Target Tracking Method, System, Terminal Device and Storage Medium

Abstract

The present application is applicable to the field of computer vision technology, and discloses a target tracking method, a terminal device, and a storage medium. After sampling the target area of the next frame according to the target tracking result of the previous frame or the initially selected target to obtain a sample set, the method of the present application calculates the first confidence of each candidate target in the sample set according to the BLAS-3 sparse discriminant classification. Moreover, the similarity of the anti-occlusion histogram of each candidate target in the sample set is calculated according to the BLAS-3 sparse generation model. Then, the collaborative confidence of each candidate target is calculated according to the first confidence and the similarity, and the candidate target corresponding to the collaborative confidence that meets the preset requirements is selected as the tracking target of the next frame, so that the computational efficiency of the robust target tracking based on sparse representation can be effectively improved without sacrificing the accuracy and robustness of the target tracking method, and further the frame rate of the target tracking can be improved.

Description

Technical Field

[0001] This application relates to the field of computer vision technology, and particularly to an object tracking method, a terminal device, and a storage medium. Background Art

[0002] Object tracking refers to relying on computer vision to identify and track a specified object in a real-time video stream, and estimating its position and size. Traditional image sparse representation represents each picture as a vector, then generates a dictionary matrix through the K-mean algorithm, and finally solves the error minimization problem of the L1 coefficient penalty term for multiple picture vectors one by one to obtain a sparse matrix composed of sparse vector solutions of multiple pictures. However, this method can only utilize inefficient second-level basic linear algebra subroutines and cannot efficiently utilize computer resources, resulting in a slow speed of the traditional image sparse representation algorithm, which restricts the calculation speed of the object tracking algorithm. Summary of the Invention

[0003] Embodiments of this application disclose an object tracking method, a terminal device, and a storage medium, which can process the object tracking process through third-level basic algebra subroutines, effectively improving the calculation efficiency of robust object tracking based on sparse representation.

[0004] Embodiments of this application disclose an object tracking method, including:

[0005] Sampling the next-frame object area according to the previous-frame object tracking result or the initially selected object to obtain a plurality of candidate region samples as a sample set;

[0006] Based on the sparse discriminant classification of a preset iterative shrinkage algorithm, calculating a first confidence level for each candidate object in the sample set, where the preset iterative shrinkage algorithm includes third-level basic algebra subroutines;

[0007] Based on the sparse generation model of the preset iterative shrinkage algorithm, calculating the similarity of the anti-occlusion histogram of each candidate object in the sample set;

[0008] Calculating a collaborative confidence level for each candidate object according to the first confidence level and the similarity;

[0009] Selecting the candidate object corresponding to the collaborative confidence level that meets the preset requirements as the tracking object for the next frame.

[0010] Optionally, the preset iterative shrinkage algorithm includes:

[0011] Calculating a basic iteration;

[0012] Calculating a first sparse matrix representation or a first non-negative sparse matrix representation according to the basic iteration;

[0013] Calculating an iteration interval;

[0014] Calculate the Taylor station expansion point according to the iteration interval and the first sparse matrix representation, or calculate the Taylor station expansion point according to the iteration interval and the first non-negative sparse matrix representation;

[0015] When the iteration error satisfies the error condition, obtain the solution of the first sparse matrix representation or obtain the solution of the first non-negative sparse matrix representation.

[0016] Optionally, the sparse discriminant classification based on the preset iterative shrinkage algorithm calculates the first confidence of each candidate target in the sample set, including:

[0017] Obtain a projection diagonal matrix;

[0018] Project the template matrix and the candidate target matrix into the dynamic discrimination space through the diagonal matrix to obtain a projected template matrix and a projected candidate target matrix, where the candidate target matrix corresponds to the sample set and the template matrix corresponds to the template set;

[0019] Calculate the first reconstruction error of the second sparse matrix representation of the candidate target according to the projected template matrix, the projected candidate target matrix and the preset iterative shrinkage algorithm;

[0020] Calculate the first confidence of each candidate target according to the first reconstruction error.

