Real-time small moving target detection method based on spatiotemporal parallel coupling robust principal component analysis

Abstract

The present application proposes a method, apparatus, equipment and medium for real-time small motion target detection using spatiotemporal parallel coupled robust principal component analysis, including: constructing a spatiotemporal data matrix based on the image data to be detected; constructing the coarse-grained sparse decomposition problem based on the rough low-rank matrix to be solved and the extreme sparse matrix to be solved, and constructing the fine-grained sparse decomposition problem based on the extreme low-rank matrix to be solved and the rough sparse matrix to be solved; solving the coarse-grained sparse decomposition problem and the fine-grained sparse decomposition problem respectively based on the Lagrange multiplier algorithm to obtain a rough low-rank matrix and an extreme low-rank matrix; using an edge continuity clustering detection algorithm to process the rough low-rank matrix and the extreme low-rank matrix to obtain small motion target detection results. The present invention can directly abandon the background modeling process, can directly detect targets from the sparse matrix, and improve the efficiency of target detection.

Description

Technical Field

[0001] The present application relates to the technical field of motion target detection, and in particular to a method, apparatus, device and medium for real-time weak motion target detection using spatiotemporal parallel coupled robust principal component analysis. Background Art

[0002] Detection of weak moving targets is a key and difficult problem in the field of target detection. On the one hand, weak targets themselves contain little information and are not obvious compared with noise. Whether using radar, sonar, infrared or camera detection equipment, there are serious detection problems; on the other hand, some weak targets, such as drones carrying warheads, can pose a serious security threat.

[0003] Xixi Chen et al. proposed a radar weak target detection method based on robust principal component analysis. This method achieves better detection performance for weak targets in complex backgrounds than conventional methods. However, due to the inherent characteristics of radar, this method has significant difficulties detecting slow-moving weak targets such as drones, unmanned vehicles, and airships. Prashant W. Patil et al. proposed an end-to-end moving target detection network. Compared with comparison methods, this method performs best in detection metrics such as prediction rate, recall rate, and F1 index. However, due to the inherent characteristics of deep convolutional neural networks, this method has difficulty effectively detecting weak moving targets.

[0004] Currently, the detection of common small moving targets still relies primarily on visual detection equipment such as infrared and cameras. Most moving target detection methods, including deep learning, optical flow, and background modeling, are poor at detecting small moving targets. Summary of the Invention

[0005] In view of this, the embodiments of the present application propose a method, device, equipment and medium for real-time weak motion target detection using spatiotemporal parallel coupling robust principal component analysis, aiming to better adapt to changes in the camera field of view and background area, while effectively improving detection accuracy, reducing false alarm rate and missed detection rate, and improving the operating efficiency of the method.

[0006] To achieve the above-mentioned objectives, an embodiment of the present application provides a real-time weak motion target detection method using robust principal component analysis with time-space parallel coupling, including: obtaining image data to be detected, and constructing a time-space data matrix based on the image data to be detected; constructing a coarse-grained sparse decomposition problem and a fine-grained sparse decomposition problem based on the time-space data matrix, wherein the coarse-grained sparse decomposition problem is constructed based on a rough low-rank matrix to be solved and an extreme sparse matrix to be solved, and the fine-grained sparse decomposition problem is constructed based on an extreme low-rank matrix to be solved and a rough sparse matrix to be solved; solving the coarse-grained sparse decomposition problem and the fine-grained sparse decomposition problem based on a Lagrange multiplier algorithm to obtain a rough low-rank matrix and an extreme low-rank matrix; processing the rough low-rank matrix and the extreme low-rank matrix using an edge continuity clustering detection algorithm to obtain a target image matrix, and determining a weak motion target detection result based on the target image matrix.

[0007] Optionally, the coarse-grained sparse decomposition problem is constructed based on the rough low-rank matrix to be solved and the extreme sparse matrix to be solved, including: constructing a first objective function based on the sum of the kernel function of the rough low-rank matrix to be solved and the L1 norm of the extreme sparse matrix to be solved with a first weighting coefficient, wherein the rough low-rank matrix to be solved and the extreme sparse matrix are obtained based on the decomposition of the spatiotemporal data matrix.

