An AI-based method and system for mapping real-person digital avatars

Abstract

The present application relates to a method and system for mapping a real - person digital avatar based on AI. The method includes obtaining first - image information of a target object from multiple angles, extracting features from the first - image information to obtain the types of clothing worn by the target object and determining a preset clothing model for the clothing type, extracting the clothing texture map and patterns of the target object's clothing, mapping the clothing texture map and patterns onto the corresponding preset clothing model to obtain a candidate clothing model, obtaining video information of the target object moving according to a preset action, determining the pleat videos of the clothing at each joint from the video information, obtaining a mapping video of the target object during real - person mapping, and processing the candidate clothing model based on the pleat videos and the mapping video to obtain the final clothing model of the target object during real - person mapping. The present application can make the changes in clothing pleats, etc., during the movement of a person in reality closer to and more accurate in virtual mapping, improving the virtual mapping effect.

Description

Technical Field

[0001] This application relates to the field of digital twins, and more particularly to a method and system for mapping a real-person digital avatar based on AI. Background Art

[0002] The metaverse will have a profound impact on fields such as office work, gaming, healthcare, social networking, and education. The most important feature is the integration of virtual and real worlds, mapping elements of real life into the virtual world. Currently, mapping a person's daily life image into the virtual world mainly involves collecting images of the person and generating corresponding virtual avatars in the virtual world according to the images. Due to the deformation of clothing caused by a person's actions in reality and other changes, it is not easy to accurately render the same changes in the virtual world as in real life for features such as clothing, resulting in a low level of realism for the person's image in the virtual world. Summary of the Invention

[0003] In order to make the changes in clothing folds and the like of a person in reality closer to accurate during virtual mapping and improve the virtual mapping effect, this application provides a method and system for mapping a real-person digital avatar based on AI.

[0004] In a first aspect, this application provides a method for mapping a real-person digital avatar based on AI, adopting the following technical solution:

[0005] A method for mapping a real-person digital avatar based on AI includes:

[0006] Obtain first image information of a target object from multiple angles;

[0007] Extract the clothing types worn by the target object by performing feature extraction on the first image information;

[0008] Search for a preset clothing model of the clothing type from a preset database, and extract clothing texture maps and patterns on the target object's clothing from the first image information from multiple angles;

[0009] Map the clothing texture maps and patterns onto the corresponding preset clothing model to obtain a candidate clothing model;

[0010] Obtain video information of the target object performing activities according to a preset action;

[0011] Determine the clothing fold videos at each joint from the video information;

[0012] Obtain a mapping video of the target object during real-person mapping, and process the candidate clothing model based on the clothing fold videos and the mapping video to obtain the final clothing model of the target object during real-person mapping.

[0013] By adopting the above technical solution, obtaining the first image information of the target object from multiple angles facilitates subsequent feature extraction of the first image information to obtain the type of clothing worn by the target object. Then, a predicted clothing model corresponding to the clothing type is found from the preset database, and an image on the target object's clothing is extracted from the first image information from multiple angles. The pattern is mapped onto the preset clothing model to obtain a candidate clothing model that is consistent with the clothing worn by the target object in reality. Video information of the target object performing activities according to preset actions is obtained. When the target object performs activities according to preset actions, wrinkles and other deformations will occur at the joints of the clothing. However, in the virtual world, the wrinkles and other deformations at the joints cannot accurately simulate the same effect as in reality. Therefore, the wrinkle videos of the clothing at each joint are determined from the video information. The wrinkle videos record the specific situations of the wrinkles that appear during the activities at each joint. Therefore, a mapping video of the target object during real-person mapping is obtained. The candidate clothing model is processed according to the wrinkle videos and the mapping video to obtain the final clothing model of the target object during real-person virtual space mapping. That is, according to the specific situations of the target object making actions and activities in the mapping video, the manifestation forms of the wrinkles at each joint are determined from the wrinkle videos, and then the final clothing model is obtained. The final clothing model is closer to the joints of the clothing worn by the target object in reality at each joint, thus making the virtual mapping effect better.

[0014] In another possible implementation manner, the step of mapping the clothing texture map and the pattern onto the corresponding preset clothing model to obtain a candidate clothing model includes:

[0015] Map the clothing texture map corresponding to each angle into a preset coordinate system to determine the coordinate points of each position on the pattern;

[0016] Obtain the second image information of the preset clothing model at the multiple angles, and perform overlapping mapping on the first image information and the second image information at the same angle;

[0017] Map the pattern into each second image information according to the coordinate points of the pattern at each angle to obtain the mapped second image information;

[0018] Map the mapped second image information back onto the preset clothing model to obtain the final clothing model.

[0019] In another possible implementation manner, the step of determining the wrinkle videos of the clothing at each joint from the video information includes:

[0020] Determine a preset range corresponding to each joint based on the position of the joint;

[0021] Perform feature recognition on the video information to obtain the feature points of each joint;

[0022] Taking the feature point as the center point, an interception area is obtained according to a preset range, and feature tracking is performed on each joint. During the feature tracking process, the video information is intercepted according to the interception area to obtain the pleat video of the clothing at each joint.

