A moving target detection method and system based on non-convex norm

Abstract

The present invention discloses a moving target detection method and system based on non-convex norm, including: obtaining video information, converting the video information into a matrix; inputting the matrix into a pre-built moving target detection model based on improved robust principal component analysis, solving the moving target detection model using an alternating direction multiplier method, and extracting the moving target according to the solution result; the moving target detection model uses a non-convex log norm and a truncated fractional norm as a sparsity function and a rank function in the robust principal component analysis. Advantages: The non-convex log norm and the truncated fractional norm in the moving target detection model based on the improved robust principal component analysis can more effectively approximate the rank function and the sparsity function in the traditional robust principal component analysis model, and the extracted moving target effect is better.

Description

Technical Field

[0001] The invention relates to a moving target detection method and system based on non-convex norm, belonging to the technical field of multimedia information processing. Background Art

[0002] Moving target detection has been a hot topic in recent years. Traditional methods include double Gaussian models, Gaussian mixture models, etc., but these methods are for single pixels and have limited functions. The principal component analysis method extracts moving targets by modeling the background, but this method does not model the moving target, resulting in poor results. In the past decade, the robust principal component analysis method has been proposed and widely used in moving target detection. The basic idea of ​​this method is to model the moving target and the background separately, such as the principal component tracking method based on robust principal component analysis, the non-convex low-rank sparse decomposition method, and the robust principal component analysis method assisted by motion information. Summary of the invention

[0003] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide a moving target detection method and system based on non-convex norm.

[0004] In order to solve the above technical problems, the present invention provides a moving target detection method based on non-convex norm, comprising:

[0005] Obtain video information and convert the video information into a measurement matrix, where each column of the measurement matrix is ​​a frame of the video and the number of rows of the measurement matrix is ​​the number of frames corresponding to the video;

[0006] The observation matrix is ​​input into a pre-built motion target detection model based on improved robust principal component analysis, the motion target detection model based on improved robust principal component analysis is solved by using the alternating direction multiplier method, and the motion target is extracted according to the solution result; the motion target detection model based on improved robust principal component analysis uses a non-convex log norm and a truncated fractional norm as a sparsity function and a rank function in the robust principal component analysis.

[0007] Furthermore, the moving target detection model is:

[0008]

[0009] stD=L+S

[0010] Among them, D is the observation matrix converted from video information, L∈R m×n is the low-rank background matrix in the observation matrix D, is the truncated fractional norm, σ t(L) is the t-th singular value of L, m, n are the rows and columns of the observation matrix, r is the cutoff parameter, β is the fractional norm parameter, S∈R m×n is the sparse foreground matrix in the observation matrix D, α is the parameter of the log norm, λ is the compromise parameter, the subscript ij represents the element of the i-th row and j-th column of the corresponding matrix, R m×n It is an m×n dimensional real space.

[0011] Furthermore, the method of solving the moving target detection model by using the alternating direction multiplier method includes:

[0012] 1) According to the moving target detection model, the corresponding augmented Lagrangian function is determined, which is expressed as:

[0013]

[0014] Among them, μ>0 is the penalty factor, Y is the Lagrange multiplier, <,> is the matrix inner product, ║·║ F is the Frobenius norm;

[0015] 2) Update the low-rank background matrix L, sparse foreground matrix S, Lagrange multiplier Y, and penalty factor μ, including:

[0016] 21) Get the initial values ​​of L, Y and μ or the updated values ​​corresponding to the previous cycle, fix L, Y and μ, update S, and get:

[0017]

[0018] S k+1 The minimization calculation converted to a scalar is shown below:

[0019]

[0020] The superscript or subscript k represents the previous update state, the superscript or subscript k+1 represents the current update state, k=1,2,…K, K represents the number of previous cycles that meet the termination condition,

[0021] make For its variable S ij Take the derivative and set it equal to 0, and we get S ij for::

[0022]

[0023] in,

[0024] 22) Get the initial values ​​of Y and μ or the updated values ​​corresponding to the previous cycle and S in step 21) k+1 , fixed S k+1 , Y and μ, update the matrix variable L and get:

[0025]

[0026] in Represents the singular value threshold operator.