[0008] Optionally, the fine-grained sparse decomposition problem is constructed based on the extreme low-rank matrix to be solved and the rough sparse matrix to be solved, including: constructing a second objective function based on the sum of the kernel function of the extreme low-rank matrix to be solved and the L1 norm of the rough sparse matrix to be solved with a second weighting coefficient, wherein the extreme low-rank matrix to be solved and the rough sparse matrix to be solved are obtained based on the decomposition of the spatiotemporal data matrix.

[0009] Optionally, the coarse-grained sparse decomposition problem and the fine-grained sparse decomposition problem are solved based on the Lagrange multiplier algorithm respectively to obtain a rough low-rank matrix and an extreme low-rank matrix, including: solving the minimum value of the first objective function based on the inexact augmented Lagrange multiplier algorithm to obtain a rough low-rank matrix; solving the minimum value of the second objective function based on the exact Lagrange multiplier algorithm to obtain the extreme low-rank matrix.

[0010] Optionally, the use of the edge continuity clustering detection algorithm to process the rough low-rank matrix and the extreme low-rank matrix to obtain the target image matrix includes: obtaining a first coordinate set in the rough low-rank matrix whose matrix values ​​are greater than a first preset value; determining a second coordinate set in the low-rank matrix to be solved that is the same as the first coordinate set based on the first coordinate point set; determining a third coordinate set belonging to a connected area with the second coordinate set according to the edge continuity clustering detection algorithm; determining the non-zero values ​​and zero values ​​in the extreme low-rank matrix to be solved based on the third coordinate set; obtaining the target image matrix based on the non-zero values ​​and zero values ​​in the extreme low-rank matrix to be solved, wherein the first preset value and the second preset value are both greater than or equal to zero and less than or equal to 1, and the first preset value is much greater than the second preset value.

[0011] Optionally, determining the third coordinate set that belongs to a connected area with the second coordinate set based on the edge continuity clustering detection algorithm includes: detecting coordinate points around the second coordinate set that are connected to each coordinate point in the second coordinate set based on the Canny operator or the deep continuous clustering algorithm; and obtaining the third coordinate set based on the coordinate points connected to each coordinate point in the second coordinate set.

[0012] Optionally, the spatiotemporal parallel coupled robust principal component analysis real-time weak motion target detection method also includes: obtaining the background image matrix based on the difference between the spatiotemporal data matrix and the target image matrix, and obtaining the background image based on the background image matrix.

[0013] To achieve the above-mentioned purpose, an embodiment of the present application also provides a real-time weak motion target detection device using a time-space parallel coupled robust principal component analysis, comprising: an acquisition module for acquiring image data to be detected and constructing a time-space data matrix based on the image data to be detected; a processing module for constructing a coarse-grained sparse decomposition problem and a fine-grained sparse decomposition problem based on the time-space data matrix, wherein the coarse-grained sparse decomposition problem is constructed based on a rough low-rank matrix to be solved and an extreme sparse matrix to be solved, and the fine-grained sparse decomposition problem is constructed based on an extreme low-rank matrix to be solved and a rough sparse matrix to be solved, and the coarse-grained sparse decomposition problem and the fine-grained sparse decomposition problem are solved based on a Lagrange multiplier algorithm to obtain a rough low-rank matrix and an extreme low-rank matrix; an output module for processing the rough low-rank matrix and the extreme low-rank matrix using an edge continuity clustering detection algorithm to obtain a target image matrix, and determining the weak motion target detection result based on the target image matrix.

[0014] To achieve the above-mentioned purpose, an embodiment of the present application also provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the real-time weak motion target detection method of spatiotemporal parallel coupling robust principal component analysis provided by any of the aforementioned embodiments.

