introducing phase information to carry out amplitude-phase joint detection, wherein the detector is expressed as: zeta_new＝S₁Claim (1-cos θ), wherein,representing the interferometric phase vector between the SAR channels,

indicates a real number field, <' > indicates a dot product operation, S₁Indicates taking a column in S, ζ_newRepresenting the final result of target detection.

Further, the method for solving the matrix decomposition problem in the step (2) is as follows:

changing the matrix decomposition problem into a first sub-problem

And the second sub-problem

Aiming at the first subproblem, in each iteration process, the fixed sparse matrix S is unchanged, and in the kth iteration, the updating formulas of u and v are respectively:

u_(k+1)＝l

v_(k+1)＝(X-S_(k))^Hl

wherein l is (X-S)_(k))v_(k)Singular value decomposition left eigenvector;

and aiming at the second subproblem, solving by a hard threshold algorithm,wherein

The representation is projected to the omega space, which takes the largest p elements in the matrix.

Further, the threshold p is lowered in step (2) to prevent false alarms.

Based on the same inventive concept, the invention discloses a multichannel synthetic aperture radar RPCA amplitude-phase combined target detection device, which comprises:

the channel stacking module is used for stacking the focused images of the channels after vectorization to obtain a matrix X, and each column of the matrix X represents one channel;

the matrix decomposition module is used for decomposing the matrix X by utilizing an improved RPCA method to obtain a low-rank matrix L and a sparse matrix S;

and the joint detection module is used for taking any channel of the sparse matrix S after obtaining the sparse matrix S and introducing phase information to carry out amplitude-phase joint detection.

Based on the same inventive concept, the device for detecting the amplitude-phase combined target of the multichannel synthetic aperture radar RPCA disclosed by the invention comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein the computer program realizes the method for detecting the amplitude-phase combined target of the multichannel synthetic aperture radar RPCA when being loaded to the processor.

Has the advantages that: the invention combines the amplitude-phase joint detection and the RPCA method together, so that the advantages of the amplitude-phase joint detection and the RPCA method are combined to have more stable and effective ground moving target detection performance. Firstly, combining the thought of an interference method along a track, comparing channel signals by space position elements after vectorization, comprehensively considering the special low-rank characteristic of clutter information and the interference phase of a moving target, and designing a brand-new RPCA optimization model, thereby greatly reducing the calculation complexity of the traditional RPCA method, and simultaneously reducing the threshold to obtain a detection result as much as possible so as to prevent alarm leakage; secondly, sending the pre-detection result into an amplitude-phase joint detector for post-detection, and removing false alarm points to obtain a final detection result. The following benefits can be obtained by the proposed RPCA amplitude-phase joint detection method: (1) specific clutter and moving target information are comprehensively considered, and the moving target detection performance is effectively improved; (2) by adopting a two-step detection method, the advantages of the RPCA and amplitude phase combined detection method are fully combined, and the false alarm rate under the condition of low signal-to-noise ratio is reduced; (3) the method of the invention adopts a matrix decomposition technology, can greatly reduce the operation complexity and improve the algorithm efficiency.

Drawings

Fig. 1 is a schematic diagram of an M-channel SAR system according to an embodiment of the present invention.

Fig. 2 is a schematic diagram of an RPCA target detection method in an embodiment of the present invention.

Fig. 3 is a graph showing the focusing results of three channels of data, wherein (a), (b), and (c) each represent one channel.

Fig. 4 is a graph comparing the indication performance of the ground moving object in the conventional GoDec method, where (a) k is 300, (b) k is 3500, (c) k is 4000, and (d) k is 5000.

Fig. 5 is a comparison graph of the indication performance of the ground moving object in the method of the present invention, where (a) k is 3000, (b) k is 4000, (c) k is 8000, and (d) k is 9000.

Fig. 6 is a comparison graph of false alarm points of the proposed method and the GoDec method under different cardinality values.

Detailed Description

The invention is further illustrated with reference to the following specific embodiments and the accompanying drawings.

