Virtual human automatic binding method and device, storage medium and electronic equipment

Abstract

The invention relates to a virtual human automatic binding method and device, a storage medium and electronic equipment, and relates to the technical field of computer animation, and the method comprises the steps: obtaining a target geometric shape of a three-dimensional human body model of a target virtual human; according to the target geometrical morphology, determining an adaptive skeleton structure corresponding to the target virtual human; based on the trained target machine learning model, determining at least one target skin weight corresponding to each model vertex in the three-dimensional human body model; determining at least one key part based on a target human body action of human body model distortion of a historical virtual human of the target role type and at least one corresponding target vertex; target skin weights of model vertexes in key parts of the three-dimensional human body model are verified, corresponding verified skin weights are obtained, and binding of the model vertexes and target bones is carried out based on the verified skin weights and the remaining skin weights. The method and the device have the effect of improving the authenticity of the action of the bound virtual human.

Description

Technical Field

[0001] The present application relates to the field of computer animation technology, and in particular to a method, device, storage medium and electronic device for automatic binding of virtual humans. Background Art

[0002] Virtual Human refers to a digital entity with humanoid form and characteristics created by technical means. With the development of underlying technologies such as deep learning and brain science, virtual images are bound to become more and more complete. In the future, virtual humans will play an important role in more fields. Virtual human binding refers to the process of automatically associating the mesh vertices of the virtual human body model, that is, the points that make up the shape of the human body model, with its internal bone structure. After binding, the operation of the skeleton can drive the mesh vertices in the human body model to move synchronously, thereby achieving the authenticity and smoothness of the virtual human's movements. This is crucial to improving the expressiveness and interactivity of virtual humans in games, film and television, virtual live broadcasts and other fields.

[0003] At present, the commonly used method for virtual human binding is: professional animators associate the mesh vertices and bone structure of the virtual human one by one, set corresponding skin weights for the mesh vertices, and finally complete the virtual human binding through manual binding. Once there are many mesh vertices that need to set skin weights, under this method, the accuracy of the skin weights assigned to each mesh vertex will be affected, resulting in poor realism of the virtual human's movements after binding. Summary of the invention

[0004] In order to improve the authenticity of virtual human actions after binding, the present application provides a virtual human automatic binding method, device, storage medium and electronic device.

[0005] In a first aspect of the present application, a method for automatic binding of a virtual human is provided, which specifically comprises: Acquire a target geometric shape of a three-dimensional human body model of a target virtual human; Determine, according to the target geometric form, an adaptation skeleton structure corresponding to the target virtual human, wherein the adaptation skeleton structure includes a plurality of target skeletons to be bound; Determine at least one target skinning weight corresponding to each model vertex in the three-dimensional human body model based on the trained target machine learning model; Determine a target role type of the target virtual human, and determine at least one key part based on a target human body action of a historical virtual human of the target role type where human body model distortion is likely to occur and at least one corresponding target vertex, wherein the target vertex is a vertex that is likely to cause model distortion under the corresponding target human body action, and the key part is a part of the human body model of the virtual human of the target role type that requires vertex skin weight verification; The target skin weights of the model vertices in the key parts of the three-dimensional human body model are verified to obtain corresponding verified skin weights, and the model vertices are bound to the target bones based on the verified skin weights and the remaining skin weights, wherein the remaining skin weights are the target skin weights of the model vertices except the key parts.

[0006] By adopting the above technical solution, after obtaining the target geometry of the three-dimensional human model, the skeletal structure of the target virtual human is automatically adapted based on the target geometry, and then the corresponding target skin weights are accurately and efficiently set for the model vertices through the target machine learning model. Furthermore, based on the target human body movements and the corresponding target vertices, the human body parts where the vertices of the virtual human of the target role type are prone to abnormalities are analyzed and determined, and they are determined as the key parts that need to be checked for vertex skin weights. Finally, the target skin weights of the model vertices in this key part of the target virtual human are checked, and finally the binding of the model vertices to the target skeleton is completed, thereby reducing the abnormalities of the model vertices in the key parts, thereby reducing the risk of distortion of the three-dimensional human model and improving the authenticity of the virtual human movements after binding.

[0007] Optionally, the determining of at least one key part based on a target human body action in which a historical virtual human of the target character type is prone to human body model distortion and at least one corresponding target vertex specifically includes: Acquire first historical human body movements that occur when human body model distortion occurs in the historical virtual human of the target character type, count the first occurrence times of each of the first historical human body movements, and select a first number of first historical human body movements from each of the first historical human body movements in descending order of the first occurrence times to determine as target human body movements; Acquire historical vertices that are accompanied by abnormalities when a human body model of a historical virtual human is distorted under a single target human body action, count the second number of occurrences of each of the historical vertices, and select a second number of historical vertices from each of the historical vertices in descending order of the second number of occurrences to determine as target vertices for the corresponding target human body action; Determine a first weight for each of the target human body movements, and determine a second weight for each of the target vertices corresponding to each of the target human body movements, wherein the first weight is a ratio of a first occurrence number of each target human body movement to a sum of first occurrence numbers of all target human body movements, and the second weight is a ratio of a second occurrence number of a single target vertex corresponding to the target human body movement to a sum of second occurrence numbers of all corresponding target vertices; At least one key part is determined according to the first weights and the corresponding second weights.

[0008] By adopting the above technical solution, the larger the number of first occurrences, the more likely the human body model will be distorted when the virtual human of the target character type makes the corresponding first historical human body, and then the target human body action is determined; the larger the number of second occurrences, when the model is distorted under a single target human body action, the corresponding historical vertex is likely to be abnormal, and then the target vertex is determined. Finally, the first weight and the corresponding second weight are combined to analyze the human body parts where the vertices of the virtual human of the target character type are prone to abnormalities, and then the key parts are accurately determined, so as to facilitate the subsequent verification of the skin weights of the model vertices in the corresponding parts of the three-dimensional human body model.