[0027] Get the initial values ​​of Y and μ or the updated values ​​corresponding to the previous cycle, S in step 21) k+1 and L in step 22) k+1 , fixed L k+1 , S k+1 , update Y and its parameters μ, and get:

[0028] Y k+1 =Y k -μ k (L k+1 +S k+1 -D)

[0029] μ k+1 =min(ρμ k ,μ max )

[0030] where ρ>1 is the step size parameter, μ max is the maximum value of μ;

[0031] 3) Repeat step 2) until the termination condition is reached, and output the low-rank matrix L corresponding to the background and the sparse matrix S corresponding to the foreground in the video information.

[0032] Further, extracting the moving target according to the solution result includes:

[0033] The moving target is determined according to the low-rank matrix L corresponding to the background and the sparse matrix S corresponding to the foreground in the video information.

[0034] A moving target detection system based on non-convex norm, comprising:

[0035] An acquisition module is used to acquire video information and convert the video information into an observation matrix, wherein each column of the observation matrix is ​​a frame of the video, and the number of rows of the observation matrix is ​​the number of frames corresponding to the video;

[0036] A processing module is used to input the observation matrix into a pre-built motion target detection model based on improved robust principal component analysis, use the alternating direction multiplier method to solve the motion target detection model based on improved robust principal component analysis, and extract the motion target according to the solution result; the motion target detection model based on improved robust principal component analysis uses a non-convex log norm and a truncated fractional norm as the sparsity function and rank function in the robust principal component analysis.

[0037] Furthermore, the moving target detection model is:

[0038]

[0039] stD=L+S

[0040] Among them, D is the observation matrix converted from video information, L∈R m×n is the low-rank background matrix in the observation matrix D, is the truncated fractional norm, σ t (L) is the t-th singular value of L, m, n are the rows and columns of the observation matrix, r is the cutoff parameter, β is the fractional norm parameter, S∈R m×n is the sparse foreground matrix in the observation matrix D, α is the parameter of the log norm, λ is the compromise parameter, the subscript ij represents the element of the i-th row and j-th column of the corresponding matrix, R m×n It is an m×n dimensional real space.

[0041] Furthermore, the processing module includes:

[0042] The first determination module is used to determine the corresponding augmented Lagrangian function according to the moving target detection model, which is expressed as:

[0043]

[0044] Among them, μ>0 is the penalty factor, Y is the Lagrange multiplier, <,> is the matrix inner product, ║·║ F is the Frobenius norm;

[0045] The solution module is used to update the low-rank background matrix L, the sparse foreground matrix S, the Lagrange multiplier Y, and the penalty factor μ, including:

[0046] 21) Get the initial values ​​of L, Y and μ or the updated values ​​corresponding to the previous cycle, fix L, Y and μ, update S, and get:

[0047]

[0048] S k+1 The minimization calculation converted to a scalar is shown below:

[0049]

[0050] The superscript or subscript k represents the previous update state, the superscript or subscript k+1 represents the current update state, k=1,2,…K, K represents the number of previous cycles that meet the termination condition,

[0051] make For its variable S ij Take the derivative and set it equal to 0, and we get S ij for::

[0052]

[0053] in, The updated value has a higher priority than the initial value;

[0054] 22) Get the initial values ​​of Y and μ or the updated values ​​corresponding to the previous cycle and S in step 21) k+1 , fixed S k+1 , Y and μ, update the matrix variable L and get:

[0055]

[0056] in Indicates the singular value threshold operator, the updated value has a higher priority than the initial value;

[0057] Get the initial values ​​of Y and μ or the updated values ​​corresponding to the previous cycle, S in step 21) k+1 and L in step 22) k+1 , fixed L k+1 , S k+1 , update Y and its parameters μ, and get:

[0058] Y k+1 =Y k -μ k (L k+1 +S k+1 -D)

[0059] μ k+1 =min(ρμ k ,μ max )

[0060] where ρ>1 is the step size parameter, μ max is the maximum value of μ, and the priority of using the updated value is greater than the initial value.

