Target detection method and device based on unitary transformation, electronic equipment and storage medium

Abstract

The present application relates to the technical field of target detection, and particularly relates to a target detection method and device based on unitary transformation and a storage medium. The method comprises the following steps: constructing a space-time tensor of multiple infrared images, wherein the space-time tensor comprises a background tensor and a target tensor; constructing a target function of the space-time tensor, wherein the target function is obtained by constraining the background tensor by using a tensor kernel norm in a unitary transformation domain and constraining the target tensor by using a joint space-time total variation and L1 norm; solving the target function to obtain the target tensor; reconstructing the obtained target tensor into multiple single-frame target images, and outputting a detection result of each target image. The method can accurately detect targets in a complex background and has strong detection capability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of target detection, and particularly relates to a target detection method and device based on unitary transformation and a storage medium. BACKGROUND

[0002] Infrared detection technology has the characteristics of strong anti-interference ability and all-weather working, so the infrared search and track (IRST) system is widely used in military and civilian fields. As a basic function of the IRST system, infrared target detection plays an important role in space reconnaissance and disaster rescue.

[0003] In the related art, when the imaging environment of a target is relatively complex, the existing target detection method cannot accurately detect the target, and the detection capability is poor.

[0004] Therefore, there is an urgent need for a target detection method based on unitary transformation to solve the above technical problems. SUMMARY

[0005] In view of the problems of low target detection accuracy and poor detection capability of the existing target detection method, the embodiments of the present application provide a target detection method and device based on unitary transformation and a storage medium, which can accurately detect targets in a complex background and has strong detection capability.

[0006] In a first aspect, the embodiments of the present application provide a target detection method based on unitary transformation, comprising:

[0007] constructing a space-time tensor of multiple infrared images, the space-time tensor comprising a background tensor and a target tensor;

[0008] constructing a target function of the space-time tensor, the target function being obtained by constraining the background tensor using a tensor kernel norm in a unitary transformation domain and constraining the target tensor using a joint space-time total variation and L1 norm;

[0009] solving the target function to obtain the target tensor;

[0010] reconstructing the obtained target tensor into multiple single-frame target images, and outputting a detection result of each target image.

[0011] In a possible design, the target function is:

[0012]

[0013]

[0014] wherein, is the background tensor, is the target tensor, is the space-time tensor, is random noise, is the tensor kernel norm in the unitary transform domain, which is equal to the sum of the kernel norms of all positive face slices of a new tensor obtained by multiplying the modulo-three fiber of is the L1 norm of the tensor, is the Frobenius norm, is the space-time total variation, λ1, λ2 and λ3 are balance coefficients, the value range of p is 1-10, the value range of λ2 is 0.01-0.1, the value range of λ3 is 100-200, m is the maximum value of the length and width of each frame of the infrared image, and l is the total frame number of the infrared image.

[0015] In a possible design, the solving the target function to obtain the target tensor comprises:

[0016] The unitary transform matrix is constructed by using the zero-frequency component in the time domain of the space-time tensor.

[0017] The target function is solved based on the unitary transform matrix to obtain the target tensor.

[0018] In a possible design, the unitary transform matrix is constructed by using the zero-frequency component in the time domain of the space-time tensor, and the method comprises:

[0019] One-dimensional Fourier transform is performed on the m*n modulo-three fibers of the space-time tensor respectively to obtain a first space-time tensor, m and n are the length and width of the infrared image respectively;

[0020] The zero-frequency component in the first space-time tensor is retained to obtain a second space-time tensor;

[0021] One-dimensional inverse Fourier transform is performed on the m*n modulo-three fibers of the second space-time tensor respectively to obtain a third space-time tensor;

[0022] The singular value decomposition is performed on the modulo-three fiber unfolding matrix of the third space-time tensor, and the conjugate transpose of the obtained left singular matrix is taken as the unitary transform matrix.

