Shielding attitude estimation method, system and equipment based on structured modeling and medium

Abstract

The invention discloses an occlusion attitude estimation method, system and device based on structured modeling and a medium, and the method comprises the following steps: obtaining an input image, and carrying out window division processing based on irregular expansion on the input image to obtain a divided image; sequentially performing down-sampling processing and feature classification quantization processing on the divided image based on a four-layer four-edge attention mechanism to obtain processed features; inputting the processed features into a decoder based on a multi-layer perceptron hybrid network for image attitude recovery; window expansion can be realized by adopting projection conversion, information loss caused by a square window is reduced, and the object information capturing capability is enhanced; a quadrilateral attention mechanism is adopted to perform down-sampling, feature extraction classification and quantification on the image, and huge calculation amount caused by window sliding is reduced; the multi-layer perceptron hybrid network is used as a decoder for attitude estimation and picture recovery, and the attitude estimation accuracy under the shielding condition is improved.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology. Specifically, the present invention is applied to the field of human body posture estimation, and in particular to an occlusion posture estimation method, system, device and medium based on structured modeling. Background Art

[0002] At present, human posture estimation solutions are widely used in motion recognition, posture tracking, security and monitoring and other fields. Human posture estimation can accurately estimate the position of human joints from images.

[0003] Most existing solutions use deep learning technology to achieve human posture estimation. However, in real scenes, the human body is often occluded by objects or crowds, and deep learning technology is difficult to solve this occlusion problem.

[0004] Specifically, the existing scheme is as follows:

[0005] (I) Single joint recognition scheme: This method performs independent recognition prediction for each joint and predicts the heat map of each joint separately, which has great limitations in occlusion scenarios.

[0006] (ii) Non-single-joint recognition schemes, such as global optimization scheme, human key point grouping recognition scheme, prediction recognition scheme based on non-occluded parts, and occlusion prediction scheme based on CPN. Among the above schemes, although the simplified scheme has better detection efficiency, the accuracy of posture estimation in occluded scenes is low; the precise scheme requires the design of complex model algorithms, with high computational complexity and large computing power requirements.

[0007] In summary, existing posture estimation methods cannot achieve fast and accurate recognition of invisible joints, and the overall performance of the scheme has certain deficiencies and limitations. Summary of the invention

[0008] The object of the present invention is to provide a method, system, device and medium for occlusion posture estimation based on structured modeling, so as to solve all or one of the above problems existing in the prior art.

[0009] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0010] In one aspect, the present invention provides a method for estimating an occlusion posture based on structured modeling, comprising the following steps:

[0011] Image windowing step: obtaining an input image, performing a window division process based on irregular expansion on the input image, and obtaining a divided image;

[0012] Feature extraction step: performing downsampling processing and feature classification and quantization processing on the divided image in sequence based on a four-layer four-edge attention mechanism to obtain processed features;

[0013] Posture recovery step: input the processed features into a decoder based on a multi-layer perceptron hybrid network to restore the image posture.

[0014] In one implementation manner, the window division process based on irregular expansion includes:

[0015] Dividing the input image into square windows;

[0016] Based on a projection transformation algorithm, the divided square windows are quadrilaterally expanded without restrictions on size, direction and shape to obtain an image covered by multiple irregular windows;

[0017] The image covered by the plurality of irregular windows is used as the divided image.

[0018] In one embodiment, each layer of the attention mechanism module of the four-layer four-edge attention mechanism includes a conditional position embedding CPE layer;

[0019] The downsampling channel size of each attention mechanism module of the four-layer four-edge attention mechanism matches the downsampling channel size of the Swintransformer.

[0020] In one embodiment, the downsampling process includes:

[0021] Using each layer of the attention mechanism module of the four-layer four-edge attention mechanism to downsample the divided image in steps of two, one, one, and one respectively;

[0022] When performing the downsampling, the number of channels of each downsampling layer is set to be twice the number of channels of the previous downsampling layer.

[0023] In one embodiment, the feature classification and quantification process includes:

[0024] Identify the pose of objects in the downsampled image;

[0025] Create a number of discrete symbols corresponding to the pose of each object in the downsampled image;

[0026] Constructing a quantitative feature set for representing structured modeling of the human body based on a number of the discrete symbols;

[0027] The quantitative feature set is used as the processed feature.