[0021] Optionally, the sparse generation model based on the preset iterative shrinkage algorithm calculates the similarity of the anti-occlusion histograms of each candidate target in the sample set, including:

[0022] Obtain a dictionary matrix;

[0023] Obtain the sample matrix of multiple standard sliding window thumbnails in the sample set;

[0024] Calculate the second non-negative sparse matrix representation of the standard sliding window thumbnail according to the dictionary matrix and the sample matrix;

[0025] Obtain the anti-occlusion histogram of the standard sliding window thumbnail;

[0026] Calculate the similarity between the template and the anti-occlusion histogram of the standard sliding window thumbnail.

[0027] Optionally, the obtaining of the projection diagonal matrix includes:

[0028] Obtain a sparse coefficient vector;

[0029] Generate a projection diagonal matrix according to the elements of the sparse coefficient vector.

[0030] Optionally, the obtaining of the dictionary matrix includes:

[0031] Calculate several cluster centers of the initial standard sliding window thumbnail;

[0032] Combine all the cluster centers into a dictionary matrix.

[0033] Optionally, obtaining the anti-occlusion histogram of the standard sliding window thumbnail includes:

[0034] Calculate a reconstruction error matrix according to the dictionary matrix, the sample matrix, and the sparse coefficient matrix;

[0035] Calculate an occlusion matrix according to the elements in the reconstruction error matrix;

[0036] Calculate the anti-occlusion histogram of the standard sliding window thumbnail according to the occlusion matrix and the sparse coefficient matrix.

[0037] An embodiment of the present application discloses an object tracking system, including:

[0038] A first module, configured to sample the target area of the next frame according to the previous frame of object tracking result or the initially selected object, and obtain several candidate region samples as a sample set;

[0039] A second module, configured to calculate the first confidence of each candidate object in the sample set based on sparse discriminant classification of a preset iterative shrinkage algorithm, and the preset iterative shrinkage algorithm includes three-level basic algebraic subroutines;

[0040] A third module, configured to calculate the similarity of the anti-occlusion histogram of each candidate object in the sample set based on a sparse generation model of a preset iterative shrinkage algorithm;

[0041] A fourth module, configured to calculate the collaborative confidence of each candidate object according to the first confidence and the similarity;

[0042] A fifth module, configured to select the candidate object corresponding to the collaborative confidence that meets the preset requirements as the tracking object of the next frame.

[0043] An embodiment of the present application discloses a terminal device, including:

[0044] A memory, in which a computer program is stored;

[0045] A processor, when the computer program is executed by the processor, enables the processor to implement the above object tracking method.

[0046] An embodiment of the present application discloses a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above object tracking method is implemented.

[0047] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0048] After sampling the target area of the next frame to obtain a sample set according to the target tracking result of the previous frame or the initially selected target, the first confidence of each candidate target in the sample set is calculated according to the sparse discriminant classification of the three-level basic algebraic subroutine. Moreover, the similarity of the anti-occlusion histogram of each candidate target in the sample set is calculated according to the sparse generation model of the three-level basic algebraic subroutine. Then, the collaborative confidence of each candidate target is calculated based on the first confidence and the similarity, and the candidate target corresponding to the collaborative confidence that meets the preset requirements is selected as the tracking target of the next frame. In this embodiment, the three-level basic algebraic subroutine is used to process the target tracking process, so that the accuracy and robustness of the target tracking method can be maintained, and the calculation efficiency of the robust target tracking based on sparse representation can be effectively improved, thereby increasing the frame rate of target tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0050] Figure 1 It is a flowchart of the implementation of the target tracking method provided by an embodiment of the present application;

[0051] Figure 2 It is a flowchart of the BLAS-3 fast iterative threshold shrinkage algorithm based on an embodiment of the present application;

[0052] Figure 3 It is a flowchart of calculating the first confidence in an embodiment of the present application;

[0053] Figure 4 It is a flowchart of step 130 in an embodiment of the present application;

[0054] Figure 5 It is a complete flowchart of the target tracking method in an embodiment of the present application;

[0055] Figure 6 It is a curve graph comparing the running times of the fast threshold shrinkage algorithm of BLAS-3 and the sparse representation algorithm in an embodiment of the present application;

[0056] Figure 7 It is a schematic structural diagram of a terminal device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] In the following description, specific details such as specific system architectures, technologies, etc. are presented for purposes of illustration and not limitation, so as to provide a thorough understanding of the embodiments of the present application. However, those skilled in the art should understand that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from obscuring the description of the present application.