[0023] In a possible design, the target function is solved based on the unitary transform matrix to obtain the target tensor, and the method comprises:

[0024] An auxiliary variable and The target function is simplified to a first target function shown in the following formula:

[0025]

[0026]

[0027] wherein, is a time-varying full variation operator;

[0028] The augmented Lagrangian function of the first objective function is:

[0029]

[0030] wherein, y1, y2, y = [y v , y h , y t ] are Lagrange multipliers, and β represents a penalty factor;

[0031] Solving the Lagrangian function, the target tensor is solved.

[0032] In a possible design, the solving the Lagrangian function, the target tensor is solved, includes:

[0033] For the background tensor, in the k+1 iteration, let

[0034] Multiply each modulo three fiber of by the unitary transformation matrix A to obtain

[0035] Perform singular value shrinkage processing on each positive face slice of to obtain the updated

[0036] Then the calculation formula of is as follows:

[0037]

[0038] wherein, A H is the conjugate transpose matrix of the unitary transformation matrix A, and i and j are natural numbers greater than 0;

[0039] For the target tensor, in the k+1 iteration, The calculation formula of

[0040]

[0041] wherein, sign() is a sign function;

[0042] In response to reaching a preset iteration stop condition, the calculation is stopped and the target tensor is output.

[0043] In a possible design, the preset iteration stopping condition is that the number of iterations reaches a preset maximum number of iterations or

[0044] In a second aspect, an embodiment of the present application further provides a target detection device based on unitary transformation, comprising:

[0045] A first constructing module, configured to construct a space-time tensor of the multiple infrared images, wherein the space-time tensor comprises a background tensor and a target tensor;

[0046] A second constructing module, configured to construct an objective function of the space-time tensor, wherein the objective function is obtained by constraining the background tensor by using a tensor core norm in a unitary transformation domain and constraining the target tensor by using a joint space-time total variation and L1 norm;

[0047] A solving module, configured to solve the objective function to obtain the target tensor;

[0048] A reconstructing and outputting module, configured to reconstruct the obtained target tensor into multiple single-frame target images, and output a detection result of each target image.

[0049] In a third aspect, an embodiment of the present application further provides a computing device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method in any of the embodiments.

[0050] In a fourth aspect, an embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and when the computer program is executed in a computer, the computer executes the method in any of the embodiments.

[0051] The embodiment of the present application provides a target detection method and device based on unitary transformation, electronic equipment and storage medium. The method constructs a space-time tensor by using multiple infrared images, can perform data processing in a high-dimensional space, not only retains the data structure of the original infrared image, but also fully utilizes the time domain information of the infrared image sequence, so that more image information is obtained, and prior information such as the shape and motion trajectory of the target is obtained. Then, the background tensor is constrained based on the tensor core norm in the unitary transformation domain, and the target tensor is constrained by using the joint space-time total variation and L1 norm, so as to construct a target function of the space-time tensor; then the target function is solved, and the target tensor is solved. In the two steps, the tensor core norm based on the unitary transformation is used to describe the low rank of the space-time tensor in the infrared image, compared with the tensor core norm based on the Fourier transformation, a lower tensor rank can be obtained. The space-time total variation is used to constrain the target tensor, fully describes the spatial and temporal continuity of the target tensor, enhances the internal smoothness of the target tensor, and improves the detection performance in a complex scene. Therefore, the method can accurately detect the target in a complex background, and has strong detection capability. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0053] Figure 1 It is a flow chart of a target detection method based on unitary transformation provided by an embodiment of the present application;

[0054] Figure 2 It is a flow chart of a target detection method based on unitary transformation provided by an embodiment of the present application; Figure 1

[0055] Fig. 3(a) is an infrared image containing a small target;

[0056] Fig. 3(b) is a three-dimensional distribution diagram of the infrared image shown in Fig. 3(a);

[0057] Fig. 4(a) is a target image obtained by detecting the target in Fig. 3(a) by using the method of the present application;