The embodiment of the invention considers M channel (M)>2) The synthetic aperture radar system of (2) operates in a front side view stripe mode, and a signal model is shown in fig. 1. Assuming that the velocity of the radar platform is v, the target has a radial velocity v on the ground_rAnd along-track velocity v_aAnd (4) moving. The fast time (i.e. distance time domain) and the slow time (i.e. azimuth time domain) are respectively represented by t_fAnd t_sAnd (4) showing. According to the schematic diagram, point target and miscellaneousThe instantaneous slope distance of wave scattering can be expressed as

Wherein R is_0tAnd R_0cRepresenting the initial nearest slope distance, x, of the moving target and the clutter scattering point_0tAnd x_0cThe initial azimuth positions of the moving target and the clutter scattering point respectively, and d is the interval baseline between two adjacent channels. Since the slant distance is large enough, the second order Taylor approximation expansion is used to obtain

Adopting a Range Doppler (RD) imaging algorithm, the focusing formula of the moving target and the clutter scattering point of the mth channel is

Wherein A is_tAnd A_cRespectively representing the reflection intensity of the moving target and the clutter point, lambda represents the wavelength of the transmitted electromagnetic wave, deltaf_rAnd Δ f_aRespectively, range resolution and azimuth resolution. Notably, the azimuthal focus offset resulting from the separation baseline "(m-1) d" in (5) and (6) can be compensated for by azimuthal compression through different filters. Therefore, the clutter component in all channels is almost the same for one clutter scatter point, and in the SAR image, the ground motionThe targets are sparsely distributed throughout the scene and, due to their different velocities of motion, the phases between the channels are also different. At this time, according to document [9 ]]The method of (3) vectorizes the focused image matrix of each channel into a column, as shown in fig. 2, and then combines the column vectors of each channel together to form a complete observation matrix X. The new observation matrix X can be regarded as a joint matrix composed of three matrices, i.e. a clutter matrix of low rank, a sparse moving object matrix and a noise matrix. The three main components of the robust principal component analysis problem which meets a standard can be used for separating the clutter and extracting the moving target by using the low-rank recovery algorithms.

And in fact, the new matrix is equivalent to X in (1) in terms of a signal model of robust principal component analysis. As long as the optimization problem in (2) is solved, the moving object can be detected. The RPCA amplitude-phase joint moving target detection method of the present invention will be described in detail below.

The robust principal component analysis method is a method widely used for data analysis and dimension reduction, and aims to recover an intrinsic low-rank information and polluted sparse matrix from damaged measurement data. Specifically, the RPCA basic mathematical model is a matrix decomposition that takes into account the following form:

X＝L+S+N (7)

wherein X, L, S and N represent an original matrix, a low-rank matrix, a sparse matrix and a noise matrix, respectively. If the information is redundant in nature, i.e., the low rank matrix is of very low rank and the non-zero elements in the sparse matrix are sparse, the original matrix can be divided into three independent matrices by solving the following optimization:

min||L||_*+μ||S||₁

s.t.||X-L-S||_F＜ (8)

in the formula, | · the luminance | |_*Representing a nuclear norm equal to the sum of singular values, | · | | luminance₁L representing a matrix₁Norm equal to the sum of absolute values of elements of the matrix, | · | | luminance_FRepresenting the Frobenius norm, which is equal to the sum of the squares of the elements of the matrix, a constant related to the observed noise level, and μ a hyper-parameter of the equilibrium targetAnd (4) counting. In view of having excellent performance of low rank background rejection, the RPCA method has been widely used to extract a moving target from a stationary clutter background. It is worth noting that although the conditions supporting separation of low rank and sparse matrices are unknown, the tight low rank and sparse nature between the components can help achieve separation of the signals. Therefore, the RPCA method can also be effectively applied to moving object detection.

In fact, the detection performance of the traditional robust principal component analysis method is reduced in the strong clutter background, which is mainly caused by the fact that the traditional robust principal component analysis method does not perform corresponding optimization and processing on a moving target detection scene, and lacks the robust performance aiming at specific problems. Therefore, specific characteristics of low-rank components and clutter components of the synthetic aperture radar ground moving target are analyzed in detail aiming at specific problems of synthetic aperture radar ground moving target indication, and then an RPCA amplitude-phase joint detection method is provided in a targeted mode and mainly comprises two steps of pre-detection and post-detection.

A. Improved RPCA pre-detection method

Firstly, the signal model in formula (2) can be solved by using the augmented lagrange multiplier method, and the RPCA problem in formula (1) can be equivalently expressed by using the following GoDec model, that is, by minimizing the decomposition error, the decomposition problem of approximate 'low rank matrix + sparse matrix' is realized:

where rank (l) denotes the rank of the low rank matrix and card(s) denotes the cardinality of the sparse matrix. The GoDec optimization model well converts the hyper-parameters of the original RPCA optimization problem into the rank of a low-rank matrix and the cardinality of a sparse matrix, thereby facilitating specific discussion of ground motion target indication of the SAR system. In fact, if the matrix arrangement method of fig. 2 is used, each column in the final received signal large matrix represents a channel. At this time, the ground clutter scenes between multiple channels are almost the same, so the rank of the clutter matrix should be 1. So for this assumption we can turn the low rank matrix L into the product of two vectors, namely:

L＝uv^H(10)

wherein the content of the first and second substances,representing the complex field, M the number of SAR channels, and N the number of pixels of the single-channel SAR focused image (since we vectorize the single-channel focused image). The original problem becomes:

s.t.card(S)≤p. (11)

we will recover a larger low rank matrix problem, becoming the problem of recovering the product of two vectors. To solve the problem in (11) above, the values of several variables can be iterated alternately by using an alternate direction multiplier method, and then the problem can be changed to the following sub-problem:

s.t.card(S)≤p. (13)

for the problem (12), in each iteration process, the fixed sparse matrix S is not changed, which is a typical least square problem, and then in the kth iteration, the update formulas of u and v are respectively:

then uv is used after the k-th iteration^HThe product of (d) can be expressed as:

wherein the content of the first and second substances,

represents to u_(k+1)The column space of (a) is projected. And in fact u_(k+1)Is also equivalent to (X-S)_(k))v_(k)The column space of (a). If pair (X-S)_(k))v_(k)Performing singular value decomposition to obtain:

(X-S_(k))v_(k)＝lλr, (17)

where l is the singular value decomposition left eigenvector (first letter of left), λ represents the singular value, and r is the singular value decomposition right eigenvalue constant (first letter of right), the above special singular value decomposition representation is due to (X-S)_(k))v_(k)Is a vector. At this time, the process of the present invention,

can be equivalently expressed as

Wherein the content of the first and second substances,

representing the projection into the column space of l. The product of u and v vectors is concerned, and the specific value of each vector is not concerned, so the projection of the vector u can be replaced by the projection of the vector l, and the projection of the vector u can be called

u_(k+1)＝l, (19)

v_(k+1)＝(X-S_(k))^Hl. (20)

The values of the u and v vectors can be finally expressed by the above equations (19) and (20). The subproblem (13) is solved by the hard threshold algorithm in GoDec, i.e. it is solved by the algorithm

Wherein

The representation is projected to the omega space, which takes the largest p elements in the matrix. Meanwhile, aiming at the specific problem of SAR ground moving target indication, the moving target is represented by the sparse matrix, and the moving target is reflected in the M channels and is not mutually independent. Therefore, after taking the first p large elements, the moving object corresponds to some rows of non-zero elements in the S matrix, and therefore the values of these non-zero elements need to be large (i.e. much higher than the noise value), which satisfies the characteristics of the ground moving object. Therefore, on the basis of the original hard threshold algorithm, the sparse matrix is further filtered. The pre-detection method fully considers the characteristics of clutter and moving targets in the ground moving target indication, so that the moving target detection method has quicker and more effective moving target detection performance. However, after the hard threshold algorithm is adopted, the hyper-parameter p needs to be accurately adjusted, and the detection performance is easily affected by the condition of false alarm or missing alarm in the adjustment process. Therefore, in order to further improve the detection performance and reduce the false-alarm or false-alarm probability, the invention provides a post-detection method based on amplitude-phase combination.

B. Amplitude-phase joint post-detection

After the pre-detection, the hard threshold algorithm is adopted, so that the pre-detected phase cannot express the specific characteristics of the moving target. Therefore, the Advantages of The Interference (ATI) along the track are fully considered in the post detection, and the interference phase information acquired by the ATI is introduced. Taking the pre-detection result of any channel (for example, taking S column 1) and recording as S₁And performing amplitude-phase joint detection by combining amplitude information acquired by pre-detection, wherein the detector can be expressed as:

ζ_new＝S₁⊙(1-cosθ) (22)

wherein the content of the first and second substances,representing the interference phase vector between the channels,

indicating a real number field, an indication of a dot product operation. In order to successfully detect real objects, the predetermined base p of the sparse matrix should be set larger during pre-detection to prevent missing some real moving objects and causing false alarms. Although a large base number may bring wrong targets and cause false alarms, a sparse matrix obtained by pre-detection is used as a new amplitude term in the formula and is combined with an interference phase term to form a new amplitude-phase joint detector, so that the false alarms can be further reduced, and better detection performance can be obtained. Here, ζ_newThe single-channel moving target detection vector is obtained, and the detection result vector is converted into the size of the original single-channel signal matrix, namely the target detection result which is finally presented. The performance of the proposed method will be analyzed in detail below using measured SAR data.

In the invention, an experiment is carried out on the X-waveband SAR original data, and the actually measured data is collected by a three-channel airborne SAR system in a front side view stripe mode so as to prove the effectiveness of the provided along-track interference RPCA method. The three channels are arranged at fixed intervals along the flight path, and specific parameters of the SAR system are shown in Table 1.