[0009] Optionally, determining at least one key part according to the first weight and the corresponding second weights specifically includes: Acquire second historical human body movements that have appeared in the animation of the historical virtual human of the target character type, count the occurrence frequency of each of the second historical human body movements, and select a third number of second historical human body movements from each of the second historical human body movements in descending order of the occurrence frequency to determine as high-frequency human body movements; Determine the high-frequency human body motion existing in each of the target human body motions as a reference human body motion, and calculate a first product of a first weight of each of the reference human body motions and a second weight of each of the corresponding target vertices; Summing the first products corresponding to at least one target vertex in the same human body part to obtain a sum of the corresponding first products; If the sum of the first products exceeds a preset first threshold, the corresponding human body part is determined as a key part.

[0010] By adopting the above technical solution, the larger the sum of the first products is, the more likely the target virtual human will have model distortion when making reference human body movements when vertex abnormalities appear in the corresponding human body parts. If the sum of the first products exceeds the first threshold, it means that the vertex abnormalities in the corresponding human body parts have a greater impact on the model distortion. In this case, the corresponding human body parts are determined as key parts, which facilitates the subsequent verification of the target skinning weights of the model vertices in the key parts of the three-dimensional human body model, reduces the risk of abnormalities in the model vertices, and thereby reduces the risk of distortion of the target virtual human in the animation.

[0011] Optionally, the method further includes: Determining at least one group of human body movement combinations of the target virtual person in the actual animation according to the movement design scheme of the target virtual person in the actual animation, wherein the actual animation is an animation produced for the target virtual person; Calculating a second product of a first weight of a target human body action in a single human body action combination and a second weight of each corresponding target vertex, and summing the second products corresponding to the target vertices in the same human body part to obtain a sum of the corresponding second products; If the sum of the second products is greater than a preset first threshold, the corresponding human body part is determined as a warning part, and a target appearance period of the corresponding human body action combination in the actual animation is determined; The playback mode of the target appearance period is set to slow playback, and when playing a single target appearance period, a check reminder for the corresponding alert area is sent to the personnel's terminal. The larger the sum of the second products, the slower the slow playback speed of the corresponding target appearance period.

[0012] By adopting the above technical solution, if the sum of the second products is greater than the preset first threshold, it means that the corresponding human body part has abnormal deformation under the combination of human body movements, resulting in a high possibility of distortion of the three-dimensional human body model, and then the corresponding human body part is determined as a warning part. Furthermore, the playback mode of the target appearance period is set to slow playback, and when playing a single target appearance period, a check reminder for the corresponding warning part is sent to the terminal of the personnel. The larger the sum of the second products, the slower the slow playback speed. This allows more targeted part checks when checking the actual animation, and timely detection and resolution of model distortion problems.

[0013] Optionally, the method further includes: Obtaining third historical human body movements when the historical virtual person of the target character type appears physically uncoordinated in the animation, counting the first occurrence frequency of each of the third historical human body movements, and selecting a fourth number of third historical human body movements from each of the third historical human body movements in descending order of the first occurrence frequencies to determine as important human body movements; Obtaining historical part combinations in which the historical virtual human has appeared incoordination under a single important human body action, counting the second occurrence frequency of each of the historical part combinations, and selecting a fifth number of historical part combinations from the historical part combinations in descending order of the second occurrence frequency to determine as important part combinations of the corresponding important human body action, wherein the historical part combinations are human body parts that have been deformed and become incoordination with each other when the historical virtual human performs the important human body action; Determine a third weight of each of the important human body movements, and determine a fourth weight of each of the important part combinations corresponding to each of the key human body movements, wherein the third weight is a ratio of a first occurrence frequency of the important human body movement to a sum of first occurrence frequencies of all the important human body movements, and the fourth weight is a ratio of a second occurrence frequency of a single important part combination corresponding to the important human body movement to a sum of second occurrence frequencies of all the corresponding important part combinations; Determining, according to the third weight and the corresponding fourth weights, whether there is a risk of incoordination in each of the key parts when the target virtual person moves; If so, the verified skin weights of the model vertices in the key parts where there is a risk of inconsistency are adjusted and optimized.

[0014] By adopting the above technical scheme, the greater the first occurrence frequency, the more likely the virtual human of the target character type will have incoordination problems due to the corresponding third historical human body movements, thereby determining the important human body movements; the greater the second occurrence frequency, the more likely the corresponding historical part combination will have incoordination problems under important human body movements, thereby determining the important part combination, and finally combining the third weight and the corresponding fourth weight to analyze the possibility of incoordination between the key parts, and then after determining that there is a risk of incoordination, the optimization of the skin weights after verification is carried out in a targeted manner, thereby reducing the risk of distortion of the three-dimensional human model while avoiding incoordination problems between different human body parts.

[0015] Optionally, determining whether there is an incoordination risk in each of the key parts when the target virtual person moves according to the third weight and the corresponding fourth weights specifically includes: If the important part combination includes multiple key parts, the corresponding important part combination is determined as the final part combination, and the important human body movements including the final part combination in each corresponding important part combination are determined as the final human body movements; Calculating a third product of the third weight of each final human body action and the fourth weight of the corresponding final part combination, and summing the third products to obtain a sum of the third products; If the sum of the third products is greater than a preset second threshold, it is determined that there is a risk of incoordination in each of the key parts.

[0016] By adopting the above technical solution, the larger the sum of the third products is, the more likely it is that incoordination problems will occur between different human body parts in the final part combination under each final human body action. Then, when the sum of the third products is greater than the preset second threshold, it means that there is a high possibility of incoordination problems between different human body parts in the final part combination, and it is determined that there is an incoordination risk in each key part. Otherwise, it is determined that there is no incoordination risk in each key part.

[0017] Optionally, if so, the verified skin weights of the model vertices in the key parts where there is a risk of incoordination are adjusted and optimized, specifically including: If so, selecting the largest third product from the third products; The key parts existing in the final part combination corresponding to the maximum third product are determined as the key parts with incoordination risks, and the verified skin weights of the model vertices in the key parts with incoordination risks are adjusted and optimized.

[0018] By adopting the above technical solution, the final part combination corresponding to the maximum third product can be understood as the combination of human body parts that are most likely to have incoordination problems. When there is an incoordination risk in each key part, the key part in the final part combination is determined as the key part with incoordination risk, and the skin weight after verification is optimized in a targeted manner, thereby reducing the risk of distortion of the three-dimensional human body model and avoiding the incoordination problem between different human body parts.