[0058] Fig. 4(b) is a three-dimensional distribution diagram of the target image shown in Fig. 4(a);

[0059] Fig. 5(a) is a target image obtained by detecting the target in Fig. 3(a) by using a local contrast (LCM) method; ​

[0060] Fig. 5(b) is a three-dimensional distribution diagram of the target image shown in Fig. 5(a);

[0061] Fig. 6 is a target image obtained by target detection on Fig. 3(a) by using an infrared patch image (IPI) method;

[0062] Fig. 6(b) is a three-dimensional distribution diagram of the target image shown in Fig. 6(a);

[0063] Fig. 7 is a target image obtained by target detection on Fig. 3(a) by using a reweighted infrared patch tensor (RIPT) method;

[0064] Fig. 7(b) is a three-dimensional distribution diagram of the target image shown in Fig. 7(a);

[0065] Fig. 8 is a target image obtained by target detection on Fig. 3(a) by using a partial sum of the tensor nuclear norm (PSTNN) method;

[0066] Fig. 8(b) is a three-dimensional distribution diagram of the target image shown in Fig. 8(a);

[0067] Figure 9 Fig. 1 is a hardware architecture diagram of a computing device according to an embodiment of the present application;

[0068] Figure 10 Fig. 2 is a structure diagram of a target detection device based on unitary transformation according to an embodiment of the present application. DETAILED DESCRIPTION

[0069] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in connection with the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0070] The specific implementation of the above concept will be described below.

[0071] Referring to Figure 1 The embodiments of the present application provide a target detection method based on unitary transformation, which comprises:

[0072] Step 100, constructing a space-time tensor of multiple infrared images, the space-time tensor comprising a background tensor and a target tensor;

[0073] In step 102, a target function of a space-time tensor is constructed, and the target function is obtained by constraining a background tensor by using a tensor kernel norm in a unitary transformation domain and constraining a target tensor by using a joint space-time total variation and L1 norm.

[0074] In step 104, the target function is solved to obtain the target tensor.

[0075] In step 106, the obtained target tensor is reconstructed into a plurality of single-frame target images, and a detection result of each target image is output.

[0076] The embodiment of the present application provides a target detection method based on a unitary transformation. The method can process data in a high-dimensional space by constructing a space-time tensor by using a plurality of infrared images, retains the data structure of the original infrared image, fully utilizes time domain information of the infrared image sequence, thereby obtains more image information, and obtains prior information such as a shape and a motion trajectory of a target. Then, the target function of the space-time tensor is constructed by constraining the background tensor by using a tensor kernel norm in a unitary transformation domain and constraining the target tensor by using a joint space-time total variation and L1 norm. Then, the target function is solved to obtain the target tensor. In the two steps, the tensor kernel norm based on the unitary transformation is used to describe the low rank of the space-time tensor in the infrared image, and compared with the tensor kernel norm based on the Fourier transformation, a lower tensor rank can be obtained. The space-time total variation is used to constrain the target tensor, fully describes the spatial and temporal continuity of the target tensor, enhances the internal smoothness of the target tensor, and improves the detection performance in a complex scene. Therefore, the method can accurately detect a target in a complex background and has strong detection capability.

[0077] In the embodiment, the target in the infrared image can be a target of various sizes, and has good detection effect on small targets.

[0078] The execution manner of each step is described below. Figure 1

[0079] Firstly, for step 100, a space-time tensor of a plurality of infrared images is constructed, and the space-time tensor includes a background tensor and a target tensor.

[0080] In the step, the number of the plurality of infrared images is at least 3 frames, and the sampling time interval between adjacent two infrared images is less than a preset time interval. In this way, the time domain information of the infrared image sequence can be fully utilized, the continuity of the target in the space and the time domain can be fully utilized, and more accurate detection results can be obtained. Of course, the plurality of infrared images are preferably continuous infrared images.