[0015] To achieve the above-mentioned purpose, an embodiment of the present application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement the above-mentioned time-space parallel coupling robust principal component analysis real-time weak motion target detection method.

[0016] The embodiments of the present application propose a method, device, equipment and medium for real-time weak motion target detection using spatiotemporal parallel coupled robust principal component analysis, including: obtaining image data to be detected, and constructing a spatiotemporal data matrix based on the image data to be detected; constructing a coarse-grained sparse decomposition problem and a fine-grained sparse decomposition problem based on the spatiotemporal data matrix, wherein the coarse-grained sparse decomposition problem is constructed based on a rough low-rank matrix to be solved and an extreme sparse matrix to be solved, and the fine-grained sparse decomposition problem is constructed based on an extreme low-rank matrix to be solved and a rough sparse matrix to be solved; solving the coarse-grained sparse decomposition problem and the fine-grained sparse decomposition problem based on the Lagrange multiplier algorithm to obtain a rough low-rank matrix and an extreme low-rank matrix; processing the rough low-rank matrix and the extreme low-rank matrix using an edge continuity clustering detection algorithm to obtain a target image matrix, and determining a weak motion target detection result based on the target image matrix. The present invention can directly abandon the background modeling process, can directly detect the target from the sparse matrix, and improve the efficiency of target detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is the process of the real-time weak moving target detection method of spatiotemporal parallel coupled robust principal component analysis provided in one embodiment of the present application Figure 1 ;

[0018] Figure 2 The process of the real-time weak moving target detection method using spatiotemporal parallel coupled robust principal component analysis is provided in one embodiment of the present application. Figure 2 ;

[0019] Figure 3 In one embodiment of the present application, four matrices generated during the operation of a real-time weak moving target detection method using spatiotemporal parallel coupled robust principal component analysis and their corresponding visualization effect display diagrams are provided;

[0020] FIG4 is a comparison diagram of the effects of a real-time weak moving target detection method using a spatiotemporal parallel coupled robust principal component analysis method provided in one embodiment of the present application;

[0021] Figure 5 This is a structural block diagram of a real-time weak moving target detection device using robust principal component analysis with time-space parallel coupling, provided in one embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, each embodiment of the present application will be described in detail below with reference to the accompanying drawings. However, it will be understood by those skilled in the art that in each embodiment of the present application, many technical details are proposed to enable the reader to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in the present application can be implemented. The division of the following embodiments is for convenience of description and should not constitute any limitation on the specific implementation of the present application. The various embodiments can be combined and referenced with each other under the premise of no contradiction.

[0023] Traditional robust principal component analysis has the following shortcomings. First, after the ideal matrix decomposition, the target information is only contained in the sparse matrix S. However, this information is difficult to be directly and effectively extracted and utilized, which causes serious information waste, while increasing the complexity of the detection method and greatly reducing the efficiency of the detection process. In addition to containing potential target information, the sparse matrix also contains a large amount of noise, error, and irrelevant information components that are difficult to filter out. These components are mixed with the effective information of the potential target, making it difficult to effectively separate the target component. The present invention effectively solves the above problems through parallel coupling ideas and clustering filtering methods, so that the detection method can directly use the sparse matrix to extract effective information, directly realize accurate detection of the target, greatly improve the operating efficiency of the algorithm, and at the same time improve the accuracy and signal-to-noise ratio of detection.

[0024] Second, in view of the problem described in question one, the traditional robust principal component analysis method cannot directly and effectively use sparse matrices for target detection. Therefore, the common operating process of the traditional robust principal component analysis method when performing target detection is to first use the low-rank matrix L to establish a robust detection background, and then subtract the background information from the input data of the current frame. The obtained content is considered as an effective foreground (i.e., a moving target) after a certain filtering algorithm. This causes the actual detection results to be mixed with a large amount of noise. When the target to be detected is relatively weak, the above problem will be more serious. The present invention directly abandons the background modeling process and detects the target directly from the sparse matrix. It can effectively circumvent the above problems, better adapt to the changes in the camera field of view and background area, and improve the operation efficiency of the algorithm.

