Human three-dimensional model construction method and human feature information acquisition method

Abstract

The present application relates to the technical field of information collection, and particularly relates to a human three-dimensional model construction method and a human feature information acquisition method. The human three-dimensional model construction method comprises the following steps: obtaining single-view human point cloud data of a to-be-tested person, inputting the single-view human point cloud data into a trained human model construction network, and obtaining a three-dimensional human model of the to-be-tested person; the human model construction network is used for deforming three-dimensional human template data according to the single-view human point cloud data and a pre-obtained corresponding relationship between the single-view human point cloud data and the three-dimensional human template data, and obtaining the three-dimensional human model of the to-be-tested person. According to the three-dimensional model construction method, the three-dimensional human model can be constructed based on the single-view human point cloud data, and the feature information on the human body can be further collected based on the three-dimensional human model.

Description

Technical Field

[0001] This invention relates to the field of information acquisition technology, and in particular to a method for constructing a three-dimensional human body model and a method for acquiring human body feature information. Background Technology

[0002] Currently, the construction of 3D human body models is based on data collected by 3D human body scanning equipment. However, existing 3D human body scanning equipment is expensive and not yet widely used in most fields requiring human body model information acquisition. Most human body information acquisition fields currently use single-view human body information acquisition equipment, such as RGB-D monocular cameras, to acquire single-view human body information. Based on the acquired single-view human body information, a complete 3D human body model cannot be constructed, thus making it impossible to accurately obtain positional information on certain parts of the 3D human body model. Summary of the Invention

[0003] This invention provides a method for constructing a three-dimensional human body model and a method for acquiring human body feature information, in order to solve the technical problem that existing technologies cannot accurately acquire human body feature information based on single-view human body information.

[0004] On one hand, the present invention provides a method for constructing a three-dimensional human body model, comprising:

[0005] Acquire single-view human point cloud data of the subject;

[0006] The single-view human point cloud data is input into the trained human model construction network to obtain a three-dimensional human model of the subject. The human model construction network is used to deform the three-dimensional human template data according to the single-view human point cloud data and the correspondence between the pre-acquired single-view human point cloud data and the three-dimensional human template data to obtain the three-dimensional human model of the subject.

[0007] According to the present invention, a method for constructing a three-dimensional human body model is provided, wherein the human body model construction network is trained and obtained through the following method:

[0008] Acquire multiple complete 3D human body point cloud data as complete 3D human body point cloud training data;

[0009] Deleting a portion of the data from the complete 3D human point cloud data yields the corresponding single-view human point cloud training data.

[0010] Set global feature constraint loss and reconstruction loss;

[0011] Based on the global feature constraint loss and reconstruction loss, the single-view human point cloud training data, the complete 3D human point cloud training data, and the 3D human template data are input into the human model construction network to train the human model construction network; so that the human model construction network can obtain the correspondence between the single-view human point cloud data and the 3D human template data.

[0012] The three-dimensional human body template data is obtained by storing standard three-dimensional human body data according to a preset storage format.

[0013] According to the present invention, a method for constructing a three-dimensional human body model is provided, wherein the human body model construction network includes an encoder and a decoder;

[0014] The encoder is used to extract features from the input complete human point cloud training data and the single-view human point cloud training data respectively to obtain the corresponding global features; the global feature constraint loss is used to represent the difference between the global features corresponding to the complete human point cloud training data and the global features corresponding to the single-view human point cloud training data output by the encoder. The smaller the global feature constraint loss, the closer the single-view human model features are to the complete human features.

[0015] The decoder is used to deform the three-dimensional human body template data in terms of posture and body shape according to the input global features.

[0016] According to the present invention, a method for constructing a three-dimensional human body model, after obtaining the three-dimensional human body model of the subject, further includes optimizing the obtained three-dimensional human body model of the subject by using a preset single-sided chamfer distance processing method.