[0058] It should be understood that when used in the specification of the present application and the appended claims, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0059] It should also be understood that the term "and / or" as used in the specification of the present application and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0060] As used in the specification of the present application and the appended claims, the term "if" can be interpreted as "when" or "once" or "in response to determining" or "in response to detecting" depending on the context. Similarly, the phrase "if determined" or "if [the described condition or event] is detected" can be interpreted as meaning "once determined" or "in response to determining" or "once [the described condition or event] is detected" or "in response to detecting [the described condition or event]" depending on the context.

[0061] In addition, in the description of the specification of the present application and the appended claims, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0062] The reference to "one embodiment" or "some embodiments" etc. described in the specification of the present application means that a specific feature, structure, or characteristic described in connection with that embodiment is included in one or more embodiments of the present application. Thus, statements such as "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments", etc. that appear in different places in this specification do not necessarily all refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in another way. The terms "comprising", "including", "having", and their variants all mean "including but not limited to", unless otherwise specifically emphasized in another way.

[0063] In the embodiments of the present application, the execution subject of the process is the terminal device to which the target tracking processing software belongs. The terminal device includes, but is not limited to: devices such as servers, computers, smart phones, and tablet computers that can execute the target tracking method disclosed in the present application. Figure 1 The flowchart showing the implementation of the target tracking method disclosed in the first embodiment of the present application is described in detail as follows:

[0064] In S110, according to the previous frame target tracking result or the initially selected target, the next frame target area is sampled to obtain a number of candidate area samples as a sample set.

[0065] In the embodiments of the present application, in a video recording scenario, the previous frame and the next frame refer to two adjacent frames. The previous frame target tracking result refers to the target tracking information at the previous time point corresponding to the current time point in the video corresponding to the video recording scenario. The initially selected target refers to the target specified at the beginning of target tracking in the video corresponding to the video recording scenario. The positions of a number of candidate areas may partially overlap. Each candidate area sample includes information such as samples and backgrounds within the candidate area.

[0066] In S120, based on the sparse discriminant classification of the preset iterative shrinkage algorithm, the first confidence of each candidate target in the sample set is calculated.

[0067] In the embodiments of the present application, it is assumed that the iterative shrinkage algorithm includes a fast iterative threshold shrinkage algorithm with three levels of basic algebraic subroutines (BLAS-3). Specifically, as Figure 2 shown, the specific execution process of the iterative shrinkage algorithm includes:

[0068] Step 1: Obtain the dictionary matrix D, set the initial iteration point B0 = 0, and generate the sample matrix Y according to the sample set.

[0069] Step 2: Calculate the basic iteration through formula (1):

[0070]

[0071] where B k is the Taylor expansion point calculated in the previous iteration and B0 = 0, D is the dictionary matrix, Y is the sample matrix, and L is the Lipschitz constant.

[0072] Step 3: If the current sparse matrix is a non-negative sparse matrix, calculate the non-negative soft threshold shrinkage (non-negative sparse matrix representation) through formula (2); otherwise, calculate the soft threshold shrinkage (sparse matrix representation) through formula (3):

[0073]

[0074]

[0075] Among them, X k is the algorithm solution of the current iteration, that is, X k is non - negative soft - thresholding shrinkage or soft - thresholding shrinkage, and λ is the regularization parameter represented by the sparse matrix.