[0025] One embodiment of the present application proposes a real-time weak moving target detection method using robust principal component analysis with time-space parallel coupling, which is applied to an electronic device, where the electronic device can be a terminal or a server. This embodiment and the following embodiments are all described using a server as an example. The following describes in detail the implementation details of the real-time weak moving target detection method using robust principal component analysis with time-space parallel coupling proposed in this embodiment. The following content is only for the convenience of understanding the implementation details provided and is not required for the implementation of this solution.

[0026] refer to Figure 1 and Figure 2 The specific process of the real-time weak moving target detection method using spatiotemporal parallel coupled robust principal component analysis proposed in this embodiment may include:

[0027] S10, acquiring image data to be detected, and constructing a spatiotemporal data matrix based on the image data to be detected;

[0028] Among them, the spatiotemporal data matrix is ​​the input data of the moving target detection method, which is usually a data matrix composed of grayscale values ​​of a video or a time-sequential image sequence.

[0029] S20. Construct a coarse-grained sparse decomposition problem and a fine-grained sparse decomposition problem based on the spatiotemporal data matrix, respectively. The coarse-grained sparse decomposition problem is constructed based on the rough low-rank matrix to be solved and the extreme sparse matrix to be solved, and the fine-grained sparse decomposition problem is constructed based on the extreme low-rank matrix to be solved and the rough sparse matrix to be solved. The coarse-grained sparse decomposition problem and the fine-grained sparse decomposition problem are respectively solved based on the Lagrange multiplier algorithm to obtain a rough low-rank matrix and an extreme low-rank matrix.

[0030] In an embodiment of the present application, step S20 may include the following execution process:

[0031] S201. Construct a first objective function based on the sum of the kernel function of the rough low-rank matrix to be solved and the L1 norm of the extreme sparse matrix to be solved with a first weighting coefficient, wherein the rough low-rank matrix to be solved and the extreme sparse matrix are obtained based on the decomposition of the spatiotemporal data matrix.

[0032] S202. Construct a second objective function based on the sum of the kernel function of the extreme low-rank matrix to be solved and the L1 norm of the rough sparse matrix to be solved with a second weighting coefficient, wherein the extreme low-rank matrix to be solved and the rough sparse matrix to be solved are obtained based on the decomposition of the spatiotemporal data matrix.

[0033] S203, solving the minimum value of the first objective function based on an inexact augmented Lagrange multiplier algorithm to obtain a rough low-rank matrix;

[0034] S204 . Solve the minimum value of the second objective function based on the exact Lagrange multiplier algorithm to obtain the extreme low-rank matrix.

[0035] Among them, the server simultaneously constructs two independent sparse decomposition problems to be optimized based on the spatiotemporal data matrix.

[0036] In principle, the first optimization problem is rougher than the second optimization problem during the optimization process, which is called a coarse-grained sparse decomposition problem; the second optimization problem is more detailed than the first optimization problem during the optimization process, which is called a fine-grained sparse decomposition problem.

[0037] refer to Figure 3 After constructing two independent sparse decomposition problems to be optimized, the server can simultaneously use different optimization methods to solve the two problems to be optimized, and obtain the rough low-rank matrix (Lr), the extreme sparse matrix (Se), the extreme low-rank matrix (Le), and the rough sparse matrix (Sr).

[0038] In the specific implementation process, two independent optimization problems are constructed at the same time. These two problems can be respectively the coarse-grained sparse decomposition problem and the fine-grained sparse decomposition problem, and the iterative optimization method is used to solve the above problems. The two independent optimization problems are expressed as follows:

[0039]

[0040] subject to X≈Lr+Se≈Le+Sr

[0041] Due to limitations in approximate substitution and approximate computation during iterative optimization, the equal sign in the formula is replaced by an approximately equal sign to indicate that the relationship between the two is only approximately equal, not a strict equality constraint. Here, λ∈[0,1], η∈[0,1], and λ>>η, X≈Lr+Se≈Le+Sr.