[0028] In one embodiment, the decoder based on the multi-layer perceptron hybrid network includes:

[0029] Fully patched fully connected layers, mixing layers, and classifiers;

[0030] The mixed layer is composed of several layers of the same size, each layer is composed of two multi-layer perceptron modules and a global pooling layer; the two multi-layer perceptron modules are respectively a small picture mixed MLP block and a channel mixed MLP block;

[0031] When the decoder based on the multi-layer perceptron hybrid network performs image posture recovery, the spatial position information of several small images is fused through the small image mixing MLP block, and the information of several channels is combined through the channel mixing MLP block.

[0032] In one embodiment, the step of restoring the image posture includes:

[0033] Using the full-patch fully connected layer to divide the downsampled image into a plurality of small images for mapping according to the processed features;

[0034] Using the multi-layer structure in the mixed layer to process the plurality of small pictures respectively;

[0035] The output result of the mixed layer is passed through the classifier to obtain the image posture restoration result.

[0036] On the other hand, the present invention also provides an occlusion posture estimation system based on structured modeling, comprising:

[0037] The image windowing module is used to: obtain an input image, perform a window division process based on irregular expansion on the input image, and obtain a divided image;

[0038] A feature extraction module is used to: perform downsampling processing and feature classification and quantization processing on the divided images in sequence based on a four-layer four-edge attention mechanism to obtain processed features;

[0039] The posture recovery module is used to: input the processed features into a decoder based on a multi-layer perceptron hybrid network to restore the image posture.

[0040] On the other hand, the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the occlusion posture estimation method based on structured modeling are implemented.

[0041] On the other hand, the present invention further provides a computer device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; wherein:

[0042] The memory is used to store computer programs;

[0043] The processor is used to execute the steps of the occlusion posture estimation method based on structured modeling by running the program stored in the memory.

[0044] The beneficial effects of the technical solution of the present invention are:

[0045] 1. The occlusion posture estimation method based on structured modeling described in the present invention can realize window expansion by using projection conversion, reduce information loss caused by square windows, and enhance the ability to capture object information; use quadrilateral attention to downsample, classify and quantize images for feature extraction, improve information extraction capabilities, and reduce the huge amount of calculations caused by window sliding; use a multi-layer perceptron hybrid network as a decoder to perform posture estimation and image restoration, improve the accuracy of posture estimation under occlusion, reduce the complexity of the model and the difficulty of training, and improve the recognition efficiency of the model and the overall performance of the model.

[0046] 2. The occlusion posture estimation system based on structured modeling described in the present invention can realize the occlusion posture estimation method based on structured modeling described in the present invention through the mutual cooperation of system modules.

[0047] 3. The computer-readable storage medium described in the present invention can realize the cooperation of the guidance system modules, thereby realizing the occlusion posture estimation method based on structured modeling described in the present invention, and the computer-readable storage medium described in the present invention also effectively improves the operability of the occlusion posture estimation method based on structured modeling.

[0048] 4. The computer device described in the present invention can store and execute the computer-readable storage medium, thereby realizing the occlusion posture estimation method based on structured modeling described in the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 is a flow chart of the occlusion posture estimation method based on structured modeling described in Example 1 of the present invention;

[0051] Figure 2It is a logical flow diagram of the occlusion posture estimation method based on structured modeling described in Example 1 of the present invention;

[0052] Figure 3 It is a schematic diagram of window expansion of the occlusion posture estimation method based on structured modeling described in Example 1 of the present invention;

[0053] Figure 4 It is a schematic diagram of the downsampling process of the occlusion posture estimation method based on structured modeling described in Example 1 of the present invention;

[0054] Figure 5 Schematic diagram of the decoder architecture of the occlusion posture estimation method based on structured modeling described in Example 1 of the present invention;

[0055] Figure 6 Schematic diagram of a hybrid layer architecture of the occlusion posture estimation method based on structured modeling described in Example 1 of the present invention;

[0056] Figure 7 Schematic diagram of the architecture of the occlusion posture estimation system based on structured modeling according to Embodiment 2 of the present invention;

[0057] Figure 8 is a schematic diagram of the structure of the computer device described in Example 4 of the present invention;

[0058] The symbols in the accompanying drawings are explained as follows:

[0059] 1501, processor; 1502, communication interface; 1503, memory; 1504, communication bus. DETAILED DESCRIPTION

[0060] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0061] In the description of the present invention, it should be noted that the embodiments described in the present invention are only part of the embodiments of the present invention, rather than all of the embodiments; based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present invention.