[0019] In a second aspect of the present application, a virtual human automatic binding device is provided, which specifically includes: An information acquisition module, used to acquire a target geometric form of a three-dimensional human body model of a target virtual human; A skeleton determination module, used to determine the matching skeleton structure corresponding to the target virtual person according to the target geometric form, wherein the matching skeleton structure includes a plurality of target skeletons that need to be bound; A weight allocation module, used to determine at least one target skinning weight corresponding to each model vertex in the three-dimensional human body model based on the trained target machine learning model; A part screening module is used to determine the target role type of the target virtual human, and determine at least one key part based on the target human body movements of the historical virtual human of the target role type that are prone to human body model distortion and at least one corresponding target vertex, wherein the target vertex is a vertex that is prone to model distortion under the corresponding target human body movement, and the key part is a part of the human body model of the virtual human of the target role type that needs to be checked for vertex skin weights; The automatic binding module is used to verify the target skin weights of the model vertices in the key parts of the three-dimensional human body model, obtain the corresponding verified skin weights, and bind the model vertices to the target bones based on the verified skin weights and the remaining skin weights, wherein the remaining skin weights are the target skin weights of the model vertices other than the key parts.

[0020] By adopting the above technical solution, after the information acquisition module obtains the target geometric shape, the skeleton determination module determines the corresponding adaptive skeleton structure of the target virtual person, then the weight distribution module determines the target skinning weight of each model vertex, and then the part screening module determines at least one key part, and finally the automatic binding module completes the binding of the model vertex and the target skeleton.

[0021] In a third aspect of the present application, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is loaded and executed by a processor, the method steps described in any one of the first aspects are performed.

[0022] In a fourth aspect of the present application, an electronic device is provided, specifically comprising: A processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the processor is used to load and execute the computer program stored in the memory so that the electronic device performs the method as described in any one of the first aspects.

[0023] In summary, the present application includes at least one of the following beneficial technical effects: automatically adapting the skeletal structure of the target virtual human based on the target geometry, and then accurately and efficiently setting the corresponding target skinning weights for the model vertices through the target machine learning model, and further, based on the target human body movements and the corresponding target vertices, analyzing and determining the human body parts where the vertices of the virtual human of the target role type are prone to abnormalities, and determining them as the key parts that need to be checked for vertex skinning weights. Finally, the target skinning weights of the model vertices in this key part of the target virtual human are checked, and finally the binding of the model vertices to the target skeleton is completed, thereby reducing the abnormalities of the model vertices in the key parts, thereby reducing the risk of distortion of the three-dimensional human model, and improving the authenticity of the virtual human movements after binding. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flowchart of a method for automatic binding of a virtual person provided in an embodiment of the present application; Figure 2 It is a flowchart of another method for automatic binding of virtual humans provided in an embodiment of the present application; Figure 3 It is a structural schematic diagram of a virtual human automatic binding device provided in an embodiment of the present application; Figure 4 It is a structural schematic diagram of another virtual human automatic binding device provided in an embodiment of the present application.

[0025] Explanation of the accompanying drawings: 11. Information acquisition module; 12. Skeleton determination module; 13. Weight allocation module; 14. Part screening module; 15. Automatic binding module; 16. Troubleshooting reminder module; 17. Weight optimization module. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0027] In the description of the embodiments of the present application, words such as "illustrative", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "illustrative", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "illustrative", "for example" or "for example" is intended to present related concepts in a concrete way.

[0028] In the description of the embodiments of the present application, the term "and / or" is only a kind of association relationship describing the associated objects, indicating that there may be three kinds of relationships, for example, A and / or B, which can represent: A exists alone, B exists alone, and A and B exist at the same time. In addition, unless otherwise specified, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" can explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0029] See also Figure 1 The present application embodiment discloses a flowchart of a virtual human automatic binding method, which can be implemented by a computer program or run on a virtual human automatic binding device based on the von Neumann system. The computer program can be integrated into an application or run as an independent tool application, specifically including: S101: Acquire a target geometric shape of a three-dimensional human body model of a target virtual human.

[0030] Specifically, in the embodiments of the present application, the role type of the virtual person will be different with different application scenarios. The target virtual person can be a cartoon character in the game. In other embodiments, the target virtual person can also be a digital human anchor, etc. The target virtual person is a virtual person whose three-dimensional human body model needs to be bound at present. The three-dimensional human body model is a virtual person form with a three-dimensional stereoscopic form created by using advanced computer graphics technology and modeling technology. Further, the three-dimensional human body model of the target virtual person is mainly created by three-dimensional modeling software, wherein the three-dimensional modeling software can use Maya software, and in other embodiments, the three-dimensional modeling software can also use Blender software. In addition, the execution subject of a virtual person automatic binding method disclosed in the present application is a server, and the server is wirelessly connected to the terminal. The terminal can be a personal computer or tablet computer of an animator, and an automatic binding client is installed in the terminal. The server can be a background server of the client, which can be an independent physical server or a server cluster composed of multiple physical servers. Further, a feasible way to obtain the target geometric form is to export the geometric data corresponding to the three-dimensional human body model through the three-dimensional modeling software, and the exported 3D file format can be obj format or fbx format. Finally, the geometric data is sent to the server through the client, and the target geometric form of the three-dimensional human body model of the target virtual person is finally obtained. It should be noted that the target geometric form is a human body geometric shape model representing the target virtual person in the form of a three-dimensional mesh, which includes geometric features such as the shape, size and proportion of the human body.

[0031] S102: Determine an adaptation skeleton structure corresponding to the target virtual human according to the target geometric shape, where the adaptation skeleton structure includes a plurality of target skeletons that need to be bound.