[0042] In another embodiment, when the question is relatively simple, the first question can also be The second problem is solved by using the inexact augmented Lagrange multiplier method. The exact Lagrange multiplier method is used to solve the problem.

[0043] S30. Processing the rough low-rank matrix and the extreme low-rank matrix using an edge continuity clustering detection algorithm to obtain a target image matrix, and determining a weak moving target detection result based on the target image matrix.

[0044] In an embodiment of the present application, step S30 may include the following execution steps:

[0045] Obtaining a first coordinate set in the rough low-rank matrix whose matrix value is greater than a first preset value;

[0046] S301: Determine, based on the first coordinate point set, a second coordinate set in a low-rank matrix to be solved that is the same as the first coordinate set;

[0047] S302: Determine, according to an edge continuity cluster detection algorithm, a third coordinate set that belongs to a connected area with the second coordinate set;

[0048] In an embodiment of the present application, step S302 may include the following execution steps:

[0049] S3021. Detecting coordinate points around the second coordinate set that are connected to each coordinate point in the second coordinate set based on a Canny operator or a deep continuous clustering algorithm;

[0050] S3022: Obtain the third coordinate set based on coordinate points connected to each coordinate point in the second coordinate set.

[0051] S303, determining non-zero values ​​and zero values ​​in the extreme low-rank matrix to be solved based on the third coordinate set;

[0052] S304, obtaining the target image matrix based on the non-zero values ​​and zero values ​​in the extreme low-rank matrix to be solved;

[0053] The first preset value and the second preset value are both greater than or equal to zero and less than or equal to 1, and the first preset value is much greater than the second preset value.

[0054] In the specific implementation process, the moving target region is centered on the target point set of the extreme sparse matrix Se (corresponding to the first coordinate set), and the region (corresponding to the third coordinate set) is detected using edge continuity clustering in the corresponding coordinate region of the Sr matrix (corresponding to the second coordinate set). From this, the present invention can directly obtain a more accurate moving target foreground region without background modeling. Its mathematical expression is as follows:

[0055]

[0056] In the formula, x and y represent the horizontal and vertical coordinates of any matrix value in the extreme sparse matrix Se, respectively. (x+i,y+j) represents the connected area with (x,y) as the core, and the specific calculation result of the edge detection operator shall prevail, that is, for the rough sparse matrix Sr, (x+i,y+j) should satisfy Se(x,y)>δ and (x+i,y+j) is within the continuous area clustering interval calculated by the edge detection operator with (x,y) as the center. δ∈[0,1], ξ∈[0,1], the present invention refers to them as positioning sensitivity coefficient and detection sensitivity coefficient respectively. Among them, the larger δ is, the more targets are detected, and it may be accompanied by a decrease in detection accuracy and a decrease in signal-to-noise ratio; the smaller δ is, the easier it is to guarantee the detection accuracy, stability and signal-to-noise ratio, but the number of detected targets may be reduced. The larger the ξ, the easier it is to preserve the details of the moving foreground, but more noise may be introduced; the smaller the ξ, the better the detection accuracy and signal-to-noise ratio, but some details of the moving foreground may be lost.

[0057] In the specific implementation process, the point that satisfies Se(x,y)>δ, δ∈[0,1] is used as the coordinate center, and the edge continuity operator (such as the Canny operator, or simple continuity clustering) is used on the relevant area of ​​the Sr matrix to calculate the connected area that meets the conditions. The coordinates of the relevant points that meet the requirements are marked as (x+i,y+j). For the point Sr(x+i,y+j)>ξ, Fa(x+i,y+j)=Sr(x+i,y+j). Other points in the Fa matrix (corresponding to the target image matrix) that do not meet the conditions remain unchanged at 0. The result is recorded as the foreground area of ​​the moving target detection, that is, the detected weak moving target.