[0062] The terms "first", "second", etc. in the specification and claims of this article and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of this article described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0063] CPE (conditional position embedding) is conditional position embedding.

[0064] Example 1

[0065] This embodiment provides an occlusion posture estimation method based on structured modeling, such as Figure 1 to Figure 6 As shown, the following steps are included:

[0066] S100, divide and expand the image window:

[0067] In this step, the input image is divided into square windows, and the divided windows are expanded based on the projection transformation so that each window covers and captures different contents in the input image. Finally, multiple quadrilaterals with unlimited size, direction and shape are obtained in the input image, and overlap is allowed between multiple quadrilaterals. The overlapping part can also promote cross-window information exchange, such as Figure 3 As shown;

[0068] Specifically, since the projection transformation operation does not retain the parallelism and direction of the original window, the position and shape of the new window are deformable. The projection transformation can be represented by the following 3×3 transformation matrix:

[0069]

[0070] Among them: x = (x1, x2, x3, x4), used to define the deformation of scaling and rotation; y = (y1, y2), used to define translation; z = (z1, z2) is the projection vector, used to determine the movement and deformation of the projected object when the observer changes the perspective in different dimensions;

[0071] Specifically, the expansion of the window also has a certain probability of generating a window covering the area outside the feature map. Therefore, a regularization term is set to reward the projected quadrilateral covering the valid area of ​​the feature map, and a penalty function is set to penalize the coordinates covering outside the feature map, as follows:

[0072]

[0073] Specifically, a zero vector is used as the sampling value of the coordinate outside the image feature map; the regularization loss of the coordinate is as follows: L reg =∑ i R(x q,i )·x q,i +R(y q,i )·y q,i .

[0074] S200, downsampling and feature extraction processing:

[0075] In this step, the image obtained in step S100 is downsampled by a four-layer quadrilateral attention mechanism module, each downsampling layer is composed of a quadrilateral attention mechanism, a standard layer, a multi-layer perceptron and a conditional position embedding CPE; then, the quadrilateral attention mechanism is used to perform feature classification, extraction and quantization on the downsampled image;

[0076] Specifically, the downsampling operation includes:

[0077] (1) Each layer of the attention mechanism module downsamples with a step size of 2, 1, 1, and 1 (i.e., a sampling factor of 4x, 2x, 2x, and 2x, respectively);

[0078] (2) When downsampling, the number of channels in each downsampled layer becomes twice the number of channels in the previous layer (the number of channels is C, 2C, 4C and 8C, respectively), thereby improving the basic representation ability, improving the model's ability to fit complex functions, improving the ability to process advanced features, and expanding the scope of application;

[0079] (3) In each layer of the attention mechanism module, the conditional position embedding (CPE) of CPVT is added to encode the spatial information into the model, thereby fully retaining the spatial position information of the objects in the image;

[0080] (4) The downsampling channel size of each layer of the attention mechanism module is set to be the same as the downsampling channel size of the Swin transformer, but the window sliding shift step is omitted to improve the simplicity and convenience of the model;

[0081] (5) The quadrilateral attention mechanism is adopted to reduce the additional computation caused by the traditional multi-head self-attention mechanism.

[0082] Specifically, feature extraction, classification and quantification processing include:

[0083] (1) Feature extraction: Identify the posture of objects in the downsampled image;

[0084] (2) Labeling, classification and quantization: The posture of each object in the downsampled image is represented by M discrete symbols, i.e., G∈RK×D; each discrete symbol represents a substructure with multiple interdependent joints, and the structured modeling of the human body is realized based on multiple substructures, i.e., T=(t 1 ,t 2 ,...t m )(i.e., the quantitative feature set); where: K is the number of body joints, D is the joint size; t i For each marked feature symbol, t i ∈R″,t i It is approximated as a substructure of a certain posture, and a substructure contains several interdependent joints.