[0076] Step Four: Calculate the iteration interval {t k} sequence through formula (4):

[0077]

[0078] Step Five: Calculate the next Taylor station expansion point through formula (5):

[0079]

[0080] Step Six: Determine whether the iteration error after several iterations meets the error condition. If it meets, end the iteration processing process; otherwise, return to Step Two.

[0081] In this embodiment, as Figure 3 shown, the step of calculating the first confidence level of each candidate target in the sample set based on the sparse discriminant classification of the preset iterative shrinkage algorithm includes, but is not limited to, the following steps:

[0082] In S310, obtain the projection diagonal matrix.

[0083] In the embodiment of the present application, the projection diagonal matrix is a matrix that adaptively selects an appropriate number of discriminant features in a dynamic environment. It can be obtained by first obtaining the sparse coefficient vector and then generating the projection diagonal matrix through the elements of the sparse coefficient vector. Exemplarily, as shown in formula (6), solve the sparse vector representation problem through the sparse representation of the level - 2 basic linear algebra subprogram (BLAS - 2):

[0084]

[0085] Among them, φ consists of positive and negative templates, and s and p are the attributes corresponding to the sparse coefficient vector and the template set respectively (+1 for the positive template, - 1 for the negative template).

[0086] Then generate the projection diagonal matrix S according to s (the vector elements being 0 correspond to the zero matrix, and the non - zero elements correspond to the identity matrix).

[0087] In S320, project the template matrix and the candidate target matrix into the dynamic discriminant space through the diagonal matrix to obtain the projected template matrix and the projected candidate target matrix.

[0088] In the embodiments of the present application, the candidate target matrix corresponds to the sample set, and the template matrix corresponds to the template set. After obtaining the shadow diagonal matrix S, the candidate target matrix X and the template matrix φ are projected onto the dynamic discrimination space through the shadow diagonal matrix to obtain the projected template matrix φ, and the projected candidate target matrix X,.

[0089] In S330, according to the projected template matrix, the projected candidate target matrix, and the preset iterative shrinkage algorithm, calculate the first reconstruction error of the second sparse matrix representation of the candidate target.

[0090] In the embodiments of the present application, the second sparse matrix representation is as shown in formula (7):

[0091]

[0092] Wherein, A represents the second sparse matrix.

[0093] In S340, calculate the first confidence level H of each candidate target according to the first reconstruction error c 。

[0094] In the embodiments of the present application, the first confidence level is calculated through formula (8):

[0095]

[0096] Wherein, σ represents a positive constant.

[0097] In S130, based on the sparse generation model of the preset iterative shrinkage algorithm, calculate the similarity of the anti-occlusion histogram of each candidate target in the sample set.

[0098] In the embodiments of the present application, as Figure 4 shown, step S130 includes, but is not limited to, the following steps:

[0099] In S410, obtain the dictionary matrix.

[0100] In this embodiment, after calculating several cluster centers of the initial quasi-sliding window thumbnail, all the cluster centers can be combined into a dictionary matrix. Specifically, the cluster centers can be calculated by the K-mean (k-means clustering algorithm, also known as the k-means clustering algorithm) algorithm. Among them, the k-means clustering algorithm is an iterative clustering analysis algorithm. The steps of this clustering algorithm are to pre-divide the data into K groups, then randomly select K objects as the initial cluster centers, and then calculate the distance between each object and each seed cluster center, and assign each object to the cluster center closest to it. The cluster centers and the objects assigned to them represent a cluster. Each time a sample is assigned, the cluster center of the cluster will be recalculated according to the existing objects in the cluster. This process will be repeated continuously until a certain termination condition is met. The termination condition can be that no (or the minimum number of) objects are reassigned to different clusters, no (or the minimum number of) cluster centers change anymore, and the sum of squared errors is locally minimized.

[0101] In S420, obtain the sample matrix of multiple standard sliding window thumbnails in the sample set; calculate the second non-negative sparse matrix representation of the standard sliding window thumbnail according to the dictionary matrix and the sample matrix;

[0102] In this embodiment, the second non-negative sparse matrix representation can be calculated by formula (9):

[0103]

[0104] Among them, Y and B respectively obtain the sample matrix and the sparse coefficient matrix of multiple standard thumbnails through a sliding window, and D represents the dictionary matrix.