Table 1 three channel SAR system parameters.

First, a three-channel focused image is acquired by a range-doppler algorithm, as shown in fig. 3. In the focused image, some cooperative objects (vehicles) are moving in both directions along the road, wherein the region of interest is enlarged within a rectangular frame. In the SAR focused image, the moving target deviates from the road and is difficult to be directly distinguished from the strong clutter scattering point.

The focused images of the three channels are vectorized and stacked into a new matrix according to the description of the signal model. The correlation of stationary clutter between channels is such that the clutter matrix is low rank, whereas the moving object matrix is sparse since the moving objects are sparsely dispersed throughout the focused image. However, the direct adoption of the conventional RPCA-like algorithm does not take into account the specific problem of ground moving object indication. In addition, in the process of separating clutter from a moving target, the value of the hyper-parameter brings about a serious reduction of detection performance, for example, a hyper-parameter p for constraining a sparse matrix base number in a GoDec algorithm, if the value of the hyper-parameter p is small, false alarm may be caused, and if the value of the hyper-parameter p is large, false alarm may be caused. Fig. 5 shows moving object indication performance of the conventional algorithm at different p-values (sparse matrix cardinality). In fig. 4 (a), since the p value is set too small, a moving object (a dotted ellipse) is missed after detection, causing a false alarm; the good detection performance resulting from the good p-values is shown in fig. 4 (b), three moving objects can be detected and no false alarm occurs. Whereas in fig. 4 (c) and (d), when the cardinality reaches 4000, false alarms begin to appear, and as the cardinality increases, false targets become more and more. Therefore, the conventional RPCA algorithm is seriously affected by the hyper-parameter, and the selection of the hyper-parameter needs to be finely adjusted for specific data, which will increase the computational complexity of the system and also reduce the detection performance.

The method of the invention fully considers the problem of SAR system moving target indication, can fully alleviate the influence of the hyper-parameter, and the detection result under different hyper-parameter k values is shown in figure 5. In fig. 5 (a), similar to the conventional GoDec algorithm, the method of the present invention generates a false alarm at a k value of 3000, which is consistent with the analysis of the third section. Therefore, in the preliminary detection process, the base number needs to be set larger in advance to avoid missing the detection target. Fig. 5 (b) and (c) show the robust detection performance obtained by the method of the present invention over a wide range of values of the over-parameter; in other words, the method of the present invention can broaden the range of values of the basis values and limit the occurrence of false alarm targets. For the example, the method of the invention can successfully detect all moving targets in the p value range of [3500,8000], only few or no false alarm targets appear, and the traditional GoDec algorithm can realize stable detection performance only in a small range with the cardinality set to be about 3500, which undoubtedly greatly enhances the engineering practice value of the algorithm. In addition, the method is more efficient than the traditional GoDec algorithm, because the characteristics of a low-rank matrix represented by clutter and a sparse matrix represented by a moving target are fully considered, and a matrix decomposition method is introduced to further accelerate the algorithm, under the same calculation condition, the calculation time of the method is 0.38s, and is one order of magnitude faster than that of the traditional GoDec method.

Fig. 6 further shows the number of false alarm points under different p-value conditions in the method of the present invention and the conventional GoDec method, specifically, the detection threshold is set to 0.03 after the signal after detection is normalized. As can be seen from the figure, the method of the present invention can keep the false alarm point at a very low level; but as cardinality increases, the traditional GoDec algorithm can detect even hundreds of false targets. Therefore, the method has larger cardinality selection tolerance and more robust moving object detection performance compared with the traditional GoDec algorithm.

Based on the same inventive concept, the multi-channel synthetic aperture radar RPCA amplitude-phase combined target detection device disclosed by the embodiment of the invention comprises: the channel stacking module is used for stacking the focused images of the channels after vectorization to obtain a matrix X, and each column of the matrix X represents one channel; the matrix decomposition module is used for decomposing the matrix X by utilizing an improved RPCA method to obtain a low-rank matrix L and a sparse matrix S; and the joint detection module is used for taking any channel of the sparse matrix S after obtaining the sparse matrix S and introducing phase information to carry out amplitude-phase joint detection. For details, reference is made to the above method embodiments, which are not described herein again.

Based on the same inventive concept, the apparatus for detecting the amplitude-phase combined target of the multi-channel synthetic aperture radar RPCA disclosed in the embodiments of the present invention includes a memory, a processor, and a computer program stored in the memory and operable on the processor, where the computer program is loaded into the processor to implement the method for detecting the amplitude-phase combined target of the multi-channel synthetic aperture radar RPCA.