[0017] On the other hand, the present invention also provides a method for obtaining human feature information, including:

[0018] Acquire single-view human point cloud data of the subject;

[0019] The single-view human point cloud data is input into the trained human model construction network to obtain a three-dimensional human model of the subject. The human model construction network is used to deform the three-dimensional human template data according to the single-view human point cloud data and the correspondence between the pre-acquired single-view human point cloud data and the three-dimensional human template data to obtain the three-dimensional human model of the subject.

[0020] The target feature information is obtained by extracting features from the target body parts on the three-dimensional human body model of the subject; the target feature information includes the position information of the target body parts.

[0021] According to the present invention, a method for constructing a three-dimensional human body model is provided, wherein the human body model construction network is trained and obtained through the following method:

[0022] Acquire multiple complete 3D human body point cloud data as complete 3D human body point cloud training data;

[0023] Deleting a portion of the data from the complete 3D human point cloud data yields the corresponding single-view human point cloud training data.

[0024] Set global feature constraint loss and reconstruction loss;

[0025] Based on the global feature constraint loss and reconstruction loss, the single-view human point cloud training data, the complete 3D human point cloud training data, and the 3D human template data are input into the human model construction network to train the human model construction network; so that the human model construction network can obtain the correspondence between the single-view human point cloud data and the 3D human template data.

[0026] The three-dimensional human body template data is obtained by storing standard three-dimensional human body data according to a preset storage format.

[0027] According to the present invention, a method for constructing a three-dimensional human body model, after obtaining the three-dimensional human body model of the subject, further includes optimizing the obtained three-dimensional human body model of the subject by using a preset single-sided chamfer distance processing method.

[0028] On the other hand, the present invention also provides a human feature information acquisition device, comprising:

[0029] The acquisition module is used to acquire single-view human point cloud data of the subject.

[0030] The model building module is used to input the single-view human point cloud data into the trained human model building network to obtain a three-dimensional human model of the test subject; the human model building network is used to deform the three-dimensional human template data according to the single-view human point cloud data and the correspondence between the pre-acquired single-view human point cloud data and the three-dimensional human template data to obtain the three-dimensional human model of the test subject.

[0031] The feature extraction module is used to extract features from the target human body parts on the three-dimensional human body model of the subject to obtain the target feature information; the target feature information includes the position information of the target human body parts.

[0032] On the other hand, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement any of the above-described methods for acquiring human feature information.

[0033] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the human feature information acquisition method as described above.

[0034] On the other hand, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the human feature information acquisition method as described above.

[0035] The present invention provides a method for constructing a 3D human body model. This method acquires single-view human point cloud data of the subject and inputs it into a trained human body model construction network to obtain a 3D human body model of the subject. The human body model construction network is used to deform the 3D human body template data based on the single-view human point cloud data and the correspondence between the pre-acquired single-view human point cloud data and 3D human body template data, thereby obtaining the 3D human body model of the subject. According to the 3D model construction method of the present invention, it is possible to construct a 3D human body model based on single-view human point cloud data, and further realize the acquisition of feature information on the human body based on the 3D human body model.

[0036] According to the present invention, a method for obtaining human feature information is provided, which obtains single-view human point cloud data of the subject, constructs a three-dimensional human model based on the single-view human point cloud data, and then extracts target feature information from the constructed three-dimensional human model based on the target part information on the three-dimensional human template data, so that the extracted target feature information is more accurate. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0038] Figure 1 This is a flowchart illustrating the method for constructing a three-dimensional human body model provided by the present invention;

[0039] Figure 2 This is a flowchart of the human body model construction network training method provided by the present invention;

[0040] Figure 3 This is one of the schematic diagrams illustrating the correspondence between single-view human point cloud and three-dimensional human point cloud provided by the present invention;

[0041] Figure 4This is the second schematic diagram illustrating the correspondence between single-view human point cloud and three-dimensional human point cloud provided by the present invention;

[0042] Figure 5 This is a flowchart illustrating the method for acquiring human feature information provided by the present invention;

[0043] Figure 6 This is a schematic diagram of the human feature information acquisition device provided by the present invention;