[0058] In another embodiment, the real-time weak moving target detection method using spatiotemporal parallel coupled robust principal component analysis further includes:

[0059] S40 , obtaining the background image matrix based on the difference between the spatiotemporal data matrix and the target image matrix, and obtaining a background image based on the background image matrix.

[0060] In specific implementations, the server subtracts the Fa matrix from the X matrix (corresponding to the spatiotemporal data matrix) to obtain a more accurate, but less robust, background image matrix Ba, which contains noise. Using the background image matrix Ba as a background model, moving target detection can also be achieved using background modeling. Compared to traditional RPCA-based moving target detection methods or other background modeling methods, this background model contains a wider range of noise and interference information. Therefore, using this background model for moving target detection (when the background model is subtracted from the current frame) can effectively improve detection accuracy and signal-to-noise ratio.

[0061] Figure 4 is a comparison diagram of the actual operation results of the present invention in Example 1, wherein Figure 4(a) is a frame of the original video; Figure 4(b) is a diagram showing the operation results of the sparse decomposition moving target detection method using the inexact Lagrange multiplier method (IALM); Figure 4(c) is a diagram showing the operation results of the sparse decomposition moving target detection method using the alternating direction iteration method (ADM); Figure 4(d) is a diagram showing the operation results of the sparse decomposition moving target detection method using the motion-assisted matrix recovery (RMAMR); and Figure 4(e) is a diagram showing the operation results of the moving target detection method using the present invention. It can be seen from the figure that the effect of detecting the target image using the method of the present application is far superior to the target image detection effect of other comparative methods.

[0062] The beneficial effects of the method of the present invention mainly include:

[0063] (1) The present invention solves the problem that RPCA-type motion target detection methods must rely on low-rank matrix components and background modeling methods to achieve target detection. It creatively extracts target information directly through sparse components for detection, and effectively suppresses the interference of a large amount of noise, interference, calculation errors and irrelevant components contained in the sparse matrix on the detection results. While simplifying the method operation process and improving the operation efficiency, it effectively improves the accuracy and stability of the detection results.

[0064] (2) Compared with the traditional RPCA-type motion target detection method, the spatiotemporal parallel coupled robust principal component analysis technology of the present invention can effectively reduce the signal-to-noise ratio of the detection results without reducing the algorithm operation efficiency; compared with other methods such as deep learning method, optical flow method, background modeling method, etc., it has obvious advantages in detecting weak targets and can greatly improve the detection rate of weak targets and the accuracy of detection results.

[0065] (3) This method can also obtain more accurate background information (including noise). Even if the background modeling method is selected for target detection, the background information provided by this method is used as the background model, and the target detection results obtained are more accurate and have a lower signal-to-noise ratio.

[0066] It should be pointed out that the step division of the various methods above is only for clear description. When implemented, they can be combined into one step or some steps can be split and decomposed into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this application; adding insignificant modifications to the algorithm or process or introducing insignificant designs without changing the core design of the algorithm and process are all within the scope of protection of this application.

[0067] On the basis of the above embodiment, the present invention further proposes a real-time weak moving target detection device using robust principal component analysis and time-space parallel coupling. The real-time weak moving target detection device using robust principal component analysis and time-space parallel coupling proposed in this embodiment is specifically described below. The following content is only for the convenience of understanding the implementation details provided and is not necessary for the implementation of this embodiment. Figure 5 It is a structural block diagram of the real-time weak motion target detection device of the spatiotemporal parallel coupled robust principal component analysis proposed in this embodiment. The target detection device 100 may include an acquisition module 1001, a processing module 1002 and an output module 1003. Among them, the acquisition module 1001 is used to acquire the image data to be detected and construct a spatiotemporal data matrix based on the image data to be detected; the processing module 1002 is used to construct a coarse-grained sparse decomposition problem and a fine-grained sparse decomposition problem based on the spatiotemporal data matrix, wherein, based on the rough low-rank matrix to be solved and the The coarse-grained sparse decomposition problem is constructed based on the extreme sparse matrix, the fine-grained sparse decomposition problem is constructed based on the extreme low-rank matrix to be solved and the rough sparse matrix to be solved, and the coarse-grained sparse decomposition problem and the fine-grained sparse decomposition problem are solved respectively based on the Lagrange multiplier algorithm to obtain a rough low-rank matrix and an extreme low-rank matrix; the output module 1003 is used to use the edge continuity clustering detection algorithm to process the rough low-rank matrix and the extreme low-rank matrix to obtain a target image matrix, and determine the weak motion target detection result based on the target image matrix.

