Method, device, equipment and medium for reconstructing real human body model based on 3D scanning

Abstract

The present invention relates to a method, device, equipment and medium for reconstructing a real human body model based on 3D scanning. The method comprises: scanning a target human body through 3D scanning technology to generate a three-dimensional point cloud model corresponding to the target human body; identifying the three-dimensional point cloud model to obtain the real dimensional values of each dimension of the target human body; deforming a pre-imported parameterized human body model according to the real dimensional values of each dimension so that the dimensional values of each dimension of the parameterized human body model tend to the real dimensional values; calculating the normal vectors of each vertex of the parameterized human body model through principal component analysis and adding the normal vectors to each vertex; performing skeletal binding, skinning, face reconstruction and motion driving on the parameterized human body model with the added normal vectors, thereby obtaining a real human body model. The virtual image generated by the present invention is highly similar to the scanned real person, and can freely control the multi-level changes of posture, body shape and body circumference.

Description

Technical Field

[0001] The present invention relates to the 3D field, and in particular to a method, device, equipment and medium for reconstructing a real human body model based on 3D scanning. Background Art

[0002] As a common three-dimensional model, the three-dimensional human body has always been an important research topic in computer graphics and computer vision, and has been widely used in film and television animation, three-dimensional games, clothing design, virtual fitting, e-commerce and other fields.

[0003] The current 3D human body reconstruction methods mainly include:

[0004] (1) Artificial construction method, which uses an artificially constructed virtual three-dimensional human body model to adjust the model shape by adjusting the fatness, thinness, height and shortness; achieve the facial features combination effect by adjusting the facial features map; realize action animation through pre-bound bone information; use part replacement technology to achieve clothing change (for example, dig out the torso and replace it with clothes) to avoid model penetration.

[0005] The disadvantage of this method is that it cannot achieve the rapid creation of virtual humans based on real-life images, and cannot extract and replicate the body characteristics of real people. Therefore, there will be a great sense of unreality when changing clothes / trying clothes, adjusting body shape / shaping transformations.

[0006] (2) Multidimensional description models based on principal component analysis. Multidimensional description models such as SMPL and STAR are parameterized models based on statistical learning that can be adjusted in multiple dimensions. They are widely used as basic models in many fields. In theory, they can achieve relatively realistic parametric virtual human reconstruction, which is close to real people in terms of body shape characteristics, and can also realize simple movement transformation.

[0007] The disadvantage of this method is that it can't perform parametric deformation of the human body based on a single, accurate piece of data, such as chest circumference, waist circumference, or arm length. Adjusting a single parameter often affects chain reactions in many other locations. Furthermore, because the model has a small number of faces and their distribution is even, joint positions are easily distorted during motion transformations.

[0008] (3) Human body reconstruction based on RGB or RGBD images. This is an image-based model reconstruction technology, with PiFu being a representative method. This method can achieve rapid reconstruction of real-life images. However, its disadvantages are that it has requirements for image quality and density, resulting in low model accuracy. This method also relies on action recognition, but the related algorithm has poor robustness, resulting in distortion of the reconstructed human posture. In addition, the reconstructed model is non-parametric and cannot be used for body shape adjustment, action transformation, and other operations. Summary of the Invention

[0009] In view of this, an object of the present invention is to provide a method, device, equipment and medium for reconstructing a real human body model based on 3D scanning to improve the above-mentioned problems.

[0010] An embodiment of the present invention provides a method for reconstructing a real human body model based on 3D scanning, which includes:

[0011] The target human body is scanned using 3D scanning technology to form point cloud data based on the three-dimensional spatial information of the target human body surface. The point cloud data obtained from multiple angles are then registered to generate a three-dimensional point cloud model corresponding to the target human body.