[0044] Figure 7 This is a schematic diagram of the physical structure of the electronic device provided by the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0046] The present invention provides a method for constructing a 3D human body model. A pre-trained human body model construction network is used to train a large amount of data to establish a correspondence between single-view human body point cloud data and 3D human body template data. Specifically, after acquiring the single-view human body point cloud data of the subject, the data is input into the human body model construction network. The network then deforms the 3D human body template data based on the single-view point cloud data and the pre-acquired correspondence between the single-view point cloud data and the 3D human body template data, obtaining a 3D human body model of the subject (this 3D human body model can be either point cloud data or mesh data, depending on the data format of the 3D human body template data). In other words, the search for single-view human body point clouds is transformed into a correspondence between the 3D human body template data and the constructed 3D human body model vertices. This allows for precise location of target feature points, as well as accurate location of nearby points.

[0047] Example 1:

[0048] This embodiment provides a method for constructing a three-dimensional human body model, such as Figure 1 The method for constructing a 3D human body model includes:

[0049] Step 101: Obtain single-view human point cloud data of the subject.

[0050] For example, in this embodiment, an RGB-D monocular camera is used to collect single-view human point cloud data.

[0051] Step 102: Input the single-view human point cloud data into the trained human model construction network to obtain the three-dimensional human model of the test subject; the human model construction network is used to deform the three-dimensional human template data according to the single-view human point cloud data and the correspondence between the pre-acquired single-view human point cloud data and the three-dimensional human template data to obtain the three-dimensional human model of the test subject.

[0052] Specifically, the human body model construction network in this embodiment includes an encoder and a decoder, and is trained and obtained through the following method:

[0053] Step 201: Acquire multiple complete 3D human point cloud data sets as complete 3D human point cloud training data. For example, use a 3D point cloud acquisition device to acquire complete 3D human point cloud data sets in multiple poses as complete 3D human point cloud training data.

[0054] Step 202: Delete part of the data in the complete 3D human point cloud data to obtain the corresponding single-view human point cloud training data.

[0055] In this embodiment, technicians can obtain single-view human point cloud training data by manually deleting the back data of the complete 3D human point cloud training data.

[0056] Step 203: Set global feature constraint loss and reconstruction loss.

[0057] Based on the global feature constraint loss and reconstruction loss, the single-view human point cloud training data, the complete human point cloud training data, and the preset 3D human template data are input into the human model construction network to train the human model construction network; so that the human model construction network can obtain the correspondence between the single-view human point cloud data and the 3D human template data.

[0058] Specifically:

[0059] The encoder is used to extract features from the input complete human point cloud training data and the single-view human point cloud training data to obtain the corresponding global features. The global feature constraint loss is used to represent the difference between the global features corresponding to the complete human point cloud training data and the global features corresponding to the single-view human point cloud training data output by the encoder. The smaller the global feature constraint loss, the closer the single-view human model features are to the complete human features, thus enabling the network to output a more accurate deformed template point cloud from the input single-view human point cloud.

[0060] The decoder is used to deform the 3D human template data in terms of posture and body shape according to the global features output by the encoder and output a 3D human model of the subject. The reconstruction loss is used to determine the reconstruction accuracy of the 3D human model of the subject output by the decoder relative to the complete human point cloud corresponding to the input single-view point cloud. For example, in this embodiment, the reconstruction loss is the point-to-point error between the vertices of the 3D human model output by the decoder and the complete human point cloud corresponding to the single view.

[0061] Among them, the global feature constraint loss and the reconstruction loss together constitute the basis for training completion.

[0062] The 3D human body template data is obtained by storing standard 3D human body data according to a preset storage format. Generally, 3D human body template data is stored using a parametric model; that is, the feature locations in the parametric model store the human body's feature information. For example, the 100th storage location in the model stores information about the left shoulder joint. Figure 3 and Figure 4 As shown, the correspondence obtained through training can be used to match single-view human point cloud with 3D human template data.