[0068] It is not difficult to find that this embodiment is an apparatus embodiment corresponding to the above-mentioned method embodiment, and this embodiment can be implemented in conjunction with the above-mentioned method embodiment. The relevant technical details and technical effects mentioned in the above-mentioned embodiments are still valid in this embodiment, and to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above-mentioned embodiments.

[0069] It is worth mentioning that all modules involved in this embodiment are logical modules. In actual applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, to highlight the innovation of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed by this application. However, this does not mean that other units do not exist in this embodiment.

[0070] Another embodiment of the present application proposes an electronic device, comprising at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the real-time weak motion target detection method using spatiotemporal parallel coupling robust principal component analysis provided in any one of the aforementioned embodiments.

[0071] The memory and processor are connected using a bus, which can include any number of interconnected buses and bridges. The bus connects various circuits of one or more processors and memories. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits. These are all well known in the art and therefore will not be described further in this article. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. Data processed by the processor is transmitted on a wireless medium via an antenna. Furthermore, the antenna also receives data and transmits it to the processor.

[0072] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory can be used to store data used by the processor when performing operations.

[0073] Another embodiment of the present application relates to a computer-readable storage medium storing a computer program, which implements the above method embodiment when executed by a processor.

[0074] That is, those skilled in the art will understand that all or part of the steps in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a program, which is stored in a storage medium and includes a number of instructions for causing a device (which may be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: a USB flash drive, a mobile hard drive, a ROM (Read-Only Memory), a RAM (Random Access Memory), a magnetic disk, or an optical disk, etc., various media that can store program code.

[0075] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application, and that in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.

Claims

1. A real-time small moving target detection method based on robust principal component analysis with time-space parallel coupling, characterized in that: include: Acquire image data to be detected, and construct a spatiotemporal data matrix based on the image data to be detected; Based on the spatiotemporal data matrix, a coarse-grained sparse decomposition problem and a fine-grained sparse decomposition problem are respectively constructed, wherein the coarse-grained sparse decomposition problem is constructed based on the rough low-rank matrix to be solved and the extreme sparse matrix to be solved, and the fine-grained sparse decomposition problem is constructed based on the extreme low-rank matrix to be solved and the rough sparse matrix to be solved, and the coarse-grained sparse decomposition problem and the fine-grained sparse decomposition problem are respectively solved based on the Lagrange multiplier algorithm to obtain a rough low-rank matrix and an extreme low-rank matrix; the coarse-grained sparse decomposition problem is decomposed into Processing the rough low-rank matrix and the extreme low-rank matrix using an edge continuity clustering detection algorithm to obtain a target image matrix, and determining a weak moving target detection result based on the target image matrix; The coarse-grained sparse decomposition problem constructed based on the rough low-rank matrix to be solved and the extreme sparse matrix to be solved includes: Constructing a first objective function based on the sum of a kernel function of the rough low-rank matrix to be solved and an L1 norm of the extreme sparse matrix to be solved with a first weighting coefficient, wherein the rough low-rank matrix to be solved and the extreme sparse matrix are obtained based on the decomposition of the spatiotemporal data matrix; The fine-grained sparse decomposition problem constructed based on the extreme low-rank matrix to be solved and the rough sparse matrix to be solved includes: A second objective function is constructed based on the sum of the kernel function of the extreme low rank matrix to be solved and the L1 norm of the rough sparse matrix to be solved with a second weighting coefficient, wherein the extreme low rank matrix to be solved and the rough sparse matrix to be solved are obtained based on the decomposition of the spatiotemporal data matrix.