[0012] Identify the three-dimensional point cloud model to obtain the real dimension value of each dimension of the target human body;

[0013] Deforming a pre-imported parameterized human body model according to the real dimensional values of each dimension, so that the dimensional values of each dimension of the parameterized human body model tend to the real dimensional values;

[0014] Calculating the normal vector of each vertex of the adjusted parameterized human body model by principal component analysis, and adding the normal vector to each vertex;

[0015] The parameterized human body model after adding the normal vector is subjected to skeleton binding, skinning, face reconstruction and action driving, thereby obtaining a real human body model.

[0016] Preferably, the dimension values of each dimension of the parameterized human body model are obtained by the following steps:

[0017] Obtaining a vertex set corresponding to each dimension through a dimension definition file; wherein the dimension definition file stores the dimension name and the subscript value of the vertex constituting the dimension in a dictionary form;

[0018] The dimension value of the dimension is obtained by calculating the sum of the distances between the vertices; assuming that there are n vertices arranged in order that constitute the dimension, the dimension value of the dimension DistSum=D1D2+D2D3+...+Dn-1Dn+DnD1, where Dn-1Dn represents the distance between the n-1th point and the nth point.

[0019] Preferably, the pre-imported parametric human body model is deformed according to the real dimensional values of each dimension so that the dimensional values of each part of the parametric human body model tend to the real dimensional values, specifically including:

[0020] Get the current dimension value and the real dimension value of the dimension to be deformed;

[0021] Calculating an adjustment factor by a dichotomy method according to the current dimension value of the dimension to be deformed and the true dimension value;

[0022] According to the adjustment factor and the preset dimension adjustment file, the coordinates of the relevant vertices are modified so that the dimension value of the dimension to be deformed approaches the real dimension value; wherein, the dimension adjustment file records the vertices that need to be adjusted when adjusting each dimension and the adjustment amount of each vertex in each direction.

[0023] Preferably, the step of face reconstruction includes:

[0024] Acquire facial point cloud data from the three-dimensional point cloud model, and preprocess the facial point cloud data, wherein the preprocessing includes filtering and denoising, data reduction, and data interpolation;

[0025] Automatically register the pre-processed facial point cloud data and reconstruct the face model;

[0026] The reconstructed face model is subjected to parametric face reconstruction so that the face is deformed toward the scanned model.

[0027] Preferably, the automatic registration technology of point cloud data transforms the point cloud data measured in different coordinate systems to obtain the overall three-dimensional point cloud data, obtains the parameters of the coordinate transformation, and minimizes the distance between the three-dimensional point cloud data measured from two perspectives after the coordinate transformation.

[0028] Preferably, parameterized face reconstruction is performed on the reconstructed face model so that the face is deformed toward the scanned model, specifically including:

[0029] Extract key points from the reconstructed face model; this includes identifying the most convex point on the face as the nose tip, calculating the normal direction using vertices within a predetermined range from the nose tip as neighboring points, and finding the nasal root along the normal projected onto the yoz plane. Since a line perpendicular to the line connecting the nose tip and the nasal root passes through both eyes, the eye points are identified near the line using the normal direction.

[0030] Acquire parameterized points corresponding to the key points one by one from the parameterized face model of the parameterized human body model;

[0031] The scaling ratio is determined by matching the extracted key points with the parameterized points of the parameterized face to obtain the matching points required for point cloud fitting.

[0032] Perform rough matching of point clouds and preliminary fitting in the horizontal and vertical directions;

[0033] After the preliminary fitting, the point cloud of the face model is fitted with the parameterized face model according to the matching points, so that the parameterized face model is deformed toward the reconstructed face model.

[0034] Preferably, the dimensions include: neck circumference, neck height, shoulder distance, upper arm circumference, upper arm length, lower arm length, wrist circumference, chest width, chest circumference, waist width, waist circumference, hip width, hip circumference, thigh length, thigh circumference, calf length, calf circumference, ankle circumference, knee circumference, and chest height.