[0032] Specifically, after the target geometry is determined, the preset skeletal structure prediction model is used to determine the skeletal structure or skeletal system that is adapted to the target virtual person based on the target geometry, that is, the adapted skeletal structure, wherein the skeletal structure prediction model is a trained deep learning model, and the deep learning model can be a Transformer model or a recurrent neural network model. The training process is briefly described as follows: the historical human body model with a determined skeletal structure is used as a training sample to train the model so that the model continuously understands the proportion, symmetry and other information of the human body model, and learns the association between each part of the human body model and the corresponding bones. During the training process, the model parameters are continuously optimized by the reverse gradient algorithm until the model converges to obtain a skeletal structure prediction model, which can well predict the corresponding skeletal structure according to the geometry of the human body model. This is a prior art and will not be repeated here. Among them, the skeletal structure is a framework for simulating the human skeletal system built for the virtual person in the virtual environment. This framework not only simulates the morphology and position of the human skeleton, but also simulates the connection mode and movement law between the bones, so that the virtual person can perform various realistic actions and postures. It should be noted that the adaptation skeleton structure includes multiple target skeletons that need to be bound, that is, includes multiple target skeletons that need to be bound to the three-dimensional human body model.

[0033] S103: Based on the trained target machine learning model, determine at least one target skinning weight corresponding to each model vertex in the three-dimensional human body model.

[0034] Specifically, the skin weights are used to determine the degree of association between vertices and bones in the human body model, that is, the extent to which the vertex is affected by the movement of a single bone or multiple bones, specifically the weight distribution value of a single bone to the vertex. The skin weight ranges from 0 to 1. The larger the skin weight of a single vertex, the higher the control degree of the corresponding bone over the vertex, and the greater the influence of the corresponding bone movement. Each vertex is usually affected by the weight distribution values ​​of 2-4 bones, that is, the deformation or movement of the vertex will be affected by the movement of 2-4 bones. After the adaptation bone structure is determined, based on the trained target machine learning model and combined with the positions of each model vertex in the three-dimensional human body model, at least one target skin weight is determined for each model vertex, that is, the weight distribution value of at least one target bone to the model vertex. Among them, the target machine learning model can be a recurrent neural network model, or a random forest model, etc. The training process is briefly described as follows: the skin weight data of each vertex in the historical virtual human body model with assigned skin weights is used as a training sample, and the target machine learning model is trained so that it continuously learns the skin weight distribution scheme until the model converges. In other embodiments, at least one target skinning weight corresponding to each model vertex may also be determined by a linear blend skinning method (LBS).

[0035] S104: Determine a target role type of the target virtual human, and determine at least one key part based on a target human body motion of a historical virtual human of the target role type that is prone to human body model distortion and at least one corresponding target vertex.

[0036] Specifically, the target vertex is the vertex that is prone to model distortion under the corresponding target human body action, and the key part is the part of the virtual human body model of the target role type where the skin weight of the vertex needs to be verified. Among them, model distortion refers to the unnatural problem that occurs in the virtual human body model during the presentation process compared with the real human body posture or form. For example, the human body model appears distorted or stiff in movement when moving. Further, to determine the target role type of this target virtual person, a feasible determination method is: receiving the role description information sent by the terminal for the target virtual person, and inputting the role description information into a preset role type prediction model to obtain the target role type. The role type prediction model can use a trained recurrent neural network model.

[0037] Furthermore, the historical verification records of the virtual human animation are retrieved, and the historical verification records include the role type of the verified virtual human, the human body movements when the human body model distortion occurs, and the abnormal vertices in the human body model. Based on the historical verification records, the first historical human body movements that occurred when the human body model distortion occurred in the historical virtual human of the target role type, that is, the movements made by the human body model, are obtained, and the first occurrence times of each first historical human body movement are counted. The larger the first occurrence times, the more likely the human body model distortion will occur when the virtual human of the target role type makes the corresponding first historical human body, and the first number of first historical human body movements are selected from each first historical human body movement in the order of the first occurrence times from large to small to determine as the target human body movement, that is, the movement that is likely to cause human body model distortion.

[0038] Furthermore, when the human body model of the historical virtual person is distorted under a single target human body action, the historical vertices that are accompanied by abnormalities are obtained, and the second occurrence number of each historical vertex is counted. The larger the second occurrence number, the more likely the corresponding historical vertex is to be abnormal when the model is distorted under a single target human body action, that is, a deformation abnormality or a movement abnormality occurs, and in descending order of the second occurrence number, a second number of historical vertices are selected from each historical vertex to be determined as the target vertices of the corresponding target human body action. Next, the first weight of each target human body action is determined, the first weight is the ratio of the first occurrence number of each target human body action to the sum of the first occurrence numbers of all target human body actions, and the second weight of each target vertex corresponding to each target human body action is determined, the second weight is the ratio of the second occurrence number of a single target vertex corresponding to the target human body action to the sum of the second occurrence numbers of all corresponding target vertices.

[0039] Finally, according to the first weight and the corresponding second weight, the key parts of the vertex skin weight that need to be checked are determined. A feasible determination method is: based on the action design records of multiple virtual people in animations, the action design records include but are not limited to the role type of the virtual people and at least one human action, obtain the second historical human action that has appeared in the animation of the historical virtual people of the target role type, and count the frequency of occurrence of each second historical human action. The greater the frequency of occurrence, the more likely the virtual person of the target role type is to appear in the corresponding second historical human action. According to the order of the frequency of occurrence from large to small, the third number of second historical human actions are selected from each second historical human action to be determined as high-frequency human actions, that is, human actions that the target virtual person is easy to make. Further, the high-frequency human actions existing in each target human action are determined as reference human actions, and the reference human actions are human actions that are easy for the target virtual person to make and are easy to induce distortion of the three-dimensional human model. Calculate the first product of the first weight of each reference human action and the second weight of each corresponding target vertex. The larger the first product, the more likely the three-dimensional human model of the target virtual person is distorted when the corresponding target vertex is abnormal under the reference human action. The first products of at least one target vertex in the same human body part are summed to obtain the corresponding sum of the first products. The larger the sum of the first products is, the more likely the target virtual person will have model distortion when making reference human body movements when there is a vertex abnormality in the corresponding human body part. Finally, the sum of the first products is compared with a preset first threshold. If the sum of the first products exceeds the first threshold, it means that the vertex abnormality in the corresponding human body part has a greater impact on the model distortion. Then the corresponding human body part is determined as a key part, which is convenient for the subsequent verification of the target skinning weights of the model vertices in the key parts of the three-dimensional human body model, thereby reducing the distortion risk of the target virtual person in the animation.