[0105] In S430, obtain the anti-occlusion histogram of the standard sliding window thumbnail.

[0106] In this embodiment, after calculating the reconstruction error matrix according to the dictionary matrix, the sample matrix and the sparse coefficient matrix, the occlusion matrix can be calculated according to the elements in the reconstruction error matrix, and then the anti-occlusion histogram of the standard sliding window thumbnail can be calculated according to the occlusion matrix and the sparse coefficient matrix. Specifically, the calculation process of the occlusion matrix can be described as follows:

[0107] Calculate the elements of the occlusion matrix O by formula (10):

[0108]

[0109] Among them, ε0 is a preset threshold, and ε ij represents the elements in the reconstruction error matrix E.

[0110] Calculate the anti-occlusion histogram by formula (11):

[0111]

[0112] Among them, q i (i = 1, 2,..., M) is the column vector in matrix Q, represents the multiplication of matrix elements.

[0113] In S440, calculate the similarity between the template and the anti-occlusion histogram of the standard sliding window thumbnail.

[0114] In this embodiment, by obtaining the similarity L c between the template and the anti-occlusion histogram of the candidate target, and judging whether to update the anti-occlusion histogram of the template according to the occlusion degree in each iteration. Specifically, the similarity between the template and the anti-occlusion histogram of the standard sliding window thumbnail can be calculated through formula (12):

[0115]

[0116] Among them, the j-th standard thumbnail and Ψ j is the template histogram.

[0117] In this embodiment, the template histogram Ψ j can be updated through formula (13):

[0118] Ψ j = μψ f +(1 - μ)ψ l If Q n < a formula (13)

[0119] Among them, ψ f , ψ l are the template histograms of the first frame and the current frame respectively, a is a preset threshold, and Q n is the occlusion matrix of the current frame.

[0120] In S140, calculate the collaborative confidence of each candidate target according to the first confidence and the similarity; select the candidate target corresponding to the collaborative confidence that meets the preset requirements as the tracking target for the next frame.

[0121] In this embodiment, the collaborative confidence can be calculated through formula (14):

[0122]

[0123] Specifically, in this embodiment, meeting the preset requirements can be the collaborative confidence with the largest value among all collaborative confidences. The candidate target corresponding to the largest collaborative confidence is used as the tracking target for the next frame for tracking, so as to effectively improve the tracking accuracy.

[0124] Such asFigure 5 As shown in the figure, the complete implementation process of the target tracking method in the embodiments of the present application is as follows:

[0125] Step 1: Based on the target tracking result of the previous frame or the initially selected target x1, use particle filtering to sample the target candidate regions in the next frame;

[0126] Step 2: Solve the sparse matrix representation problem through the fast iterative shrinkage thresholding algorithm based on BLAS-3, specifically as shown in formula (15):

[0127]

[0128] where T, X, and Γ are the candidate template set, the sample set, and the sparse coefficient matrix respectively. Use Γ to obtain the weights of each candidate region, and then obtain the motion model p(x t |x t-1 );

[0129] Step 3: Solve the sparse vector representation problem through BLAS-2 sparse representation, specifically as shown in formula (6):

[0130]

[0131] where φ consists of positive and negative templates, s and p are the sparse coefficient vector and the attributes corresponding to the template set (+1 for positive template, -1 for negative template) respectively. Use s to generate the projection diagonal matrix S (the vector elements are 0 corresponding to the zero matrix, and non-zero corresponding to the identity matrix).

[0132] Step 4: Project X and φ into X' and φ' respectively through the projection matrix S;

[0133] Step 5: Solve the sparse matrix representation problem through the fast iterative shrinkage thresholding algorithm based on BLAS-3, specifically as shown in formula (7):

[0134]

[0135] where A represents the second sparse matrix.

[0136] Step 6: Calculate the confidence H of the candidate region by formula (8) c :

[0137]

[0138] where σ represents a positive constant.