[0085] S300, image restoration processing:

[0086] In this step, MLP-Mixer (multi-layer perceptron hybrid network) is used as the decoder (the decoder architecture is as follows Figure 5 As shown in the figure, the marked classified and quantized features are used as input data to restore the final posture of the image; among them, MLP-Mixer can better realize parallel computing and improve the training speed. At the same time, it can retain the spatial information of objects in the image to a greater extent through spatial position encoding, maintain the invariance of image information, and reduce the loss of image information;

[0087] Specifically, the decoder consists of a Per-patch Fully-connected layer, a Mixer layer, and a classifier (global pooling layer + fully connected layer);

[0088] Among them, Figure 6 As shown in the figure, Mixer consists of multiple layers of the same size, each layer consists of 2 MLP (multi-layer perceptron) modules and 1 global pooling layer; the 2 MLPs are small picture mixing MLP blocks and channel mixing MLP blocks, each MLP block consists of 2 fully connected layers and 1 activation function; the small picture mixing MLP block can fuse the spatial position information of different small pictures to realize automatic learning and representation of position information; the channel mixing MLP block can combine the information of different channels (or different feature maps);

[0089] Specifically, the image restoration operation includes:

[0090] (1) Per-patch Fully-connected is used to divide the input image after the downsampling operation into multiple small images for mapping, and finally converted into a matrix of X = S × C;

[0091] (2) Input multiple small images into Mixer, and use the multi-layer structure in Mixer to process the multiple small images separately;

[0092] (3) After being processed by the Mixer’s mixing layer, the output result is passed through the final classifier to restore the image’s posture.

[0093] In summary, this method uses projection conversion to achieve window expansion, enhance the ability to capture object information, and reduce information loss; uses quadrilateral attention to downsample, extract features, classify and quantize images, improve information extraction capabilities, and reduce the huge amount of calculations caused by window sliding; uses a multi-layer perceptron hybrid network as a decoder for posture estimation and image restoration, improves the accuracy of posture estimation under occlusion, and improves recognition efficiency and the overall performance of the model.

[0094] It should be noted that all the above examples are only for explaining the present invention and cannot limit the protection scope of the present invention.

[0095] Example 2

[0096] This embodiment is based on the same inventive concept as the occlusion posture estimation method based on structured modeling described in Embodiment 1, and provides an occlusion posture estimation system based on structured modeling, such as Figure 7 As shown, including:

[0097] The image windowing module is used to: obtain an input image, perform a window division process based on irregular expansion on the input image, and obtain a divided image;

[0098] A feature extraction module is used to: perform downsampling processing and feature classification and quantization processing on the divided images in sequence based on a four-layer four-edge attention mechanism to obtain processed features;

[0099] The posture recovery module is used to: input the processed features into a decoder based on a multi-layer perceptron hybrid network to restore the image posture.

[0100] Example 3

[0101] This embodiment provides a computer-readable storage medium, including:

[0102] The storage medium is used to store computer software instructions used to implement the occlusion posture estimation method based on structured modeling described in the above-mentioned Example 1, which includes a program for executing the above-mentioned program set for the occlusion posture estimation method based on structured modeling; specifically, the executable program can be built into the occlusion posture estimation system based on structured modeling described in Example 2, so that the occlusion posture estimation system based on structured modeling can implement the occlusion posture estimation method based on structured modeling described in Example 1 by executing the built-in executable program.

[0103] In addition, the computer-readable storage medium of this embodiment may adopt any combination of one or more computer-readable storage media, wherein the computer-readable storage medium includes electrical, optical, electromagnetic, infrared or semiconductor systems, devices or components, or any combination thereof.

[0104] Example 4

[0105] This embodiment provides an electronic device, such as Figure 8 As shown, the electronic device may include: a processor 1501 , a communication interface 1502 , a memory 1503 and a communication bus 1504 , wherein the processor 1501 , the communication interface 1502 , and the memory 1503 communicate with each other via the communication bus 1504 .

[0106] Memory 1503, used for storing computer programs;

[0107] The processor 1501 is used to implement the steps of the occlusion posture estimation method based on structured modeling described in the above-mentioned embodiment 1 when executing the computer program stored in the memory 1503.

[0108] As an embodiment of the present invention, the communication bus mentioned in the above terminal can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0109] As an implementation mode of the present invention, the communication interface is used for communication between the above-mentioned terminal and other devices.

[0110] As an embodiment of the present invention, the memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0111] As an embodiment of the present invention, the above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0112] Different from the prior art, the present application adopts an occlusion posture estimation method, system, device and medium based on structured modeling, which can use projection conversion to achieve window expansion, reduce information loss caused by square windows, and enhance the ability to capture object information; use quadrilateral attention to downsample, feature extract, classify and quantify images, improve information extraction capabilities, and reduce the huge amount of calculations caused by window sliding; use a multi-layer perceptron hybrid network as a decoder to perform posture estimation and image restoration, improve the accuracy of posture estimation under occlusion, reduce the complexity of the model and the difficulty of training, and improve the recognition efficiency of the model and the overall performance of the model.