[0061] The loop module is used to loop the processing process of the solution module until the termination condition is reached, and output the low-rank matrix L corresponding to the background and the sparse matrix S corresponding to the foreground in the video information.

[0062] Furthermore, the processing module includes:

[0063] The second determination module is used to determine the moving target according to the low-rank matrix L corresponding to the background and the sparse matrix S corresponding to the foreground in the video information.

[0064] The beneficial effects achieved by the present invention are:

[0065] The non-convex log norm and truncated fractional norm are used to replace the sparsity function and rank function in the traditional robust principal component analysis, and the alternating direction multiplier method is used to solve the proposed method and extract the moving targets. Compared with the existing robust principal component analysis algorithm, the non-convex log norm and truncated fractional norm can more effectively approximate the rank function and sparsity function in the traditional robust principal component analysis model, and the extracted moving targets are better. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a structural diagram of a moving target detection method based on non-convex norm;

[0067] Figure 2 It is a visual contrast effect diagram of a moving target detection method based on non-convex norm;

[0068] Figure 3 This is a comparison diagram of the F-measure values ​​of the moving target detection method based on the non-convex norm;

[0069] Figure 4 This is a comparison diagram of the running time values ​​of a moving target detection method based on a non-convex norm. DETAILED DESCRIPTION

[0070] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0071] like Figure 1As shown in the figure, it is a structural block diagram of a motion target detection method based on a non-convex norm. It can be seen from the figure that the present invention proposes a new motion target detection method based on a non-convex norm, constructs a motion target detection method model based on a non-convex norm, and the model respectively adopts a non-convex log norm and a truncated fractional norm to replace the sparsity function and the rank function in the traditional robust principal component analysis; the alternating direction multiplier method is used to solve the motion target detection method model based on the non-convex norm; the proposed motion target detection method based on the non-convex norm is applied to motion target detection, and the performance of the motion target detection method based on the non-convex norm proposed by the present invention is verified from the perspectives of vision, F-measure value, processing time, etc.

[0072] First, the moving target detection model based on the improved robust principal component analysis is:

[0073]

[0074] stD=L+S

[0075] Among them, D is the observation matrix converted from video information, L∈R m×n is the low-rank background matrix in the observation matrix D, is the truncated fractional norm, σ t (L) is the t-th singular value of L, m, n are the rows and columns of the observation matrix, r is the cutoff parameter, β is the fractional norm parameter, S∈R m×n is the sparse foreground matrix in the observation matrix D, α is the parameter of the log norm, λ is the compromise parameter, the subscript ij represents the element of the i-th row and j-th column of the corresponding matrix, R m×n It is an m×n dimensional real space.

[0076] Secondly, the alternating direction multiplier method is used to solve the moving target detection model based on the improved robust principal component analysis. The augmented Lagrangian function corresponding to the moving target detection model based on the improved robust principal component analysis is:

[0077]

[0078] Where μ>0 is the penalty factor, Y is the Lagrange multiplier, and <,> is the matrix inner product ║·║ F is the Frobenius norm.

[0079] The solution process of the above problem is as follows:

[0080] (1) Get the initial values ​​of L, Y and μ or the updated values ​​corresponding to the previous cycle, fix L, Y and μ, and update S, so that:

[0081]

[0082] The above problem can be converted into a scalar minimization problem as shown below:

[0083]

[0084] The superscript or subscript k represents the previous update state, the superscript or subscript k+1 represents the current update state, k=1,2,…K, K represents the number of previous cycles that meet the termination condition,

[0085] set up: For its variable S ij Take the derivative and set it equal to 0, and we get S ij for::

[0086]

[0087] in Updated values ​​take precedence over initial values.

[0088] (2) Obtain the initial values ​​of Y and μ or the updated values ​​corresponding to the previous cycle and S in step 21) k+1 , fixed S k+1 , Y and μ, update the matrix variable L, and then get:

[0089]

[0090] in Represents the singular value threshold operator. The updated value has a higher priority than the initial value.