[0139] Step 7: Use the K-mean algorithm to find several clustering centers of the initial target template sliding subgraph, and then combine them into the dictionary matrix D.

[0140] Step 8: Solve the sparse matrix representation problem through the fast iterative threshold shrinkage algorithm based on BLAS-3 using formula (9):

[0141]

[0142] Among them, Y and B respectively obtain the sample matrix and the sparse coefficient matrix of multiple standard small graphs through a sliding window, and D represents the dictionary matrix.

[0143] Step 10: Calculate the occlusion matrix O, and its elements can be calculated by formula (10) for the elements of the occlusion matrix O:

[0144]

[0145] Among them, ε0 is a preset threshold, and ε ij represents the elements in the reconstruction error matrix E.

[0146] Specifically, after calculating all the occlusion elements, determine whether the sum of all the occlusion elements is greater than the preset threshold. If so, the template histogram is the weighted sum of the first-frame initial target histogram and the current histogram; otherwise, the template histogram is the first-frame initial target histogram.

[0147] The reconstruction error matrix E is shown in formula (16):

[0148]

[0149] Calculate the anti-occlusion histogram using formula (11):

[0150]

[0151] Among them, q i (i = 1, 2,..., M) is the column vector in the matrix Q, represents the multiplication of matrix elements.

[0152] Step 11: Calculate the similarity between the template and the candidate target anti-occlusion histogram using formula (12):

[0153]

[0154] Among them, The jth standard small graph and Ψ j is the template histogram, and the template histogram Ψ j is updated using formula (13)

[0155] Ψ j = μψ f +(1 - μ)ψ l If Q n < a formula (13)

[0156] Among them, ψ f and ψ l are the template histograms of the first frame and the current frame respectively, a is a preset threshold, and Q n is the occlusion matrix of the current frame.

[0157] Step Twelve: Calculate the collaborative confidence of the candidate regions And take the one with the maximum value as the tracking target of the next frame. And determine whether the tracking ends. If it does not end, return to Step One to execute. Otherwise, end the tracking process.

[0158] In some embodiments, by comparing the running time of the fast threshold shrinkage algorithm of BLAS-3 involved in the target tracking method of the embodiments of the present application with other sparse representation algorithms, the Figure 6 shown curve graph can be obtained. From Figure 6 it can be seen that with the increase of the sparse level, the running time of the fast threshold shrinkage algorithm of BLAS-3 in the embodiments of the present application does not increase. Therefore, the fast threshold shrinkage algorithm of BLAS-3 in the embodiments of the present application can improve the computational efficiency of robust target tracking based on sparse representation without losing the accuracy and robustness of the target tracking method, thereby improving the frame rate of target tracking.

[0159] The embodiments of the present application disclose a target tracking system, including:

[0160] A first module, configured to sample the target region of the next frame according to the target tracking result of the previous frame or the initially selected target, and obtain a plurality of candidate region samples as a sample set;

[0161] A second module, configured to calculate the first confidence of each candidate target in the sample set based on the sparse discriminant classification of the preset iterative shrinkage algorithm, and the preset iterative shrinkage algorithm includes three-level basic algebraic subroutines;

[0162] A third module, configured to calculate the similarity of the anti-occlusion histogram of each candidate target in the sample set based on the sparse generation model of the preset iterative shrinkage algorithm;

[0163] A fourth module, configured to calculate the collaborative confidence of each candidate target according to the first confidence and the similarity;

[0164] A fifth module, configured to select the candidate target corresponding to the collaborative confidence that meets the preset requirements as the tracking target of the next frame.

[0165] The content of the above method embodiments is applicable to the system embodiments of the present application, and the effects achieved by the system embodiments of the present application are the same as those of the above method embodiments.

[0166] The embodiments of the present application disclose a terminal device, including:

[0167] A memory in which a computer program is stored;

[0168] A processor, when the computer program is executed by the processor, enables the processor to implement the above-mentioned various target tracking methods.