[0063] For example, training is considered complete when the training criteria, which consist of the global feature constraint loss and the reconstruction loss, are met.

[0064] In one embodiment, after obtaining the three-dimensional human body model of the subject, the method further includes optimizing the obtained three-dimensional human body model of the subject using a preset single-sided chamfer distance processing method. The correspondence between the single-view human body point cloud and the three-dimensional human body template data is transformed into the correspondence between two complete parameterized human body vertices, and further, the nearest neighbor points of all point clouds are found using a kd-tree (short for k-dimensional tree).

[0065] The 3D model construction method according to this embodiment can construct a 3D human body model based on single-view human body point cloud data, facilitating the acquisition of feature information on the human body based on the 3D human body model. Furthermore, the method described in this embodiment does not necessarily require a 3D human body scanning device when constructing a 3D human body model; the use of a single-view information acquisition device reduces equipment costs.

[0066] Example 2:

[0067] This embodiment provides a method for obtaining human feature information, such as Figure 5 The method for obtaining human feature information includes:

[0068] Step 301: Obtain single-view human point cloud data of the subject.

[0069] Step 302: Input the single-view human point cloud data into the trained human model construction network to obtain the three-dimensional human model of the test subject; the human model construction network is used to deform the three-dimensional human template data according to the single-view human point cloud data and the correspondence between the pre-acquired single-view human point cloud data and the three-dimensional human template data to obtain the three-dimensional human model of the test subject.

[0070] Step 303: Extract features from the target body parts on the three-dimensional human body model of the subject to obtain target feature information; the target feature information includes the position information of the target body parts.

[0071] The training method for the human model construction network in this embodiment is the same as that in Embodiment 1 above, and will not be described again in this embodiment.

[0072] In this embodiment, after obtaining the three-dimensional human body model of the subject, a preset single-sided chamfer distance processing method is used to optimize the obtained three-dimensional human body model of the subject. The specific optimization processing method is the same as that in Embodiment 1 above, and will not be repeated here.

[0073] According to the human feature information acquisition method of this embodiment, single-view human point cloud data of the subject is acquired. Based on the correspondence between the single-view human point cloud data and the three-dimensional human template data, the three-dimensional human template data is deformed to obtain the three-dimensional human model of the subject. Then, based on the position information of the target part of the human body, the target feature information is extracted from the three-dimensional human model. In this way, the target feature information (such as the position information of the target point cloud data) and the position information of the surrounding point cloud can be accurately extracted according to the correspondence.

[0074] Example 3:

[0075] like Figure 6 This embodiment provides a human feature information acquisition device, which includes: an acquisition module 601, a model construction module 602, and a feature extraction module 603.

[0076] The acquisition module 601 is used to acquire single-view human point cloud data of the subject.

[0077] The model building module 602 is used to input single-view human point cloud data into the trained human model building network to obtain a three-dimensional human model of the test subject. The human model building network is used to deform the three-dimensional human template data according to the single-view human point cloud data and the correspondence between the pre-acquired single-view human point cloud data and the three-dimensional human template data to obtain a three-dimensional human model of the test subject.

[0078] The feature extraction module 603 is used to extract features from the target human body parts on the three-dimensional human body model of the test subject to obtain target feature information; the target feature information includes the position information of the target human body parts.

[0079] The implementation methods of the functions of each module in this embodiment correspond to the methods in the above embodiments, and will not be repeated here.

[0080] The human feature information acquisition device of this embodiment can accurately acquire target feature information on a three-dimensional human body model based on a single-view human information acquisition device.

[0081] Figure 7 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 7 As shown, the electronic device may include a processor 710, a communication interface 720, a memory 730, and a communication bus 740, wherein the processor 710, the communication interface 720, and the memory 730 communicate with each other via the communication bus 740. The processor 710 can call logical instructions in the memory 730 to execute a method for acquiring human feature information. This method includes: acquiring single-viewpoint human point cloud data of the subject; inputting the single-viewpoint human point cloud data into a trained human model construction network to obtain a three-dimensional human model of the subject; the human model construction network is used to deform the three-dimensional human template data according to the single-viewpoint human point cloud data and the correspondence between the pre-acquired single-viewpoint human point cloud data and the three-dimensional human template data to obtain a three-dimensional human model of the subject; and extracting features from the target human parts on the three-dimensional human model of the subject to obtain target feature information; the target feature information includes the position information of the target human parts.