2. The method for real-time small moving target detection using robust principal component analysis with spatiotemporal parallel coupling according to claim 1, wherein the coarse-grained sparse decomposition problem and the fine-grained sparse decomposition problem are solved respectively based on the Lagrange multiplier algorithm to obtain a rough low-rank matrix and an extreme low-rank matrix, comprising: Solving the minimum value of the first objective function based on an inexact augmented Lagrange multiplier algorithm to obtain a rough low-rank matrix; The minimum value of the second objective function is solved based on the exact Lagrange multiplier algorithm to obtain the extreme low-rank matrix.

3. The real-time small moving target detection method based on spatiotemporal parallel coupling robust principal component analysis according to claim 1 is characterized in that: The method of using an edge continuity clustering detection algorithm to process the rough low-rank matrix and the extreme low-rank matrix to obtain a target image matrix includes: Obtaining a first coordinate set in the rough low-rank matrix whose matrix value is greater than a first preset value; Determine, based on the first coordinate set, a second coordinate set in the low-rank matrix to be solved that is the same as the first coordinate set; determining, according to an edge continuity cluster detection algorithm, a third coordinate set that belongs to a connected region with the second coordinate set; Determine non-zero values ​​and zero values ​​in the extreme low-rank matrix to be solved based on the third coordinate set; Obtaining the target image matrix based on the non-zero values ​​and zero values ​​in the extreme low-rank matrix to be solved; The first preset value and the second preset value are both greater than or equal to zero and less than or equal to 1, and the first preset value is much greater than the second preset value.

4. The method for real-time small moving target detection using robust principal component analysis with time-space parallel coupling according to claim 3 is characterized in that: The determining, according to the edge continuity cluster detection algorithm, a third coordinate set belonging to a connected area with the second coordinate set includes: Detecting coordinate points around the second coordinate set that are connected to each coordinate point in the second coordinate set based on a Canny operator or a deep continuous clustering algorithm; The third coordinate set is obtained based on coordinate points connected to each coordinate point in the second coordinate set.

5. The method for real-time small moving target detection using robust principal component analysis with time-space parallel coupling according to claim 1 is characterized in that: After determining the small moving target detection result based on the target image matrix, the time-space parallel coupled robust principal component analysis real-time small moving target detection method further includes: A background image matrix is ​​obtained based on the difference between the spatiotemporal data matrix and the target image matrix, and a background image is obtained based on the background image matrix.

6. A device for real-time small moving target detection using robust principal component analysis and time-space parallel coupling, which executes the real-time small moving target detection method using robust principal component analysis and time-space parallel coupling according to any one of claims 1 to 5, characterized in that: include: An acquisition module, configured to acquire image data to be detected and construct a spatiotemporal data matrix based on the image data to be detected; A processing module, used to construct a coarse-grained sparse decomposition problem and a fine-grained sparse decomposition problem based on the spatiotemporal data matrix, wherein the coarse-grained sparse decomposition problem is constructed based on a rough low-rank matrix to be solved and an extreme sparse matrix to be solved, and the fine-grained sparse decomposition problem is constructed based on the extreme low-rank matrix to be solved and the rough sparse matrix to be solved, and the coarse-grained sparse decomposition problem and the fine-grained sparse decomposition problem are respectively solved based on a Lagrange multiplier algorithm to obtain a rough low-rank matrix and an extreme low-rank matrix; An output module is used to process the rough low-rank matrix and the extreme low-rank matrix using an edge continuity clustering detection algorithm to obtain a target image matrix, and determine a weak moving target detection result based on the target image matrix.

7. An electronic device, characterized in that: include: at least one processor; and, a memory communicatively coupled to the at least one processor; In which, the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the real-time weak motion target detection method of spatiotemporal parallel coupling robust principal component analysis as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the real-time weak moving target detection method using spatiotemporal parallel coupling robust principal component analysis as claimed in any one of claims 1 to 5 can be implemented.

Images