[0035] An embodiment of the present invention further provides a device for reconstructing a real human body model based on 3D scanning, comprising:

[0036] A scanning unit is used to scan a target human body using 3D scanning technology to form point cloud data based on the three-dimensional spatial information of the target human body surface, and to generate a three-dimensional point cloud model corresponding to the target human body by registering the point cloud data obtained from multiple angles;

[0037] an identification unit, configured to identify the three-dimensional point cloud model to obtain the true dimension values of each dimension of the target human body;

[0038] a deformation unit, configured to deform the pre-imported parameterized human body model according to the dimensional values of the respective dimensions, so that the dimensional values of the respective parts of the parameterized human body model tend to the real dimensional values;

[0039] a normal adding unit, configured to calculate the normal vectors of the respective vertices of the adjusted parameterized human body model by principal component analysis, and add the normal vectors to the respective vertices;

[0040] The reconstruction unit is used to perform skeleton binding, skinning, face reconstruction and action driving on the parameterized human body model after the normal vector is added, so as to obtain a real human body model.

[0041] An embodiment of the present invention also provides a real human body model reconstruction device based on 3D scanning, which includes a memory and a processor. The memory stores a computer program, and the computer program can be executed by the processor to implement the real human body model reconstruction method based on 3D scanning as described above.

[0042] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program, wherein the computer program can be executed by a processor of a device where the computer-readable storage medium is located to implement the above-mentioned method for reconstructing a real human body model based on 3D scanning.

[0043] In summary, this embodiment provides a complete solution for reconstructing a real human model from 3D scanning to creating a parametric, drivable virtual human avatar. The virtual avatar generated in this embodiment is highly similar to the scanned real person, and can freely control the multi-level changes in body shape, body shape, and body circumference.

[0044] Furthermore, the virtual avatar generated by this embodiment can be integrated with other technologies to achieve applications in many different fields. Combining it with motion capture equipment can create the effect of real-life action animation; combining it with clothing simulation software can achieve the effect of rendering a real person trying on clothing. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 Schematic diagram of the process of reconstructing a real human body model based on 3D scanning according to the first embodiment of the present invention.

[0046] Figure 2 This is a diagram showing some dimension definitions.

[0047] Figure 3 This is the front view of the base model.

[0048] Figure 4 Adjust the model waistline front view for factor=0.5.

[0049] Figure 5 Adjust the model waistline front view for factor=2.

[0050] Figure 6 Side view of the base model.

[0051] Figure 7 Adjust the model waistline side view for factor=0.5.

[0052] Figure 8 Adjust the model waistline side view for factor=2.

[0053] Figures 9-11 Schematic diagram of the calculation of adjustment factors using the dichotomy method.

[0054] Figure 12 This is a schematic diagram of the model without adding normal information.

[0055] Figure 13 Schematic diagram of the model with normal information added.

[0056] Figure 14 Schematic diagram of the structure of a real human body model reconstruction device based on 3D scanning according to the second embodiment of the present invention. DETAILED DESCRIPTION

[0057] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0058] In order to better understand the technical solution of the present invention, the embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0059] It should be understood that the embodiments described are only a portion of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative work are within the scope of protection of the present invention.

[0060] See also Figure 1 A first embodiment of the present invention provides a method for reconstructing a real human body model based on 3D scanning, which can be performed by a real human body model reconstruction device based on 3D scanning (hereinafter referred to as a reconstruction device), and in particular, is executed by one or more processors in the reconstruction device to implement the following steps:

[0061] S101, scanning a target human body using 3D scanning technology to form point cloud data based on the three-dimensional spatial information of the target human body surface, and generating a three-dimensional point cloud model corresponding to the target human body by registering the point cloud data obtained from multiple angles.

[0062] In this embodiment, the reconstruction device may be a desktop computer, a laptop computer, a workstation or other device with image processing and computing capabilities, and the present invention does not make any specific limitation.

[0063] In this embodiment, when a real human body model needs to be reconstructed, the target human body must first be scanned using 3D scanning technology. 3D scanning refers to 3D laser scanning technology, also known as real-life replication technology. It uses non-contact laser scanning to quickly and high-resolution obtain three-dimensional spatial information of the human body surface, forming point cloud data. By registering the point cloud data acquired from multiple angles, a 3D point cloud model with the same scale as the human body can be generated.