[0169] Figure 7 The figure shows a schematic structural diagram of a terminal device disclosed in an embodiment of the present application. As Figure 7 shown, the terminal device 500 of this embodiment includes: a memory 510 and a processor 520, and a computer program 511 is stored in the memory 510; when the computer program 511 is executed by the processor 520, the processor 520 is enabled to implement the steps of any of the above-mentioned methods.

[0170] The terminal device 500 may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The terminal device may include, but is not limited to, a processor 520 and a memory 510. Those skilled in the art can understand that Figure 7 merely examples of the terminal device 500, and do not constitute a limitation on the terminal device 500, and may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, it may also include input / output devices, network access devices, etc.

[0171] The so-called processor 520 may be a central processing unit (CPU), and the processor 520 may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0172] The memory 510 may be an internal storage unit of the terminal device 500 in some embodiments, such as the hard disk or memory of the terminal device 500. The memory 510 may also be an external storage device of the terminal device 500 in some other embodiments, such as a plug-in hard disk equipped on the terminal device 500, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. Further, the memory 510 may also include both the internal storage unit and the external storage device of the terminal device 500. The memory 510 is used to store an operating system, application programs, a BootLoader, data, and other programs, such as the program code of the computer program. The memory 510 may also be used to temporarily store data that has been output or will be output.

[0173] An embodiment of the present application also discloses a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned method embodiments can be implemented.

[0174] An embodiment of the present application discloses a computer program product, and when the computer program product runs on a mobile terminal, the mobile terminal can implement the steps in the above-mentioned method embodiments when executed.

[0175] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-mentioned method embodiments of the present application, a computer program may be used to instruct relevant hardware to complete. The computer program may be stored in a computer-readable storage medium, and when the computer program is executed by a processor, the steps in the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code may be in the form of source code, object code, an executable file, or some intermediate form, etc. The computer-readable medium may at least include: any entity or device capable of carrying the computer program code to the photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk, or an optical disc, etc. In some jurisdictions, according to legislation and patent practice, the computer-readable medium may not be an electrical carrier signal and a telecommunication signal.

[0176] In the above embodiments, the descriptions of the various embodiments each have their own focuses. For parts not detailed or recorded in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0177] Those of ordinary skill in the art will realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0178] In the embodiments disclosed in this application, it should be understood that the disclosed device / terminal device and method can be implemented in other ways. For example, the device / terminal device embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the device or unit can be in electrical, mechanical or other forms.

[0179] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0180] The above embodiments are only used to illustrate the technical solutions of this application, not to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. And these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included in the protection scope of this application.

Claims

1. A target tracking method, characterized in that, Including: Sampling the target area of the next frame according to the target tracking result of the previous frame or the initially selected target to obtain a number of candidate region samples as a sample set; Based on the sparse discriminant classification of the preset iterative shrinkage algorithm, calculating the first confidence of each candidate target in the sample set, and the preset iterative shrinkage algorithm includes three-level basic algebraic subroutines; Based on the sparse generation model of the preset iterative shrinkage algorithm, calculating the similarity of the anti-occlusion histogram of each candidate target in the sample set; Calculating the collaborative confidence of each candidate target according to the first confidence and the similarity; Selecting the candidate target corresponding to the collaborative confidence that meets the preset requirements as the tracking target of the next frame; Wherein, the preset iterative shrinkage algorithm includes: Calculating the basic iteration; Calculating the first sparse matrix representation or the first non-negative sparse matrix representation according to the basic iteration; Calculating the iteration interval; Calculating the Taylor station expansion point according to the iteration interval and the first sparse matrix representation, or calculating the Taylor station expansion point according to the iteration interval and the first non-negative sparse matrix representation; When the iteration error meets the error condition, obtaining the solution of the first sparse matrix representation or obtaining the solution of the first non-negative sparse matrix representation; Wherein, the sparse discriminant classification based on the preset iterative shrinkage algorithm, calculating the first confidence of each candidate target in the sample set, includes: Obtaining a projection diagonal matrix; Projecting the template matrix and the candidate target matrix into the dynamic discrimination space through the diagonal matrix to obtain a projected template matrix and a projected candidate target matrix, the candidate target matrix corresponding to the sample set, and the template matrix corresponding to the template set; Calculating the first reconstruction error of the second sparse matrix representation of the candidate target according to the projected template matrix, the projected candidate target matrix and the preset iterative shrinkage algorithm; Calculating the first confidence of each candidate target according to the first reconstruction error.