[0091] (3) Obtain the initial values ​​of Y and μ or the updated values ​​corresponding to the previous cycle, S in step 21) k+1 and L in step 22) k+1 , fixed L k+1 , S k+1 , update Y and its parameters μ, we can get:

[0092] Y k+1 =Y k -μ k (L k+1 +S k+1 -D)

[0093] μ k+1 =min(ρμ k ,μ max )

[0094] where ρ>1 is the step size parameter, μ max is the maximum value of μ, and the priority of using the updated value is greater than the initial value.

[0095] The process of steps (1)-(3) is repeated until the termination condition is reached, and the low-rank matrix L corresponding to the background and the sparse matrix S corresponding to the foreground in the video information are output, and the moving target is determined based on the low-rank matrix L corresponding to the background and the sparse matrix S corresponding to the foreground in the video information.

[0096] Correspondingly, the present invention also provides a moving target detection system based on non-convex norm, comprising:

[0097] An acquisition module is used to acquire video information and convert the video information into an observation matrix, wherein each column of the observation matrix is ​​a frame of the video, and the number of rows of the observation matrix is ​​the number of frames corresponding to the video;

[0098] A processing module is used to input the matrix into a pre-built motion target detection model, use the alternating direction multiplication method to solve the motion target detection model based on the improved robust principal component analysis, and extract the motion target according to the solution result; the motion target detection model based on the improved robust principal component analysis uses a non-convex log norm and a truncated fractional norm as the sparsity function and rank function in the robust principal component analysis.

[0099] Furthermore, the moving target detection model is:

[0100]

[0101] stD=L+S

[0102] Among them, D is the observation matrix converted from video information, L∈R m×n is the low-rank background matrix in the observation matrix D, is the truncated fractional norm, σ t (L) is the t-th singular value of L, m, n are the rows and columns of the observation matrix, r is the cutoff parameter, β is the fractional norm parameter, S∈R m×n is the sparse foreground matrix in the observation matrix D, α is the parameter of the log norm, λ is the compromise parameter, the subscript ij represents the element of the i-th row and j-th column of the corresponding matrix, R m×n It is an m×n dimensional real space.

[0103] Furthermore, the processing module includes:

[0104] The first determination module is used to determine the corresponding augmented Lagrangian function according to the moving target detection model, which is expressed as:

[0105]

[0106] Among them, μ>0 is the penalty factor, Y is the Lagrange multiplier, <,> is the matrix inner product, ║·║F is the Frobenius norm;

[0107] The solution module is used to update the low-rank background matrix L, the sparse foreground matrix S, the Lagrange multiplier Y, and the penalty factor μ, including:

[0108] 21) Get the initial values ​​of L, Y and μ or the updated values ​​corresponding to the previous cycle, fix L, Y and μ, update S, and get:

[0109]

[0110] S k+1 The minimization calculation converted to a scalar is shown below:

[0111]

[0112] The superscript or subscript k represents the previous update state, the superscript or subscript k+1 represents the current update state, k=1,2,…K, K represents the number of previous cycles that meet the termination condition,

[0113] make For its variable S ij Take the derivative and set it equal to 0, and we get S ij for:

[0114]

[0115] in, Updated values ​​take precedence over initial values.

[0116] 22) Get the initial values ​​of Y and μ or the updated values ​​corresponding to the previous cycle and S in step 21) k+1 , fixed S k+1 , Y and μ, update the matrix variable L and get:

[0117]

[0118] in Represents the singular value threshold operator. The updated value has a higher priority than the initial value.

[0119] Get the initial values ​​of Y and μ or the updated values ​​corresponding to the previous cycle, S in step 21) k+1 and L in step 22) k+1 , fixed L k+1 , S k+1 , update Y and its parameters μ, and get:

[0120] Y k+1 =Yk -μ k (L k+1 +S k+1 -D)

[0121] μ k+1 =min(ρμ k ,μ max )

[0122] where ρ>1 is the step size parameter, μ max is the maximum value of μ, and the priority of using the updated value is greater than the initial value.