[0064] S102: Identify the three-dimensional point cloud model to obtain the real dimension value of each dimension of the target human body.

[0065] Among them, after generating a three-dimensional point cloud model with the same proportion as the human body, the algorithm identifies the different parts of the human body model and quickly obtains the dimensional values of different parts, including neck circumference, neck height, shoulder distance, upper arm circumference, upper arm length, lower arm length, wrist circumference, chest width, chest circumference, waist width, waist circumference, hip width, hip circumference, thigh length, thigh circumference, calf length, calf circumference, ankle circumference, knee circumference, chest height and more than 100 body dimensions and other data.

[0066] S103: deforming the pre-imported parameterized human body model according to the real dimensional values of each dimension, so that the dimensional values of each dimension of the parameterized human body model tend to the real dimensional values.

[0067] In this embodiment, the parametric human body model is an OBJ file. OBJ files are a standard 3D model file format that is well-suited for interoperability between 3D software models. In the parametric human body model, the OBJ file contains geometric vertices, face information, and vertex normals.

[0068] In this embodiment, the parametric human body models include two male and two female basic topological models, including a normal-weight male basic model, an overweight male basic model, a normal-weight female basic model, and an overweight female basic model. All male and female parametric human body models are derived by transforming appropriate basic models selected based on weight.

[0069] In this embodiment, each parameterized human body model includes, in addition to the model data, a dimension definition file and a dimension adjustment file; wherein:

[0070] Dimension definition file: saves the dimension name and the corresponding subscript value of the vertex that constitutes the dimension in the form of a dictionary. The position of some dimensions in the model is as follows: Figure 2 As shown. By calculating the sum of the distances between vertices, the value of the dimension can be obtained. Taking waist circumference as an example, assuming that there are n points arranged in order that constitute this dimension, the sum of the distances between n points and the vertices they are connected to is calculated as DistSum=D1D2+D2D3+...+D n-1 D n +D n D1, D n-1 D n Represents the distance between the n-1th point and the nth point, that is, the dimension value of the waist circumference.

[0071] Dimension adjustment file: This file stores the vertices, movement directions, and distances required to adjust a dimension. The file contains an n*4-column matrix, where n represents the number of vertices to adjust, the first column represents the subscript of the vertex to be adjusted, and columns 2-4 represent the x, y, and z displacements, respectively. For example, "1 0.1 0.05 -0.04" indicates that the x position of the first vertex in the parametric human body model needs to be increased by 0.1 units, the y position needs to be increased by 0.05 units, and the z position needs to be decreased by 0.04 units. By moving multiple vertices, the dimensions of the parametric human body model can be modified naturally.

[0072] In this embodiment, based on the above-mentioned dimension definition file and dimension adjustment file, step S103 is specifically as follows:

[0073] S1031, obtaining the current dimension value and the real dimension value of the dimension to be deformed.

[0074] The current dimension value can be obtained by calculating the above-mentioned dimension definition file in combination with the parameterized model.

[0075] S1032: Calculate an adjustment factor by a dichotomy method according to the current dimension value of the dimension to be deformed and the real dimension value.

[0076] S1033. Modify the coordinates of the relevant vertices according to the adjustment factor and the preset dimension adjustment file so that the dimension value of the dimension to be deformed approaches the real dimension value; wherein the dimension adjustment file records the vertices that need to be adjusted when adjusting each dimension and the adjustment amount of each vertex in each direction.

[0077] Among them, such as Figure 3-Figure 8 As shown, to adjust the dimension to a certain precise value, a suitable adjustment factor factor needs to be selected. This embodiment uses the idea of dichotomy to obtain the adjustment factor factor.