[0081] ​In this step, the target tensor and the background tensor are unknown tensors, and a target function of the space-time tensor is constructed to solve the two tensors.

[0082] Then, for step 102, a target function of the space-time tensor is constructed, which is obtained by constraining the background tensor by using the tensor core norm in the unitary transform domain and constraining the target tensor by using the joint space-time total variation and L1 norm.

[0083] In this step, the tensor core norm based on the unitary transform is used to characterize the low rank of the space-time tensor in the infrared image, and the unitary transform matrix used is related to the zero frequency component in the time dimension of the space-time tensor. Compared with the tensor core norm based on the Fourier transform, a lower tensor rank is obtained. The space-time total variation is used to constrain the target tensor, which fully describes the spatial and temporal continuity of the target tensor, enhances the internal smoothness of the target tensor, and improves the detection performance in complex scenes. Compared with the infrared small target detection method of using total variation to constrain the background tensor, the running time can be effectively reduced.

[0084] In some embodiments, the target function is:

[0085]

[0086]

[0087] In the formula, is the background tensor, is the target tensor, is the space-time tensor, is random noise, is the tensor core norm in the unitary transform domain, which is equal to the sum of the core norms of all positive slices of the new tensor obtained by multiplying the modulus three fiber of is the L1 norm of the tensor, is the Frobenius norm, is the space-time total variation, λ1, λ2 and λ3 are balance coefficients, the value range of p is 1-10, the value range of λ2 is 0.01-0.1, the value range of λ3 is 100-200, and m is the maximum value of the length and width of each infrared image, and l is the total number of infrared images.

[0088] Then, for step 104, the target function is solved to obtain the target tensor.

[0089] Of course, by solving the target function, the background tensor can also be solved.

[0090] In some embodiments, solving the target function to obtain the target tensor includes:

[0091] The unitary transformation matrix is constructed by using the zero frequency component of the space-time tensor time domain;

[0092] The target function is solved based on the unitary transformation matrix, and the target tensor is solved.

[0093] In some embodiments, the unitary transformation matrix is constructed by using the zero frequency component of the space-time tensor time domain, comprising:

[0094] One-dimensional Fourier transform is performed on the m*n mod 3 fibers of the space-time tensor respectively, to obtain a first space-time tensor, m and n are the length and width of the infrared image respectively;

[0095] The zero frequency component in the first space-time tensor is retained to obtain a second space-time tensor;

[0096] One-dimensional inverse Fourier transform is performed on the m*n mod 3 fibers of the second space-time tensor respectively, to obtain a third space-time tensor; the third space-time tensor can be used as an estimated value of the background tensor;

[0097] The singular value decomposition is performed on the mod 3 fiber expansion matrix of the third space-time tensor, and the conjugate transpose of the obtained left singular matrix is used as the unitary transformation matrix.

[0098] In some embodiments, the target function is solved based on the unitary transformation matrix, and the target tensor is solved, comprising:

[0099] An auxiliary variable is introduced And The target function is simplified to a first target function shown in the following formula:

[0100]

[0101]

[0102] In the formula, is a space-time total variation operator, which includes three components, D v , D h and D t , which respectively represent the difference operators in the vertical direction, the horizontal direction and the time domain.

[0103] The augmented Lagrange function of the first target function is:

[0104]

[0105] In the formula, y1, y2, y = [y v , y h , y t ] are Lagrange multipliers, and β represents a penalty factor;

[0106] The Lagrange function is solved to solve the target tensor.