[0123] The loop module is used to loop the processing process of the solution module until the termination condition is reached, and output the low-rank matrix L corresponding to the background and the sparse matrix S corresponding to the foreground in the video information.

[0124] Furthermore, the processing module includes:

[0125] The second determination module is used to determine the moving target according to the low-rank matrix L corresponding to the background and the sparse matrix S corresponding to the foreground in the video information.

[0126] Furthermore, if Figure 2 As shown, the figure is a visual comparison effect diagram of a motion target detection method based on a non-convex norm. The third column (NLTFN) in the article is a visual effect diagram of a motion target detection method based on a non-convex norm. It can be seen from the figure that the motion target detection method based on a non-convex norm proposed in the present invention has a better effect in extracting motion targets.

[0127] Furthermore, if Figure 3 As shown in the figure, it is a comparison effect diagram of the F-measure value of a motion target detection method based on non-convex norm. The F-measure value is between 0 and 1. The larger the value, the better the effect. From this figure, it can be seen that the motion target detection method based on non-convex norm proposed in this paper has the best average effect on different videos.

[0128] Finally, if Figure 4 As shown, it is a comparison effect diagram of the running time value of a moving target detection method based on non-convex norm. From this figure, it can be seen that the time for the moving target detection method based on non-convex norm proposed by the present invention to extract moving targets from different videos ranks second among the comparison algorithms. Under the premise of ensuring the extraction effect, the running time is relatively good.

[0129] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0130] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0131] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0132] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0133] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A moving target detection method based on non-convex norm, characterized in that: include: Obtain video information and convert the video information into a measurement matrix, where each column of the measurement matrix is ​​a frame of the video and the number of rows of the measurement matrix is ​​the number of frames corresponding to the video; The observation matrix is ​​input into a pre-built moving target detection model based on improved robust principal component analysis, the moving target detection model based on improved robust principal component analysis is solved by using an alternating direction multiplier method, and the moving target is extracted according to the solution result; the moving target detection model based on improved robust principal component analysis uses a non-convex log norm and a truncated fractional norm as a sparsity function and a rank function in the robust principal component analysis; The moving target detection model is: stD=L+S Among them, D is the observation matrix converted from video information, L∈R m×n is the low-rank background matrix in the observation matrix D, is the truncated fractional norm, σ t (L) is the t-th singular value of L, m, n are the rows and columns of the observation matrix, r is the cutoff parameter, β is the fractional norm parameter, S∈R m×n is the sparse foreground matrix in the observation matrix D, α is the parameter of the log norm, λ is the compromise parameter, the subscript ij represents the element of the i-th row and j-th column of the corresponding matrix, R m×n It is an m×n dimensional real space.

2. The moving target detection method based on non-convex norm according to claim 1, characterized in that: The method of using the alternating direction multiplier method to solve the moving target detection model includes: 1) According to the moving target detection model, the corresponding augmented Lagrangian function is determined, which is expressed as: Among them, μ>0 is the penalty factor, Y is the Lagrange multiplier, <,> is the matrix inner product, ║·║ F is the Frobenius norm; 2) Update the low-rank background matrix L, sparse foreground matrix S, Lagrange multiplier Y, and penalty factor μ, including: 21) Get the initial values ​​of L, Y and μ or the updated values ​​corresponding to the previous cycle, fix L, Y and μ, update S, and get: S k+1 The minimization calculation converted to a scalar is shown below: The superscript or subscript k represents the previous update state, the superscript or subscript k+1 represents the current update state, k=1,2,…K, K represents the number of previous cycles that meet the termination condition, make For its variable S ij Take the derivative and set it equal to 0, and we get S ij for: in, The updated value has a higher priority than the initial value; 22) Get the initial values ​​of Y and μ or the updated values ​​corresponding to the previous cycle and S in step 21) k+1 , fixed S k+1 , Y and μ, update the low-rank background matrix L and obtain: in Indicates the singular value threshold operator, the updated value has a higher priority than the initial value; 23) Get the initial values ​​of Y and μ or the updated values ​​corresponding to the previous cycle, S in step 21) k+1 and L in step 22) k+1 , fixed L k+1 , S k+1 , update Y and its parameters μ, and get: Y k+1 =Y k -μ k (L k+1 +S k+1 -D) m k+1 =min(ρμ k ,m max ) where ρ>1 is the step size parameter, μ max is the maximum value of μ, and the priority of the updated value is greater than the initial value; 3) Repeat step 2) until the termination condition is reached, and output the low-rank background matrix L corresponding to the background and the sparse foreground matrix S corresponding to the foreground in the video information.