[0078] like Figures 9-11 As shown in the figure, for example, the waist circumference of the current base model is adjusted from 60cm to the true dimension value of 70cm, and the acceptable error range between the waist circumference and the true dimension value is 0.1cm. The range of factor is [0, 2]. Factor first takes the middle value "1" to calculate the waist circumference, resulting in a waist circumference of 65cm, which is less than 70cm. Factor then takes the middle value of "1" and the value on the right, and takes the value "1.5" to calculate the waist circumference, resulting in a waist circumference of 80cm. This is greater than 70cm. Factor then takes the middle value of "1.5" and the value on the left, and takes the value "1.25" to calculate the waist circumference, resulting in a waist circumference of 70.1cm. The difference from the target value of 70cm is within the error range, so the final value of factor is 1.25.

[0079] In this embodiment, through the above-mentioned adjustment method, the dimension values of each dimension of the parameterized human body model can be adjusted to approach the real dimension values of the target human body.

[0080] It should be noted that, in the above embodiment, the range of factor can be set according to actual needs and is not limited to the above range.

[0081] S104 , calculating the normal vector of each vertex of the adjusted parameterized human body model by principal component analysis, and adding the normal vector to each vertex.

[0082] In order to make the adjusted parameterized model have a good rendering effect, it is necessary to add a normal vector to each vertex.

[0083] Specifically, first find N adjacent points connected to the vertex, then perform principal component analysis on these adjacent points, and find the normal vector of the adjacent point with the smallest principal component as the normal vector of the vertex.

[0084] Figure 12 This is a schematic diagram of the model before adding the normal. You can see that the model looks rough. Figure 13 This is a schematic diagram of the model after adding normals. It can be seen that the appearance of the model after adding normals is obviously better. Figure 12 smooth.

[0085] S105 , performing skeleton binding, skinning, face reconstruction, and motion driving on the parameterized human body model after the normal vector is added, thereby obtaining a real human body model.

[0086] In this embodiment, the model is composed of a large number of vertices. If such a large number of vertices are manually moved to the specified position in each frame, it is obviously an impossible task. Therefore, people draw lessons from the relationship between animal bones and skin and design a virtual skeleton for the model.

[0087] Skeletons, sometimes called skeletons, are composed of individual bones, just like the skeleton in the human body. Bones are connected by joints, called bones. One or both ends of a bone are connected to two joints, and a joint may be connected to several bones. Each bone controls the vertices in its vicinity. When a bone moves, it pulls the bones it controls along with it.

[0088] Skinning is the process of combining a skeleton with a model. Since the skeleton and model are independent, skinning allows the skeleton to drive the model's movement in a reasonable manner. A single bone can control many vertices, and a single vertex can be controlled by multiple bones. Vertices are assigned the control rights of bones, known as weights. By controlling the position and rotation of joints, the entire skeleton can be controlled, and the entire skeleton, in turn, affects the model's outer skin (the vertex mesh), thus establishing an animation model.

[0089] Action Driven

[0090] BVH (Biovision Hierarchy) is the file extension for files generated by devices like BioVision after capturing human motion. It contains data on a character's skeleton and limb joint rotations. BVH is a universal file format for human feature animation. The animations depicted in this format are highly realistic because they are captured using a real mannequin wearing specialized sensor-equipped clothing. It is widely supported by popular animation software. Furthermore, this file is stored in text format, making it easy to use and develop.

[0091] A BVH file consists of two parts: a header and a data section. The header section describes the skeleton hierarchy and initial pose. The BVH skeleton information begins with the keyword "HIERARCHY," followed by 'ROOT', which represents the root node. Then, the joints below the root node, 'JOINT', are listed, recursively defining the child joints below the parent joint. 'OFFSET' represents the offset of the child joint relative to the parent joint, and can also indicate the length and orientation of the corresponding parent joint. If there are multiple child joints, the data for the first child joint is used. Next, 'CHANNELS' specifies the number and names of the channels. The ROOT always has six channels, while a typical JOINT only has three. Unlike the ROOT, the data section lacks XYZ position information. This is because the child joint's position in the coordinate system can be calculated based on its offset relative to the parent joint. The data section of the BVH file contains motion data. The data section begins with "MOTION," followed by a number indicating the frame rate. The next line, "Frame Time," defines the sampling frequency. The remainder of the file contains the motion data, with each line representing a sample of the motion data.