[0082] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0083] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the human feature information acquisition method provided by the above methods. The method includes: acquiring single-view human point cloud data of the subject; inputting the single-view human point cloud data into a trained human model construction network to obtain a three-dimensional human model of the subject; the human model construction network is used to deform the three-dimensional human template data according to the single-view human point cloud data and the correspondence between the pre-acquired single-view human point cloud data and the three-dimensional human template data to obtain a three-dimensional human model of the subject; extracting features from the target human parts on the three-dimensional human model of the subject to obtain target feature information; the target feature information includes the position information of the target human parts.

[0084] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements a method for acquiring human feature information provided by the above-described methods. The method includes: acquiring single-viewpoint human point cloud data of a subject; inputting the single-viewpoint human point cloud data into a trained human model construction network to obtain a three-dimensional human model of the subject; the human model construction network is used to deform the three-dimensional human template data according to the single-viewpoint human point cloud data and the correspondence between the pre-acquired single-viewpoint human point cloud data and the three-dimensional human template data to obtain a three-dimensional human model of the subject; and extracting features from the target human body parts on the three-dimensional human model of the subject to obtain target feature information; the target feature information includes the position information of the target human body parts.

[0085] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0086] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions 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 the present invention.

Claims

1. A method for constructing a three-dimensional human body model, characterized in that, include: Acquire single-view human point cloud data of the subject; The single-view human point cloud data is input into the trained human model construction network to obtain a three-dimensional human model of the subject. The human model construction network is used to deform the three-dimensional human template data in terms of posture and body shape according to the single-view human point cloud data and the correspondence between the pre-acquired single-view human point cloud data and the three-dimensional human template data to obtain the three-dimensional human model of the subject. The training process of the human body model construction network includes a global feature constraint loss and a reconstruction loss. The global feature constraint loss represents the difference between the global features corresponding to the complete human body point cloud training data output by the encoder in the human body model construction network and the global features corresponding to the single-view human body point cloud training data. The reconstruction loss is used to determine the point-to-point error between the vertices of the 3D human body model output by the decoder in the human body model construction network and the complete human body point cloud corresponding to the single-view human body point cloud data. The human body model construction network is trained using the following method: acquiring multiple complete human body point cloud data as complete human body point cloud training data; deleting partial regions of the complete human body point cloud data to obtain corresponding single-view human body point cloud training data; setting global feature constraint loss and reconstruction loss; and inputting the single-view human body point cloud training data, complete human body point cloud training data, and preset 3D human body template data into the human body model construction network based on the global feature constraint loss and reconstruction loss to train the human body model construction network; enabling the human body model construction network to obtain the correspondence between the single-view human body point cloud data and the 3D human body template data; the 3D human body template data is obtained by storing standard 3D human body data according to a preset storage format; The human body model construction network includes an encoder and a decoder. The encoder is used to extract features from the input complete human body point cloud training data and the single-view human body point cloud training data to obtain corresponding global features. The smaller the global feature constraint loss, the closer the single-view human body model features are to the complete human body features. The decoder is used to deform the three-dimensional human body template data in terms of posture and body shape according to the input global features.

2. The method for constructing a three-dimensional human body model according to claim 1, characterized in that, After obtaining the three-dimensional human body model of the subject, the method further includes optimizing the obtained three-dimensional human body model of the subject using a preset single-sided chamfer distance processing method.