[0107] In some embodiments, the Lagrangian function is solved to obtain the target tensor, including:

[0108] For the background tensor, in the k+1 iteration, let

[0109] Multiply each modulo three fiber of by the unitary transformation matrix A to obtain

[0110] For each frontal slice of , singular value shrinkage processing is performed to obtain the updated

[0111] Taking singular value shrinkage processing on the Lth frontal slice as an example:

[0112] Singular value decomposition is performed on the Lth frontal slice of to obtain the left singular matrix U L , the right singular matrix V L , and the diagonal matrix ∑ L , and singular value shrinkage processing is performed on the frontal slice to obtain

[0113]

[0114] Iterate through all the frontal slices of and perform singular value shrinkage processing to obtain the updated

[0115] Thus, the calculation formula of is obtained:

[0116]

[0117] In the formula, A H is the conjugate transpose matrix of the unitary transformation matrix A, and i and j are natural numbers greater than 0;

[0118] For the target tensor, in the k+1 iteration, the calculation formula is as follows:

[0119]

[0120] Wherein, sign() is the sign function;

[0121] In response to reaching the preset iteration stopping condition, the calculation is stopped and the target tensor is output. Of course, after the iteration calculation is completed, the user can also solve the background tensor according to the need.

[0122] In some embodiments, the preset iteration stopping condition is that the number of iterations reaches a preset maximum number of iterations or

[0123] Finally, for step 106, the solved target tensor is reconstructed into a plurality of single-frame target images, and a detection result of each target image is output.

[0124] The beneficial effects of the method of the present application are illustrated below with a specific embodiment.

[0125] As Figure 2 shown, a flowchart of the method of the present application is shown in the figure, in which the number of infrared images is 3, and the size of each infrared image is 256x256. One of the infrared images contains a small target, and its three-dimensional distribution map is shown in FIG. 3(a) and FIG. 3(b). In the infrared image, the center of the target is located at (145, 115). The imaging background is complex, and the background contains forests, grasslands, farmlands, etc. Due to the existence of many strong radiation and noise from the ground, in addition to the target, there are many high-brightness areas and high-brightness noise points in the image, and these high-brightness areas will interfere with the detection of the target. If the inter-frame information in the infrared image sequence is not used, it is difficult to detect the target.

[0126] The inventors respectively used the method of the present application, the LCM method, the IPI method, the RIPT method and the PSTNN method to process the infrared image shown in FIG. 3(a), and the detection results are shown in FIG. 4(a), FIG. 4(b), FIG. 5(a), FIG. 5(b), FIG. 6(a), FIG. 6(b), FIG. 7(a), FIG. 7(b) and FIG. 8(a), FIG. 8(b) respectively. Figures 4(b) to 8(a) From FIG. 4(a) and FIG. 4(b), it can be seen that, by using the method of the present application, the background is effectively suppressed, and only the target is retained in the image. From FIG. 5(a) and FIG. 5(b), it can be seen that the image obtained by the LCM method has a significant "blocky" effect, and is very sensitive to noise. From FIG. 6(a), FIG. 6(b), FIG. 7(a) and FIG. 7(b), it can be seen that the IPI method, the RIPT method and the PSTNN method have a high gray value for the pixels other than the small target, which is easy to cause false alarm, and these three methods are difficult to completely solve the influence of high-brightness noise points and clutter on the target. Figures 6(b) to 8(a)

[0127] It can be seen that, by constructing a space-time tensor and simultaneously using space-time total variation to constrain the target tensor in time, the time domain information is effectively used, the influence of noise and high-brightness areas is well reduced, and the target is detected.

[0128] As Figure 9 , Figure 10 shown, the embodiment of the present application provides a target detection device based on unitary transformation. The device embodiment can be realized by software, or realized by hardware or a combination of software and hardware. From the hardware layer, as shown in FIG. 9, the device embodiment includes a receiving unit 901, a processing unit 902 and a sending unit 903.​Figure 9 The diagram shown is a hardware architecture diagram of a computing device housing a target detection device based on unitary transform according to an embodiment of the present invention. (Except for...) Figure 9 In addition to the processor, memory, network interface, and non-volatile memory shown, the computing device in the embodiment may also include other hardware, such as a forwarding chip responsible for processing packets. Taking software implementation as an example, such as... Figure 10 As shown, a logical device is formed by the CPU of its computing device reading the corresponding computer program from non-volatile memory into memory and running it. This embodiment provides a target detection device based on unitary transform, comprising:

[0129] The first construction module 1000 is used to construct the spatiotemporal tensor of multiple frames of infrared images. The spatiotemporal tensor includes a background tensor and a target tensor.