[0092] Face reconstruction

[0093] In this embodiment, the steps of face reconstruction include:

[0094] Acquire facial point cloud data from the three-dimensional point cloud model, and preprocess the facial point cloud data, wherein the preprocessing includes filtering and denoising, data reduction, and data interpolation;

[0095] Among them, face reconstruction is obtained by aligning multiple facial point cloud data. Since the data obtained by laser scanning often contains noise points and holes, in order to avoid affecting the subsequent point cloud processing, the data needs to be preprocessed. The preprocessing includes filtering and denoising, data simplification, data interpolation, etc.

[0096] The pre-processed facial point cloud data is automatically registered to reconstruct the face model.

[0097] After point cloud preprocessing, they are then registered. Due to incompleteness, rotational misalignment, and translational misalignment, obtaining a complete point cloud requires registering partial point clouds. To obtain a complete data model of the measured object, a suitable coordinate system must be determined. Point sets obtained from various viewpoints are then merged into a unified coordinate system to form a complete point cloud, which then facilitates visualization. This is point cloud data registration. Automatic point cloud registration technology automatically registers two point clouds by calculating the misalignment between them. Essentially, it transforms the point clouds measured in different coordinate systems to obtain a complete data model. The coordinate transformation parameters R (rotation matrix) and T (translation vector) are then determined to minimize the distance between the three-dimensional data measured from the two viewpoints after the coordinate transformation. The classic ICP registration algorithm first calculates the distance between all points in the initial point cloud and the target point cloud, ensuring that these points correspond to the closest points in the target point cloud. It also constructs an objective function based on the sum of squared residuals. The error function is minimized based on the least squares method, and the process is repeated until the mean square error is less than the set threshold.

[0098] The reconstructed face model is subjected to parametric face reconstruction so that the face is deformed toward the scanned model.

[0099] In order to make the parametric model closer to a real person, it is necessary to change the face and hairstyle of the model. By recognizing the contours of the face, the face shape of the parametric human body model is adjusted. Specifically, the following steps are performed:

[0100] Extract key points from the reconstructed face model; this includes identifying the most convex point on the face as the nose tip, calculating the normal direction using vertices within a predetermined range from the nose tip as neighboring points, and finding the nasal root along the normal projected onto the yoz plane. Since a line perpendicular to the line connecting the nose tip and the nasal root passes through both eyes, the eye points are identified near the line using the normal direction.

[0101] Acquire parameterized points corresponding to the key points one by one from the parameterized face model of the parameterized human body model;

[0102] The scaling ratio is determined by matching the extracted key points with the parameterized points of the parameterized face to obtain the matching points required for point cloud fitting.

[0103] Perform rough matching of point clouds and preliminary fitting in the horizontal and vertical directions;

[0104] After the preliminary fitting, the point cloud of the face model is fitted with the parameterized face model according to the matching points, so that the parameterized face model is deformed toward the reconstructed face model.

[0105] In summary, this embodiment provides a complete solution for reconstructing a real human model from 3D scanning to creating a parametric, drivable virtual human avatar. The virtual avatar generated in this embodiment is highly similar to the scanned real person, and can freely control the multi-level changes in body shape, body shape, and body circumference.

[0106] Furthermore, the virtual avatar generated by this embodiment can be integrated with other technologies to achieve applications in many different fields. Combining it with motion capture equipment can create the effect of real-life action animation; combining it with clothing simulation software can achieve the effect of rendering a real person trying on clothing.