[0040] In other embodiments, after the target virtual person is bound and the actual animation is produced according to the corresponding action design scheme, the animator determines at least one human action combination of the target virtual person in the actual animation based on the action design scheme during the verification of the actual animation, and the human action combination includes at least one human action. The actual animation is an animation produced for the target virtual person. Further, the second product of the first weight of the target human action existing in a single human action combination and the second weight of each corresponding target vertex is calculated, and then the second products corresponding to the target vertices in the same human body part are summed to obtain the sum of the second products of the corresponding human body parts. If the sum of the second products is greater than the preset first threshold, it means that the corresponding human body part has an abnormal deformation under the human action combination, resulting in a high possibility of distortion of the three-dimensional human body model, then the corresponding human body part is determined as a warning part, and the target appearance period of the corresponding human action combination in the actual animation is determined, that is, the period from the beginning to the end of the human action combination.

[0041] Furthermore, the playback mode of the target appearance period is set to slow playback, and when playing a single target appearance period, a check reminder for the corresponding alert part is sent to the terminal of the personnel. The larger the sum of the second product, the slower the slow playback speed of the corresponding target appearance period. This allows more targeted part checks to be carried out when checking the actual animation, and timely discover and solve the problem of model distortion.

[0042] S105: Verify the target skin weights of the model vertices in the key parts of the 3D human body model to obtain corresponding verified skin weights, and bind the model vertices to the target bones based on the verified skin weights and the remaining skin weights.

[0043] Specifically, after the key parts are determined, the target skin weights of each model vertex in the key parts of the three-dimensional human body model are verified. A feasible verification method is to send a weight verification reminder to the animator's terminal for this key part, so that the animator can verify and adjust the target skin weights of each model vertex in the key part after automatically assigning the target skin weights, thereby improving the rationality of the skin weights. Finally, through the preset Maya tool, based on the skin weights of each model vertex, the model vertices in the three-dimensional human body model are bound to the target bones. This is a prior art and will not be repeated here. It should be noted that the above-mentioned model vertices or vertices can be understood as mesh vertices.

[0044] See also Figure 2The present application embodiment discloses a flowchart of another method for automatically binding a virtual human, which can be implemented by a computer program or run on a virtual human automatically binding device based on the von Neumann system. The computer program can be integrated into an application or run as an independent tool application, specifically including: S201: Acquire a target geometric shape of a three-dimensional human body model of a target virtual human.

[0045] S202: Determine an adaptation skeleton structure corresponding to the target virtual human according to the target geometric shape, where the adaptation skeleton structure includes a plurality of target skeletons that need to be bound.

[0046] S203: Based on the trained target machine learning model, determine at least one target skinning weight corresponding to each model vertex in the three-dimensional human body model.

[0047] S204: Determine a target role type of the target virtual human, and determine at least one key part based on a target human body motion of a historical virtual human of the target role type that is prone to human body model distortion and at least one corresponding target vertex.

[0048] S205: Verify the target skin weights of the model vertices in the key parts of the 3D human body model to obtain corresponding verified skin weights, and bind the model vertices to the target bones based on the verified skin weights and the remaining skin weights.

[0049] For details, please refer to steps S101-S105, which will not be described in detail here.

[0050] S206: Obtain the third historical human body movements of the historical virtual human of the target character type when body incoordination occurs in the animation, count the first occurrence frequency of each third historical human body movement, and select a fourth number of third historical human body movements from each third historical human body movement in descending order of the first occurrence frequency to determine as important human body movements.

[0051] S207: Obtain historical part combinations in which the virtual human has appeared inconsistent under a single important human body action, count the second occurrence frequency of each historical part combination, and select a fifth number of historical part combinations from each historical part combination in descending order of the second occurrence frequency to determine as the important part combinations of the corresponding important human body action.

[0052] Specifically, after the binding is completed, based on the animator's historical inspection records of the overall coordination problems of the human body model in the virtual human animation, the historical inspection records include whether there are incoordination problems in the virtual human body model, the incoordination of the human body parts, and the human body movements of the virtual human when the incoordination occurs, etc., and then obtain the third historical human body movements of the historical virtual human of the target character type when the body is incoordinating in the animation, count the first occurrence frequency of the third historical human body movements, and select the fourth number of third historical human body movements from each third historical human body movement in order from large to small according to the first occurrence frequency to determine as important human body movements, that is, human body movements that are likely to cause incoordination problems with the human body model. Among them, the incoordination problem refers to the lack of coordination between the movements of the various human body parts of the human body model, resulting in an unnatural overall posture of the human body model. For example, when walking, the legs of the human body model should be coordinated with the movements of the upper body. Further, based on the above historical inspection records, historical part combinations that have been inconsistent with the historical virtual human under a single important human body action are obtained, the second occurrence frequency of each historical part combination is counted, and the fifth number of historical part combinations are selected from each historical part combination in descending order of the second occurrence frequency to determine as the important part combinations of the corresponding important human body action, that is, the human body parts that are prone to inconsistency. The historical part combinations are human body parts that are inconsistent with each other when the historical virtual human performs important human body actions.

[0053] S208: Determine a third weight of each important human body motion, and determine a fourth weight of each important part combination corresponding to each key human body motion.

[0054] S209: Determine, based on the third weight and the corresponding fourth weights, whether there is an incoordination risk in each key part when the target virtual human moves.

[0055] S210: If yes, the verified skin weights of the model vertices in the key parts where there is a risk of incompatibility are adjusted and optimized.

[0056] Specifically, the third weight is the ratio of the first occurrence frequency of the important human body action to the sum of the first occurrence frequencies of all important human body actions, and the fourth weight is the ratio of the second occurrence frequency of a single important part combination corresponding to the important human body action to the sum of the second occurrence frequencies of all corresponding important part combinations. Then, if the important part combination contains multiple key parts, the corresponding important part combination is determined as the final part combination, and the important human body action containing this final part combination in each corresponding important part combination is determined as the final human body action. The third product of the third weight of each final human body action and the fourth weight of the corresponding final part combination is calculated, and each third product is summed to obtain the sum of the third products. The larger the sum of the third products, the easier it is for different human body parts in the final part combination to have incoordination problems under each final human body action. When the sum of the third products is greater than the preset second threshold, it means that there is a greater possibility of incoordination problems between different human body parts in the final part combination, and it is determined that there is an incoordination risk in each key part. Otherwise, it is determined that there is no incoordination risk in each key part.