[0113] It should be understood that in the various embodiments of this document, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this document.

[0114] It should also be understood that in the embodiments of this article, the term "and / or" is only a description of the association relationship of the associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0115] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this article.

[0116] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0117] In the several embodiments provided herein, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or can be electrical, mechanical or other forms of connection.

[0118] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of this article.

[0119] In addition, each functional unit in each embodiment of this invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional unit.

[0120] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this article is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of this article. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0121] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for estimating occlusion posture based on structured modeling, characterized in that: The following steps are involved: Image windowing step: obtaining an input image, performing a window division process based on irregular expansion on the input image, and obtaining a divided image; Feature extraction step: performing downsampling processing and feature classification and quantization processing on the divided image in sequence based on a four-layer four-edge attention mechanism to obtain processed features; Posture recovery step: input the processed features into a decoder based on a multi-layer perceptron hybrid network to restore the image posture.

2. The occlusion posture estimation method based on structured modeling according to claim 1, characterized in that: The window division process based on irregular expansion includes: Dividing the input image into square windows; Based on a projection transformation algorithm, the divided square windows are quadrilaterally expanded without restrictions on size, direction and shape to obtain an image covered by multiple irregular windows; The image covered by the plurality of irregular windows is used as the divided image.

3. The occlusion posture estimation method based on structured modeling according to claim 1, characterized in that: Each attention mechanism module of the four-layer four-edge attention mechanism includes a conditional position embedding CP E layer; The downsampling channel size of each attention mechanism module of the four-layer four-edge attention mechanism matches the downsampling channel size of the Swintransformer.

4. The occlusion posture estimation method based on structured modeling according to claim 1, characterized in that: The downsampling process comprises: Using each layer of the attention mechanism module of the four-layer four-edge attention mechanism to downsample the divided image in steps of two, one, one, and one respectively; When performing the downsampling, the number of channels of each downsampling layer is set to be twice the number of channels of the previous downsampling layer.

5. The occlusion posture estimation method based on structured modeling according to claim 1, characterized in that: The feature classification and quantification processing includes: Identify the pose of objects in the downsampled image; Create a number of discrete symbols corresponding to the pose of each object in the downsampled image; Constructing a quantitative feature set for representing structured modeling of the human body based on a number of the discrete symbols; The quantitative feature set is used as the processed feature.

6. The occlusion posture estimation method based on structured modeling according to claim 1, characterized in that: The decoder based on the multi-layer perceptron hybrid network includes: Fully patched fully connected layers, mixing layers, and classifiers; The mixed layer is composed of several layers of the same size, each layer is composed of two multi-layer perceptron modules and a global pooling layer; the two multi-layer perceptron modules are respectively a small picture mixed MLP block and a channel mixed MLP block; When the decoder based on the multi-layer perceptron hybrid network performs image posture recovery, the spatial position information of several small images is fused through the small image mixing MLP block, and the information of several channels is combined through the channel mixing MLP block.

7. The method for estimating occlusion posture based on structured modeling according to claim 6, characterized in that: The step of restoring the image posture comprises: Using the full-patch fully connected layer to divide the downsampled image into a plurality of small images for mapping according to the processed features; Using the multi-layer structure in the mixed layer to process the plurality of small pictures respectively; The output result of the mixed layer is passed through the classifier to obtain the image posture restoration result.

8. An occlusion posture estimation system based on structured modeling, characterized in that: include: The image windowing module is used to: obtain an input image, perform a window division process based on irregular expansion on the input image, and obtain a divided image; A feature extraction module is used to: perform downsampling processing and feature classification and quantization processing on the divided images in sequence based on a four-layer four-edge attention mechanism to obtain processed features; The posture recovery module is used to: input the processed features into a decoder based on a multi-layer perceptron hybrid network to restore the image posture.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the occlusion posture estimation method based on structured modeling in any one of claims 1 to 7 are implemented.

10. A computer device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; wherein: The memory is used to store computer programs; The processor is used to execute the steps of the occlusion posture estimation method based on structured modeling in any one of claims 1 to 7 by running the program stored in the memory.

Images