[0107] See also Figure 14 The second embodiment of the present invention further provides a device for reconstructing a real human body model based on 3D scanning, which includes:

[0108] The scanning unit 210 is configured to scan a target human body using 3D scanning technology to generate point cloud data based on the three-dimensional spatial information of the target human body surface, and to generate a three-dimensional point cloud model corresponding to the target human body by registering the point cloud data acquired from multiple angles;

[0109] The recognition unit 220 is used to recognize the three-dimensional point cloud model to obtain the real dimension value of each dimension of the target human body;

[0110] a deformation unit 230 configured to deform the pre-imported parameterized human body model according to the dimensional values of the respective dimensions, so that the dimensional values of the respective parts of the parameterized human body model approach the real dimensional values;

[0111] a normal adding unit 240 for calculating the normal vectors of the adjusted vertices of the parameterized human body model by principal component analysis, and adding the normal vectors to the vertices;

[0112] The reconstruction unit 250 is used to perform skeleton binding, skinning, face reconstruction and action driving on the parameterized human body model after the normal vector is added, so as to obtain a real human body model.

[0113] The third embodiment of the present invention also provides a real human body model reconstruction device based on 3D scanning, which includes a memory and a processor. The memory stores a computer program, and the computer program can be executed by the processor to implement the real human body model reconstruction method based on 3D scanning as described above.

[0114] The fourth embodiment of the present invention further provides a computer-readable storage medium storing a computer program, which can be executed by a processor of a device where the computer-readable storage medium is located to implement the above-mentioned method for reconstructing a real human body model based on 3D scanning.

[0115] In the several embodiments provided in the embodiments of the present invention, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of the devices, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or part of the code, which contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the boxes can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, as well as the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified functions or actions, or can be implemented using a combination of dedicated hardware and computer instructions.

[0116] In addition, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0117] If the functions are implemented in the form of software modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion 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 for causing a computer device (which can be a personal computer, electronic device, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks. It should be noted that, in this document, the terms "comprise," "include," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or device that includes a series of elements includes not only those elements but also other elements not explicitly listed, or also includes elements inherent to such process, method, article, or device. Without further constraints, an element defined by the phrase "comprises a..." does not preclude the existence of additional identical elements in the process, method, article or apparatus that includes the element.

[0118] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for reconstructing a real human body model based on 3D scanning, characterized in that: include: The target human body is scanned using 3D scanning technology to form point cloud data based on the three-dimensional spatial information of the target human body surface. The point cloud data obtained from multiple angles are then registered to generate a three-dimensional point cloud model corresponding to the target human body. Identify the three-dimensional point cloud model to obtain the real dimension value of each dimension of the target human body; The pre-imported parametric human body model is deformed according to the real dimensional values of each dimension, so that the dimensional values of each dimension of the parametric human body model tend to the real dimensional values; the dimensional values of each dimension of the parametric human body model are obtained by the following steps: Obtaining a vertex set corresponding to each dimension through a dimension definition file; wherein the dimension definition file stores the dimension name and the subscript value of the vertex constituting the dimension in a dictionary form; The dimension value of the dimension is obtained by calculating the sum of the distances between the vertices. If there are n vertices arranged in order that constitute the dimension, the dimension value of the dimension DistSum=D1D2+D2D3+...+D n-1 D n +D n D1, D n-1 D n Represents the distance between the n-1th point and the nth point; Calculating the normal vector of each vertex of the adjusted parameterized human body model by principal component analysis, and adding the normal vector to each vertex; The parameterized human body model after adding the normal vector is subjected to skeleton binding, skinning, face reconstruction and action driving, thereby obtaining a real human body model.

2. The method for reconstructing a real human body model based on 3D scanning according to claim 1, characterized in that: Deforming a pre-imported parametric human body model according to the real dimensional values of each dimension so that the dimensional values of each part of the parametric human body model tend to the real dimensional values specifically includes: Get the current dimension value and the real dimension value of the dimension to be deformed; Calculating an adjustment factor by a dichotomy method according to the current dimension value of the dimension to be deformed and the true dimension value; According to the adjustment factor and the preset dimension adjustment file, the coordinates of the relevant vertices are modified so that the dimension value of the dimension to be deformed approaches the real dimension value; wherein, the dimension adjustment file records the vertices that need to be adjusted when adjusting each dimension and the adjustment amount of each vertex in each direction.