3. The moving target detection method based on non-convex norm according to claim 2 is characterized in that: The step of extracting the moving target according to the solution result comprises: The moving target is determined according to the low-rank background matrix L corresponding to the background and the sparse foreground matrix S corresponding to the foreground in the video information.

4. A moving target detection system based on non-convex norm, characterized in that: include: An acquisition module is used to acquire video information and convert the video information into an observation matrix, wherein each column of the observation matrix is ​​a frame of the video, and the number of rows of the observation matrix is ​​the number of frames corresponding to the video; A processing module is used to input the observation matrix into a pre-built moving target detection model based on improved robust principal component analysis, solve the moving target detection model based on improved robust principal component analysis by using an alternating direction multiplier method, and extract the moving target according to the solution result; the moving target detection model based on improved robust principal component analysis uses a non-convex log norm and a truncated fractional norm as a sparsity function and a rank function in the robust principal component analysis; The moving target detection model is: stD=L+S Among them, D is the observation matrix converted from video information, L∈R m×n is the low-rank background matrix in the observation matrix D, is the truncated fractional norm, σ t (L) is the t-th singular value of L, m, n are the rows and columns of the observation matrix, r is the cutoff parameter, β is the fractional norm parameter, S∈R m×n is the sparse foreground matrix in the observation matrix D, α is the parameter of the log norm, λ is the compromise parameter, the subscript ij represents the element of the i-th row and j-th column of the corresponding matrix, R m×n It is an m×n dimensional real space.

5. The moving target detection system based on non-convex norm according to claim 4, characterized in that: The processing module comprises: The first determination module is used to determine the corresponding augmented Lagrangian function according to the moving target detection model, which is expressed as: Among them, μ>0 is the penalty factor, Y is the Lagrange multiplier, <,> is the matrix inner product, ║·║ F is the Frobenius norm; The solution module is used to update the low-rank background matrix L, the sparse foreground matrix S, the Lagrange multiplier Y, and the penalty factor μ, including: 21) Get the initial values ​​of L, Y and μ or the updated values ​​corresponding to the previous cycle, fix L, Y and μ, update S, and get: S k+1 The minimization calculation converted to a scalar is shown below: The superscript or subscript k represents the previous update state, the superscript or subscript k+1 represents the current update state, k=1,2,…K, K represents the number of previous cycles that meet the termination condition, make For its variable S ij Take the derivative and set it equal to 0, and we get S ij for: in, The updated value has a higher priority than the initial value; 22) Get the initial values ​​of Y and μ or the updated values ​​corresponding to the previous cycle and S in step 21) k+1 , fixed S k+1 , Y and μ, update the matrix variable L and get: in Indicates the singular value threshold operator, the updated value has a higher priority than the initial value; 23) Get the initial values ​​of Y and μ or the updated values ​​corresponding to the previous cycle, S in step 21) k+1 and L in step 22) k+1 , fixed L k+1 , S k+1 , update Y and its parameters μ, and get: Y k+1 =Y k -μ k (L k+1 +S k+1 -D) m k+1 =min(ρμ k ,m max ) where ρ>1 is the step size parameter, μ max is the maximum value of μ, and the priority of the updated value is greater than the initial value; The loop module is used to loop the processing process of the solution module until the termination condition is reached, and output the low-rank background matrix L corresponding to the background and the sparse foreground matrix S corresponding to the foreground in the video information.

6. The moving target detection system based on non-convex norm according to claim 5, characterized in that: The processing module comprises: The second determination module is used to determine the moving target according to the low-rank background matrix L corresponding to the background and the sparse foreground matrix S corresponding to the foreground in the video information.

Images