2. The target tracking method according to claim 1, wherein, The sparse generation model based on the preset iterative shrinkage algorithm, calculating the similarity of the anti-occlusion histogram of each candidate target in the sample set, includes: Obtaining a dictionary matrix; Obtaining the sample matrix of multiple standard sliding window small images in the sample set; Calculating the second non-negative sparse matrix representation of the standard sliding window small image according to the dictionary matrix and the sample matrix; Obtaining the anti-occlusion histogram of the standard sliding window small image; Calculating the similarity of the anti-occlusion histogram between the template and the standard sliding window small image.

3. The target tracking method according to claim 1, wherein The obtaining the projection diagonal matrix includes: Obtaining a sparse coefficient vector; Generating a projection diagonal matrix according to the elements of the sparse coefficient vector.

4. The target tracking method according to claim 2, wherein The obtaining the dictionary matrix includes: Calculating several cluster centers of the initial standard sliding window small image; Combining all cluster centers into a dictionary matrix.

5. The target tracking method according to claim 2, wherein The obtaining the anti-occlusion histogram of the standard sliding window small image includes: Calculating a reconstruction error matrix according to the dictionary matrix, the sample matrix and the sparse coefficient matrix; Calculating an occlusion matrix according to the elements in the reconstruction error matrix; Calculating the anti-occlusion histogram of the standard sliding window small image according to the occlusion matrix and the sparse coefficient matrix.

6. A target tracking system, characterized in that, Including: The first module is used to sample the target area of the next frame according to the target tracking result of the previous frame or the initially selected target, and obtain a number of candidate region samples as a sample set; The second module is used to calculate the first confidence of each candidate target in the sample set based on the sparse discriminant classification of the preset iterative shrinkage algorithm, and the preset iterative shrinkage algorithm includes three-level basic algebraic subroutines; The third module is used to calculate the similarity of the anti-occlusion histogram of each candidate target in the sample set based on the sparse generation model of the preset iterative shrinkage algorithm; The fourth module is used to calculate the collaborative confidence of each candidate target according to the first confidence and the similarity; The fifth module is used to select the candidate target corresponding to the collaborative confidence that meets the preset requirements as the tracking target of the next frame; Wherein, the preset iterative shrinkage algorithm includes: Calculating the basic iteration; Calculating the first sparse matrix representation or the first non-negative sparse matrix representation according to the basic iteration; Calculating the iteration interval; Calculating the Taylor station expansion point according to the iteration interval and the first sparse matrix representation, or calculating the Taylor station expansion point according to the iteration interval and the first non-negative sparse matrix representation; When the iteration error meets the error condition, obtaining the solution of the first sparse matrix representation or obtaining the solution of the first non-negative sparse matrix representation; Wherein, the sparse discriminant classification based on the preset iterative shrinkage algorithm, calculating the first confidence of each candidate target in the sample set, includes: Obtaining a projection diagonal matrix; Projecting the template matrix and the candidate target matrix into the dynamic discrimination space through the diagonal matrix to obtain a projected template matrix and a projected candidate target matrix, the candidate target matrix corresponding to the sample set, and the template matrix corresponding to the template set; Calculating the first reconstruction error of the second sparse matrix representation of the candidate target according to the projected template matrix, the projected candidate target matrix and the preset iterative shrinkage algorithm; Calculating the first confidence of each candidate target according to the first reconstruction error.

7. A terminal device, characterized in that, Including: A memory in which a computer program is stored; A processor, when the computer program is executed by the processor, causes the processor to implement the target tracking method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the target tracking method according to any one of claims 1 to 5.

Images