[0057] If it is determined that there is a risk of incoordination, then the maximum third product is selected from each third product, and the multiple key parts in the final part combination corresponding to the maximum third product are determined as key parts with incoordination risks, and the skin weights of the model vertices in these key parts are adjusted and optimized after verification, specifically, the optimization reminder information is sent to the animator's terminal, so that the animator optimizes the skin weights after verification for coordination problems, thereby reducing the risk of distortion of the three-dimensional human body model and avoiding the problem of incoordination between different human body parts. Among them, the final part combination corresponding to the maximum third product is: the human body part combination that is most likely to have incoordination problems. In other embodiments, the fourth product of the third weight of each important human body action and the fourth product of the corresponding important part combination is calculated, and the fourth product greater than the preset product threshold is selected from each fourth product, and determined as the target product, and the important human body action corresponding to the target product is determined as a warning human body action. When the animator subsequently checks the actual animation of the target virtual person, when this warning human body action appears, a coordination problem inspection reminder is issued to remind the animator to focus on checking whether there is an incoordination problem in the human body parts.

[0058] The implementation principle of a method for automatic binding of virtual humans in an embodiment of the present application is as follows: based on the target geometric form, the skeletal structure of the target virtual human is automatically adapted, and then the corresponding target skinning weights are accurately and efficiently set for the model vertices through the target machine learning model. Furthermore, based on the target human body movements and the corresponding target vertices, the human body parts where the vertices of the virtual human of the target role type are prone to abnormalities are analyzed and determined, and they are determined as the key parts that need to be checked for vertex skinning weights. Finally, the target skinning weights of the model vertices in this key part of the target virtual human are checked, and finally the binding of the model vertices to the target skeleton is completed, thereby reducing the abnormalities of the model vertices in the key parts, thereby reducing the risk of distortion of the three-dimensional human model, and improving the authenticity of the virtual human movements after binding.

[0059] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0060] See also Figure 3 , which is a schematic diagram of the structure of the virtual human automatic binding device provided in the embodiment of the present application. The virtual human automatic binding device can be implemented as all or part of the device through software, hardware or a combination of both. The device includes an information acquisition module 11, a skeleton determination module 12, a weight allocation module 13, a part screening module 14 and an automatic binding module 15.

[0061] An information acquisition module 11 is used to acquire a target geometric shape of a three-dimensional human body model of a target virtual human; The skeleton determination module 12 is used to determine the matching skeleton structure corresponding to the target virtual person according to the target geometric form, and the matching skeleton structure includes a plurality of target skeletons that need to be bound; A weight allocation module 13 is used to determine at least one target skinning weight corresponding to each model vertex in the three-dimensional human body model based on the trained target machine learning model; The part screening module 14 is used to determine the target role type of the target virtual human, and determine at least one key part based on the target human body movements of the historical virtual human of the target role type that are prone to human body model distortion and at least one corresponding target vertex, wherein the target vertex is a vertex that is prone to model distortion under the corresponding target human body movement, and the key part is a part of the human body model of the virtual human of the target role type that needs to be checked for vertex skin weights; The automatic binding module 15 is used to verify the target skin weights of the model vertices in the key parts of the three-dimensional human body model, obtain the corresponding verified skin weights, and bind the model vertices to the target bones based on the verified skin weights and the remaining skin weights. The remaining skin weights are the target skin weights of the model vertices except the key parts.

[0062] Optionally, the part screening module 14 is specifically used for: Acquire the first historical human body movements that occur when the human body model of the historical virtual human of the target character type is distorted, count the first occurrence times of each first historical human body movement, and select the first number of first historical human body movements from each first historical human body movement in descending order of the first occurrence times to determine as the target human body movements; Acquire abnormal historical vertices when a human body model of a historical virtual human under a single target human body action is distorted, count the second occurrence times of each historical vertex, and select a second number of historical vertices from each historical vertex in descending order of the second occurrence times to determine as target vertices for the corresponding target human body action; Determine a first weight for each target human action, and determine a second weight for each target vertex corresponding to each target human action, wherein the first weight is a ratio of a first occurrence number of each target human action to a sum of first occurrence numbers of all target human actions, and the second weight is a ratio of a second occurrence number of a single target vertex corresponding to the target human action to a sum of second occurrence numbers of all corresponding target vertices; At least one key part is determined according to the first weight and the corresponding second weights.

[0063] Optionally, the part screening module 14 is specifically used for: Obtaining second historical human body movements that have appeared in the animation of the historical virtual human of the target character type, counting the occurrence frequency of each second historical human body movement, and selecting a third number of second historical human body movements from each second historical human body movement in descending order of occurrence frequency to determine as high-frequency human body movements; Determine the high-frequency human body motions in each target human body motion as reference human body motions, and calculate a first product of a first weight of each reference human body motion and a second weight of each corresponding target vertex; Summing the first products corresponding to at least one target vertex in the same human body part to obtain a sum of the corresponding first products; If the sum of the first products exceeds a preset first threshold, the corresponding human body part is determined as a key part.

[0064] Optional, such as Figure 4 As shown, the device also includes a troubleshooting reminder module 16, which is specifically used to: Determine at least one set of human body movement combinations of the target virtual person in the actual animation according to the movement design scheme of the target virtual person in the actual animation, where the actual animation is an animation produced for the target virtual person; Calculate the second product of the first weight of the target human body action in the single human body action combination and the second weight of each corresponding target vertex, and sum the second products corresponding to the target vertices in the same human body part to obtain the sum of the corresponding second products; If the sum of the second products is greater than a preset first threshold, the corresponding human body part is determined as a warning part, and a target appearance period of the corresponding human body action combination in the actual animation is determined; The playback mode of the target appearance period is set to slow playback, and when playing a single target appearance period, a check reminder for the corresponding alert area is sent to the personnel's terminal. The larger the sum of the second products, the slower the slow playback speed of the corresponding target appearance period.