[0130] The second construction module 1002 is used to construct the objective function of the space-time tensor. The objective function is obtained by constraining the background tensor using the tensor nuclear norm in the unitary transform domain and constraining the target tensor using the joint space-time total variation and L1 norm.

[0131] Solver module 1004 is used to solve for the objective function and obtain the objective tensor;

[0132] The reconstruction output module 1006 is used to reconstruct the solved target tensor into multiple single-frame target images and output the detection results of each target image.

[0133] In this embodiment of the invention, the first construction module 1000 can be used to execute step 100 in the above method embodiment, the second construction module 1002 can be used to execute step 102 in the above method embodiment, the solving module 1004 can be used to execute step 104 in the above method embodiment, and the reconstruction output module 1006 can be used to execute step 106 in the above method embodiment.

[0134] In one embodiment of the present invention, the objective function is:

[0135]

[0136]

[0137] In the formula, For the background tensor, For the target tensor, When the tensor is empty, It is random noise. The tensor nuclear norm in the unitary transformation domain is numerically equal to The sum of the nuclear norms of all frontal slices of the new tensor obtained by multiplying the modulo-3 fiber by the unitary transformation matrix A. is the L1 norm of the tensor, is the Frobenius norm, is the total variation in time, λ1, λ2 and λ3 are balance coefficients, p is in the range of 1-10, λ2 is in the range of 0.01-0.1, λ3 is in the range of 100-200, m is the maximum value of length and width of each infrared image, and l is the total number of infrared images.

[0138] In the embodiment of the application, the solving module 1004 is configured to perform:

[0139] constructing the unitary transformation matrix by using the zero-frequency component of the space-time tensor in the time domain;

[0140] solving the target function based on the unitary transformation matrix to obtain the target tensor.

[0141] In the embodiment of the application, the unitary transformation matrix is constructed by using the zero-frequency component of the space-time tensor in the time domain, including:

[0142] performing one-dimensional Fourier transform on the m*n mod-3 fibers of the space-time tensor to obtain a first space-time tensor, m and n are respectively the length and width of the infrared image;

[0143] retaining the zero-frequency component in the first space-time tensor to obtain a second space-time tensor;

[0144] performing one-dimensional inverse Fourier transform on the m*n mod-3 fibers of the second space-time tensor to obtain a third space-time tensor;

[0145] performing singular value decomposition on the mod-3 fiber expansion matrix of the third space-time tensor, and taking the conjugate transpose of the obtained left singular matrix as the unitary transformation matrix.

[0146] In the embodiment of the application, the target function is solved based on the unitary transformation matrix to obtain the target tensor, including:

[0147] introducing auxiliary variables and simplifying the target function into a first target function shown in the following formula:

[0148]

[0149]

[0150] in the formula, is a space-time total variation operator;

[0151] the augmented Lagrangian function of the first target function is:

[0152]

[0153] In the formula, y1, y2, y = [y v h t are Lagrange multipliers, and β represents a penalty factor.

[0154] Solve the Lagrange function to solve the target tensor.

[0155] In the embodiment of the present application, solving the Lagrange function to solve the target tensor comprises:

[0156] For the background tensor, in the k+1 iteration, let

[0157] Multiply each modulo three fiber of by the unitary transformation matrix A to obtain

[0158] Perform singular value shrinkage processing on each positive face slice of to obtain the updated

[0159] Then the calculation formula of is as follows:

[0160]

[0161] In the formula, A H is the conjugate transpose matrix of the unitary transformation matrix A, and i and j are natural numbers greater than 0.

[0162] For the target tensor, in the k+1 iteration, The calculation formula of

[0163]

[0164] Wherein, sign() is a sign function.