3. A method for acquiring human body feature information, characterized in that, include: Acquire single-view human point cloud data of the subject; The single-view human point cloud data is input into a trained human model construction network to obtain a 3D human model of the test subject. The human model construction network is used to deform the 3D human template data in terms of posture and body shape according to the correspondence between the single-view human point cloud data and the pre-acquired single-view human point cloud data 3D human template data to obtain the 3D human model of the test subject. The training process of the human model construction network is equipped with global feature constraint loss and reconstruction loss. The global feature constraint loss is used to represent the difference between the global features corresponding to the complete human point cloud training data output by the encoder in the human model construction network and the global features corresponding to the single-view human point cloud training data. The reconstruction loss is used to determine the point-to-point error between the vertices of the 3D human model output by the decoder in the human model construction network and the complete human point cloud corresponding to the single-view human point cloud data. Target feature information is obtained by extracting features from target human body parts on a three-dimensional human body model of the subject; the target feature information includes the positional information of the target human body parts. The human body model construction network is trained using the following method: acquiring multiple complete human body point cloud data as complete human body point cloud training data; deleting partial regions of the complete human body point cloud data to obtain corresponding single-view human body point cloud training data; setting global feature constraint loss and reconstruction loss; and inputting the single-view human body point cloud training data, complete human body point cloud training data, and preset 3D human body template data into the human body model construction network based on the global feature constraint loss and reconstruction loss to train the human body model construction network; enabling the human body model construction network to obtain the correspondence between the single-view human body point cloud data and the 3D human body template data; the 3D human body template data is obtained by storing standard 3D human body data according to a preset storage format; The human body model construction network includes an encoder and a decoder. The encoder is used to extract features from the input complete human body point cloud training data and the single-view human body point cloud training data to obtain corresponding global features. The smaller the global feature constraint loss, the closer the single-view human body model features are to the complete human body features. The decoder is used to deform the three-dimensional human body template data in terms of posture and body shape according to the input global features.

4. A device for acquiring human body feature information, characterized in that, include: The acquisition module is used to acquire single-view human point cloud data of the subject. The model building module is used to input the single-view human point cloud data into the trained human model building network to obtain a three-dimensional human model of the test subject. The human model building network is used to deform the three-dimensional human template data in posture and body shape according to the single-view human point cloud data and the correspondence between the pre-acquired single-view human point cloud data and the three-dimensional human template data, thereby obtaining the three-dimensional human model of the test subject. The training process of the human model building network includes a global feature constraint loss and a reconstruction loss. The global feature constraint loss represents the difference between the global features corresponding to the complete human point cloud training data output by the encoder in the human model building network and the global features corresponding to the single-view human point cloud training data. The reconstruction loss is used to determine the point-to-point error between the vertices of the three-dimensional human model output by the decoder in the human model building network and the complete human point cloud corresponding to the single-view human point cloud data. The feature extraction module is used to extract features from target human body parts on the three-dimensional human body model of the subject to obtain target feature information; the target feature information includes the position information of the target human body parts; The human body model construction network is trained using the following method: acquiring multiple complete human body point cloud data as complete human body point cloud training data; deleting partial regions of the complete human body point cloud data to obtain corresponding single-view human body point cloud training data; setting global feature constraint loss and reconstruction loss; and inputting the single-view human body point cloud training data, complete human body point cloud training data, and preset 3D human body template data into the human body model construction network based on the global feature constraint loss and reconstruction loss to train the human body model construction network; enabling the human body model construction network to obtain the correspondence between the single-view human body point cloud data and the 3D human body template data; the 3D human body template data is obtained by storing standard 3D human body data according to a preset storage format; The human body model construction network includes an encoder and a decoder. The encoder is used to extract features from the input complete human body point cloud training data and the single-view human body point cloud training data to obtain corresponding global features. The smaller the global feature constraint loss, the closer the single-view human body model features are to the complete human body features. The decoder is used to deform the three-dimensional human body template data in terms of posture and body shape according to the input global features.

5. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the human feature information acquisition method as described in claim 3.

6. A non-transitory 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 human feature information acquisition method as described in claim 3.

7. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the human feature information acquisition method as described in claim 3.

Images