[0065] Optionally, the device further includes a weight optimization module 17, which is specifically used to: Obtaining the third historical human body movements when the historical virtual person of the target character type appears physically uncoordinated in the animation, counting the first occurrence frequency of each third historical human body movement, and selecting a fourth number of third historical human body movements from each third historical human body movement in descending order of the first occurrence frequency to determine as important human body movements; Obtain historical part combinations in which the historical virtual human has appeared incoordination under a single important human body action, count the second occurrence frequency of each historical part combination, and select a fifth number of historical part combinations from each historical part combination in descending order of the second occurrence frequency to determine as the important part combinations of the corresponding important human body action, the historical part combinations being the human body parts that have been deformed and incoordinated when the historical virtual human performs the important human body action; Determine a third weight for each important human body action, and determine a fourth weight for each important part combination corresponding to each key human body action, the third weight being a ratio of a first occurrence frequency of the important human body action to a sum of first occurrence frequencies of all important human body actions, and the fourth weight being a ratio of a second occurrence frequency of a single important part combination corresponding to the important human body action to a sum of second occurrence frequencies of all corresponding important part combinations; Determining whether there is a risk of incoordination in each key part when the target virtual person moves according to the third weight and the corresponding fourth weights; If so, the verified skin weights of the model vertices in the key parts where there is a risk of inconsistency are adjusted and optimized.

[0066] Optionally, the weight optimization module 17 is specifically used for: If the important part combination includes multiple key parts, the corresponding important part combination is determined as the final part combination, and the important human body movements including the final part combination in the corresponding important part combinations are determined as the final human body movements; Calculating a third product of the third weight of each final human body action and the fourth weight of the corresponding final part combination, and summing the third products to obtain a sum of the third products; If the sum of the third products is greater than a preset second threshold, it is determined that there is a risk of incoordination in each key area.

[0067] Optionally, the weight optimization module 17 is specifically used for: If so, select the largest third product from each third product; The key parts in the final part combination corresponding to the maximum third product are determined as the key parts with incoordination risks, and the verified skin weights of the model vertices in the key parts with incoordination risks are adjusted and optimized.

[0068] It should be noted that the above embodiment provides a virtual human automatic binding device, when executing the virtual human automatic binding method, only uses the division of the above functional modules as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the above embodiment provides a virtual human automatic binding device and a virtual human automatic binding method embodiment, which belong to the same concept. The implementation process is detailed in the method embodiment, which will not be repeated here.

[0069] The embodiment of the present application further discloses a computer-readable storage medium, and the computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, a virtual human automatic binding method of the above embodiment is adopted.

[0070] Among them, the computer program can be stored in a computer-readable medium, the computer program includes computer program code, the computer program code can be in the form of source code, object code, executable file or certain middleware, etc. The computer-readable medium includes any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the computer-readable medium includes but is not limited to the above-mentioned components.

[0071] Among them, through this computer-readable storage medium, a virtual human automatic binding method of the above embodiment is stored in a computer-readable storage medium, and is loaded and executed on a processor to facilitate the storage and application of the above method.

[0072] An embodiment of the present application further discloses an electronic device, wherein a computer program is stored in a computer-readable storage medium, and when the computer program is loaded and executed by a processor, the above-mentioned virtual human automatic binding method is adopted.

[0073] The electronic device may be a desktop computer, a laptop computer, a cloud server or other electronic device, and the electronic device includes but is not limited to a processor and a memory. For example, the electronic device may also include input and output devices, a network access device, and a bus.

[0074] Among them, the processor can adopt a central processing unit (CPU). Of course, according to actual usage, other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. can also be adopted. The general-purpose processor can adopt a microprocessor or any conventional processor, etc., and this application does not impose any restrictions on this.

[0075] Among them, the memory can be an internal storage unit of the electronic device, such as a hard disk or memory of the electronic device, or it can be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital card (SD) or a flash memory card (FC) equipped on the electronic device. Moreover, the memory can also be a combination of an internal storage unit and an external storage device of the electronic device. The memory is used to store computer programs and other programs and data required by the electronic device. The memory can also be used to temporarily store data that has been output or is to be output, and this application does not impose any restrictions on this.

[0076] Among them, through this electronic device, a virtual human automatic binding method of the above embodiment is stored in the memory of the electronic device, and is loaded and executed on the processor of the electronic device for easy use.

[0077] The above is only an exemplary embodiment of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. This application is intended to cover any variation, use or adaptive change of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field that are not recorded in the present disclosure. The description and examples are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A virtual human automatic binding method, characterized in that: The method comprises: Acquire a target geometric shape of a three-dimensional human body model of a target virtual human; Determine, according to the target geometric form, an adaptation skeleton structure corresponding to the target virtual human, wherein the adaptation skeleton structure includes a plurality of target skeletons to be bound; Determine at least one target skinning weight corresponding to each model vertex in the three-dimensional human body model based on the trained target machine learning model; Determine a target role type of the target virtual human, and determine at least one key part based on a target human body action of a historical virtual human of the target role type where human body model distortion is likely to occur and at least one corresponding target vertex, wherein the target vertex is a vertex that is likely to cause model distortion under the corresponding target human body action, and the key part is a part of the human body model of the virtual human of the target role type that requires vertex skin weight verification; The target skin weights of the model vertices in the key parts of the three-dimensional human body model are verified to obtain corresponding verified skin weights, and the model vertices are bound to the target bones based on the verified skin weights and the remaining skin weights, wherein the remaining skin weights are the target skin weights of the model vertices except the key parts.

2. The method for automatic binding of virtual human according to claim 1, characterized in that: The determining of at least one key part based on the target human body action in which the historical virtual human of the target character type is prone to human body model distortion and the corresponding at least one target vertex specifically includes: Acquire first historical human body movements that occur when human body model distortion occurs in the historical virtual human of the target character type, count the first occurrence times of each of the first historical human body movements, and select a first number of first historical human body movements from each of the first historical human body movements in descending order of the first occurrence times to determine as target human body movements; Acquire historical vertices that are accompanied by abnormalities when a human body model of a historical virtual human is distorted under a single target human body action, count the second number of occurrences of each of the historical vertices, and select a second number of historical vertices from each of the historical vertices in descending order of the second number of occurrences to determine as target vertices for the corresponding target human body action; Determine a first weight for each of the target human body movements, and determine a second weight for each of the target vertices corresponding to each of the target human body movements, wherein the first weight is a ratio of a first occurrence number of each target human body movement to a sum of first occurrence numbers of all target human body movements, and the second weight is a ratio of a second occurrence number of a single target vertex corresponding to the target human body movement to a sum of second occurrence numbers of all corresponding target vertices; At least one key part is determined according to the first weights and the corresponding second weights.