[0165] In response to reaching the preset iteration stopping condition, stop calculating and output the target tensor.

[0166] In the embodiment of the present application, the preset iteration stopping condition is that the number of iterations reaches a preset maximum number of iterations or

[0167] It can be understood that the structure illustrated in the embodiment of the present application does not constitute a specific limitation on a target detection device based on unitary transformation. In other embodiments of the present application, a target detection device based on unitary transformation can include more or fewer components than the illustration, or combine certain components, or split certain components, or different component arrangement. The components illustrated can be implemented in hardware, software, or a combination of software and hardware.​​

[0168] The information interaction, execution process and the like between the modules in the above device are based on the same concept as the method embodiments of the present application, and the specific content can be referred to the description in the method embodiments of the present application, which will not be described here.

[0169] The embodiment of the present application further provides a computing device, comprising a memory and a processor, the memory stores a computer program, and the processor executes the computer program to realize the unitary transformation-based target detection method in any of the embodiments of the present application.

[0170] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the computer program, when executed by a processor, causes the processor to execute the unitary transformation-based target detection method in any of the embodiments of the present application.

[0171] Specifically, a system or device equipped with a storage medium can be provided, the storage medium stores a software program code for realizing the functions of any of the above embodiments, and the computer (or CPU or MPU) of the system or device reads out and executes the program code stored in the storage medium.

[0172] In this case, the program code read from the storage medium itself can realize the functions of any of the above embodiments, and therefore the program code and the storage medium storing the program code constitute a part of the present application.

[0173] The storage medium for providing the program code includes a floppy disk, a hard disk, a magneto-optical disk, an optical disk (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), a magnetic tape, a non-volatile memory card, and a ROM. Alternatively, the program code can be downloaded from a server computer via a communication network.

[0174] In addition, it should be clear that not only the program code read by the computer can be executed, but also part or all of the actual operations can be completed by the operating system and the like operating on the computer based on the instructions of the program code, so as to realize the functions of any of the above embodiments.

[0175] In addition, it can be understood that the program code read from the storage medium is written into the memory provided in the expansion board inserted into the computer or the memory provided in the expansion module connected to the computer, and then part and all of the actual operations are executed by the CPU and the like installed on the expansion board or the expansion module based on the instructions of the program code, so as to realize the functions of any of the above embodiments.

[0176] It should be noted that, in the present document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0177] Those skilled in the art can understand that all or part of the steps of the above-mentioned method embodiments can be completed by program instruction related hardware, and the foregoing program can be stored in a computer readable storage medium, and the program performs the steps of the above-mentioned method embodiments when executed; and the foregoing storage medium includes various storage media that can store program codes, such as ROM, RAM, magnetic disk or optical disk.

[0178] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for target detection based on unitary transformation, characterized in that, The method comprises the following steps: constructing a space-time tensor of multiple infrared images, the space-time tensor comprising a background tensor and a target tensor; constructing an objective function of the space-time tensor, the objective function being obtained by constraining the background tensor by using a tensor kernel norm in a unitary transformation domain and constraining the target tensor by using a joint space-time total variation and L1 norm; solving the objective function to obtain the target tensor; reconstructing the obtained target tensor into multiple single-frame target images, and outputting a detection result of each target image; the solving of the objective function to obtain the target tensor comprises: constructing a unitary transformation matrix by using zero-frequency components in a time domain of the space-time tensor; solving the objective function based on the unitary transformation matrix to obtain the target tensor; the construction of the unitary transformation matrix by using the zero-frequency components in the time domain of the space-time tensor comprises: performing one-dimensional Fourier transform on m*n mod-3 fibers of the space-time tensor respectively to obtain a first space-time tensor, m and n being the length and the width of the infrared image respectively; retaining zero-frequency components in the first space-time tensor to obtain a second space-time tensor; performing one-dimensional inverse Fourier transform on m*n mod-3 fibers of the second space-time tensor respectively to obtain a third space-time tensor; performing singular value decomposition on a mod-3 fiber expansion matrix of the third space-time tensor, and taking a conjugate transpose of a left singular matrix obtained as the unitary transformation matrix.