3. The method for reconstructing a real human body model based on 3D scanning according to claim 1, characterized in that: The steps of face reconstruction include: Acquire facial point cloud data from the three-dimensional point cloud model, and preprocess the facial point cloud data, wherein the preprocessing includes filtering and denoising, data reduction, and data interpolation; Automatically register the pre-processed facial point cloud data and reconstruct the face model; The reconstructed face model is subjected to parametric face reconstruction so that the face is deformed toward the scanned model.

4. The method for reconstructing a real human body model based on 3D scanning according to claim 3, characterized in that: The automatic registration technology of point cloud data transforms the point cloud data measured in different coordinate systems to obtain the overall three-dimensional point cloud data and obtain the parameters of the coordinate transformation, so that the distance between the three-dimensional point cloud data measured from two perspectives after coordinate transformation is minimized.

5. The method for reconstructing a real human body model based on 3D scanning according to claim 3, characterized in that: Perform parameterized face reconstruction on the reconstructed face model so that the face deforms toward the scanned model, specifically including: Extract key points from the reconstructed face model; this includes identifying the most convex point on the face as the nose tip, calculating the normal direction using vertices within a predetermined range from the nose tip as neighboring points, and finding the nasal root along the normal projected onto the yoz plane. Since a line perpendicular to the line connecting the nose tip and the nasal root passes through both eyes, the eye points are identified near the line using the normal direction. Acquire parameterized points corresponding to the key points one by one from the parameterized face model of the parameterized human body model; The scaling ratio is determined by matching the extracted key points with the parameterized points of the parameterized face to obtain the matching points required for point cloud fitting. Perform rough matching of point clouds and preliminary fitting in the horizontal and vertical directions; After the preliminary fitting, the point cloud of the face model is fitted with the parameterized face model according to the matching points, so that the parameterized face model is deformed toward the reconstructed face model.

6. The method for reconstructing a real human body model based on 3D scanning according to claim 1, characterized in that: The dimensions include: neck circumference, neck height, shoulder distance, upper arm circumference, upper arm length, lower arm length, wrist circumference, chest width, chest circumference, waist width, waist circumference, hip width, hip circumference, thigh length, thigh circumference, calf length, calf circumference, ankle circumference, knee circumference, and chest height.

7. A real human body model reconstruction device based on 3D scanning, characterized in that: include: A scanning unit is used to scan a target human body using 3D scanning technology to form point cloud data based on the three-dimensional spatial information of the target human body surface, and to generate a three-dimensional point cloud model corresponding to the target human body by registering the point cloud data obtained from multiple angles; an identification unit, configured to identify the three-dimensional point cloud model to obtain the true dimension values of each dimension of the target human body; A deformation unit is used to deform the pre-imported parametric human body model according to the dimensional values of each dimension so that the dimensional values of each part of the parametric human body model tend to the real dimensional values; the dimensional values of each dimension of the parametric human body model are obtained by the following steps: Obtaining a vertex set corresponding to each dimension through a dimension definition file; wherein the dimension definition file stores the dimension name and the subscript value of the vertex constituting the dimension in a dictionary form; The dimension value of the dimension is obtained by calculating the sum of the distances between the vertices. If there are n vertices arranged in order that constitute the dimension, the dimension value of the dimension DistSum=D1D2+D2D3+...+D n-1 D n +D n D1, D n-1 D n Represents the distance between the n-1th point and the nth point; a normal adding unit, configured to calculate the normal vectors of the respective vertices of the adjusted parameterized human body model by principal component analysis, and add the normal vectors to the respective vertices; The reconstruction unit is used to perform skeleton binding, skinning, face reconstruction and action driving on the parameterized human body model after the normal vector is added, so as to obtain a real human body model.

8. A device for reconstructing a real human body model based on 3D scanning, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the computer program can be executed by the processor to implement the method for reconstructing a real human body model based on 3D scanning as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that A computer program is stored, and the computer program can be executed by a processor of the device where the computer-readable storage medium is located to implement the real human body model reconstruction method based on 3D scanning as described in any one of claims 1 to 6.

Images