3. The method for automatic binding of virtual human according to claim 2, characterized in that: The determining at least one key part according to the first weight and the corresponding second weights specifically includes: Acquire second historical human body movements that have appeared in the animation of the historical virtual human of the target character type, count the occurrence frequency of each of the second historical human body movements, and select a third number of second historical human body movements from each of the second historical human body movements in descending order of the occurrence frequency to determine as high-frequency human body movements; Determine the high-frequency human body motion existing in each of the target human body motions as a reference human body motion, and calculate a first product of a first weight of each of the reference human body motions and a second weight of each of the corresponding target vertices; Summing the first products corresponding to at least one target vertex in the same human body part to obtain a sum of the corresponding first products; If the sum of the first products exceeds a preset first threshold, the corresponding human body part is determined as a key part.

4. The method for automatic virtual human binding according to claim 2, characterized in that: The method further comprises: Determining at least one group of human body movement combinations of the target virtual person in the actual animation according to the movement design scheme of the target virtual person in the actual animation, wherein the actual animation is an animation produced for the target virtual person; Calculating a second product of a first weight of a target human body action in a single human body action combination and a second weight of each corresponding target vertex, and summing the second products corresponding to the target vertices in the same human body part to obtain a sum of the corresponding second products; If the sum of the second products is greater than a preset first threshold, the corresponding human body part is determined as a warning part, and a target appearance period of the corresponding human body action combination in the actual animation is determined; The playback mode of the target appearance period is set to slow playback, and when playing a single target appearance period, a check reminder for the corresponding alert area is sent to the personnel's terminal. The larger the sum of the second products, the slower the slow playback speed of the corresponding target appearance period.

5. The method for automatic virtual human binding according to claim 1, characterized in that: The method further comprises: Obtaining third historical human body movements when the historical virtual person of the target character type appears physically uncoordinated in the animation, counting the first occurrence frequency of each of the third historical human body movements, and selecting a fourth number of third historical human body movements from each of the third historical human body movements in descending order of the first occurrence frequencies to determine as important human body movements; Obtaining historical part combinations in which the historical virtual human has appeared incoordination under a single important human body action, counting the second occurrence frequency of each of the historical part combinations, and selecting a fifth number of historical part combinations from the historical part combinations in descending order of the second occurrence frequency to determine as important part combinations of the corresponding important human body action, wherein the historical part combinations are human body parts that have been deformed and become incoordination with each other when the historical virtual human performs the important human body action; Determine a third weight of each of the important human body movements, and determine a fourth weight of each of the important part combinations corresponding to each of the key human body movements, wherein the third weight is a ratio of a first occurrence frequency of the important human body movement to a sum of first occurrence frequencies of all the important human body movements, and the fourth weight is a ratio of a second occurrence frequency of a single important part combination corresponding to the important human body movement to a sum of second occurrence frequencies of all the corresponding important part combinations; Determining, according to the third weight and the corresponding fourth weights, whether there is a risk of incoordination in each of the key parts when the target virtual person moves; If so, the verified skin weights of the model vertices in the key parts where there is a risk of inconsistency are adjusted and optimized.

6. The method for automatic virtual human binding according to claim 5, characterized in that: The determining, according to the third weight and the corresponding fourth weights, whether there is an incoordination risk in each of the key parts when the target virtual person moves, specifically includes: If the important part combination includes multiple key parts, the corresponding important part combination is determined as the final part combination, and the important human body movements including the final part combination in each corresponding important part combination are determined as the final human body movements; Calculating a third product of the third weight of each final human body action and the fourth weight of the corresponding final part combination, and summing the third products to obtain a sum of the third products; If the sum of the third products is greater than a preset second threshold, it is determined that there is an incoordination risk in each of the key parts.

7. The method for automatic virtual human binding according to claim 6, characterized in that: If so, the verified skin weights of the model vertices in the key parts where there is a risk of incoordination are adjusted and optimized, specifically including: If so, selecting the largest third product from the third products; The key parts existing in the final part combination corresponding to the maximum third product are determined as the key parts with incoordination risks, and the verified skin weights of the model vertices in the key parts with incoordination risks are adjusted and optimized.

8. A virtual human automatic binding device, characterized in that: include: An information acquisition module (11) is used to acquire a target geometric shape of a three-dimensional human body model of a target virtual person; A skeleton determination module (12) is used to determine the matching skeleton structure corresponding to the target virtual person according to the target geometric form, wherein the matching skeleton structure includes a plurality of target skeletons that need to be bound; A weight allocation module (13) is used to determine at least one target skinning weight corresponding to each model vertex in the three-dimensional human body model based on the trained target machine learning model; A part screening module (14) is used to determine the target role type of the target virtual human, and determine at least one key part based on the target human body movements of the historical virtual human of the target role type that are prone to human body model distortion and at least one corresponding target vertex, wherein the target vertex is a vertex that is prone to model distortion under the corresponding target human body movement, and the key part is a part of the human body model of the virtual human of the target role type that needs to be checked for vertex skin weight; The automatic binding module (15) is used to verify the target skin weights of the model vertices in the key parts of the three-dimensional human body model, obtain corresponding verified skin weights, and bind the model vertices to the target bones based on the verified skin weights and the remaining skin weights, wherein the remaining skin weights are the target skin weights of the model vertices other than the key parts.

9. A computer-readable storage medium having a computer program stored therein, characterized in that: When the computer program is loaded and executed by a processor, the method according to any one of claims 1 to 7 is adopted.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: When the processor loads and executes the computer program, the method according to any one of claims 1 to 7 is adopted.

Images