2. The method of claim 1, wherein, the objective function is as follows: In the formula, is the background tensor, is the target tensor, is the space-time tensor, is random noise, is the tensor kernel norm in the unitary transform domain, and the numerical value is equal to the modulus three fiber multiplied by the unitary transformation matrix the sum of the kernel norms of all positive slices of the new tensor obtained after is the L1 norm of the tensor, is the Frobenius norm, is the space-time total variation, , and is the balance coefficient, , the value range of p is 1-10, the value range of is 0.01-0.1, the value range of is 100-200, m is the maximum value of the length and width of each frame of the infrared image, is the total number of frames of the infrared image.

3. The method of claim 1, wherein, the solving of the objective function based on the unitary transformation matrix to obtain the target tensor comprises: Introducing auxiliary variables and simplifying the objective function to a first objective function shown in the following equation: In the formula, for the spatial total variation operator; an augmented Lagrange function of the first objective function is as follows: wherein is a Lagrange multiplier, denotes a penalty factor; solving the Lagrange function to obtain the target tensor.

4. The method of claim 3, wherein, the solving of the Lagrange function to obtain the target tensor comprises: For the background tensor, in the k+1 iteration, let ; right Each modulo 3 fiber is multiplied by the unitary transformation matrix. get ; For each frontal slice of the singular value shrinkage processing is performed to obtain an updated ; Then The calculation formula is: wherein is the conjugate transpose matrix of the unitary transformation matrix A, and i, j are natural numbers greater than 0, respectively. For the target tensor, in k+1 iteration, The calculation formula is as follows: wherein , is the sign function; in response to a preset iteration stop condition being reached, stopping calculation and outputting the target tensor.

5. The method of claim 4, wherein, The preset iteration stopping condition is that the iteration number reaches a preset maximum iteration number or .

6. An apparatus for target detection based on unitary transformation, characterized by The method comprises the following steps: a first construction module is configured to construct a space-time tensor of multiple infrared images, the space-time tensor comprising a background tensor and a target tensor; a second construction module is configured to construct an objective function of the space-time tensor, the objective function being obtained by constraining the background tensor by using a tensor kernel norm in a unitary transformation domain and constraining the target tensor by using a joint space-time total variation and L1 norm; a solving module is configured to solve the objective function to obtain the target tensor; a reconstruction and output module is configured to reconstruct the obtained target tensor into multiple single-frame target images, and output a detection result of each target image; the solving module is configured to perform the following operations: construct a unitary transformation matrix by using zero-frequency components in a time domain of the space-time tensor; solve the objective function based on the unitary transformation matrix to obtain the target tensor; the construction of the unitary transformation matrix by using the zero-frequency components in the time domain of the space-time tensor comprises: perform one-dimensional Fourier transform on m*n mod-3 fibers of the space-time tensor respectively to obtain a first space-time tensor, m and n being the length and the width of the infrared image respectively; retain zero-frequency components in the first space-time tensor to obtain a second space-time tensor; perform one-dimensional inverse Fourier transform on m*n mod-3 fibers of the second space-time tensor respectively to obtain a third space-time tensor; performing one-dimensional inverse Fourier transform on the m*n modulo-3 fibers of the second space-time tensor to obtain a third space-time tensor; performing singular value decomposition on the modulo-3 fiber unfolding matrix of the third space-time tensor, and taking the conjugate transpose of the left singular matrix obtained as the unitary transformation matrix. 7.A computing device comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method according to any one of claims 1-5. 8.A computer readable storage medium storing a computer program, wherein the computer program, when executed in a computer, causes the computer to perform the method according to any one of claims 1-5.

Images