[0023] In another possible implementation manner, processing the candidate clothing model based on the pleat video and the mapping video to obtain the final clothing model of the target object during real-person mapping includes:

[0024] Determining first pleat features at each bending angle from the pleat video corresponding to each joint;

[0025] Performing rotation simulation on the first pleat features at each bending angle to obtain second pleat features when the pleat features at each bending angle are rotated to each angle. Each second pleat feature corresponds to a bending angle and a rotation angle;

[0026] Performing action recognition on each frame of the mapping video to obtain the bending angle formed at each joint in each frame of the picture and the relative rotation angle of the limbs on both sides of the joint;

[0027] Based on the bending angle formed at each joint in each frame of the picture and the relative rotation angle of the limbs on both sides of the joint, searching for the target pleat features corresponding to each joint from the second pleat features of each joint;

[0028] Mapping the target pleat features corresponding to each joint to the corresponding positions on the candidate clothing model to obtain the final clothing model.

[0029] In another possible implementation manner, the method further includes:

[0030] Determining third image information at each joint in each frame of the picture from the mapping video. Each third image information corresponds to the bending angle formed at the corresponding joint and the relative rotation angle of the limbs on both sides of the joint;

[0031] Searching for associated second pleat features with the same bending angle and rotation angle as the third image information;

[0032] Calculating a first similarity between the associated second pleat feature and the third image information;

[0033] If the first similarity does not reach a preset similarity threshold, adjusting the associated second pleat feature based on the third image information.

[0034] In another possible implementation manner, adjusting the associated second pleat feature based on the third image information includes:

[0035] Perform texture extraction on the third image information and the second wrinkle feature respectively to obtain a first texture map for the third image information and a second texture map for the second wrinkle feature;

[0036] Perform overlapping mapping on the first texture map and the second texture map to determine the non-overlapping texture;

[0037] Determine the gray value of the area corresponding to the non-overlapping texture from the third image information;

[0038] Adjust the associated second wrinkle feature according to the gray value and the non-overlapping texture to obtain the adjusted second wrinkle feature.

[0039] In another possible implementation manner, the method further includes:

[0040] Determine each frame of the mapping video and perform edge detection to obtain the fourth image information of the clothing;

[0041] Calculate the second similarity between the fourth image information and the final clothing model corresponding to each frame of the video;

[0042] Perform an averaging process on the second similarities of each frame of the mapping video to obtain the average value of the second similarities;

[0043] Output the average value of the second similarities.

[0044] In a second aspect, the present application provides an AI-based real-person digital avatar mapping system, adopting the following technical solutions:

[0045] An AI-based real-person digital avatar mapping system, comprising:

[0046] An image acquisition module, configured to acquire first image information of a target object from multiple angles;

[0047] A first extraction module, configured to perform feature extraction on the first image information to obtain the clothing types worn by the target object;

[0048] A second extraction module, configured to find the preset clothing models of the clothing types from a preset database, and extract the clothing texture maps and the patterns on the clothing of the target object from the first image information of the multiple angles;

[0049] A mapping module, configured to map the clothing texture maps and patterns to the corresponding preset clothing models to obtain candidate clothing models;

[0050] A video acquisition module, configured to acquire video information of the target object performing activities according to preset actions;

[0051] A video determination module, configured to determine the pleat videos of the clothing at each joint from the video information;

[0052] A processing module, configured to obtain a mapping video of the target object during real-person mapping, and process a candidate clothing model based on the pleat videos and the mapping video to obtain the final clothing model of the target object during real-person mapping.

[0053] By adopting the above technical solution, the image acquisition module acquires the first image information of the target object from multiple angles, which facilitates the subsequent first extraction module to extract features from the first image information to obtain the clothing types worn by the target object. Then, a predicted clothing model corresponding to the clothing type is found from the preset database, and the second extraction module extracts the images on the target object's clothing from the first image information from multiple angles. The mapping module maps the pattern onto the preset clothing model to obtain a candidate clothing model that is consistent with the clothing worn by the target object in reality. The video acquisition module acquires the video information of the target object moving according to a preset action. When the target object moves according to the preset action, the clothing at each joint will have deformations such as pleats, but in the virtual world, the deformations such as pleats at the joints cannot accurately simulate the same effect as in reality. Therefore, the video determination module determines the pleat videos of the clothing at each joint from the video information. The pleat videos record the specific situations of the pleats that appear during the activities at each joint. Therefore, the processing module acquires the mapping video of the target object during real-person mapping, and processes the candidate clothing model according to the pleat videos and the mapping video to obtain the final clothing model of the target object during real-person virtual space mapping, that is, determines the manifestation forms of the pleats at each joint from the pleat videos according to the specific situations of the actions and activities made by the target object in the mapping video, and then obtains the final clothing model. The final clothing model is closer to the joints of the clothing worn by the target object in reality at each joint, thus making the virtual mapping effect better.

[0054] In another possible implementation manner, when the mapping module maps the clothing texture map and the pattern onto the corresponding preset clothing model to obtain a candidate clothing model, it is specifically configured to:

[0055] Map the clothing texture map corresponding to each angle into a preset coordinate system to determine the coordinate points of each position on the pattern;

[0056] Obtain the second image information of the preset clothing model at the multiple angles, and perform overlapping mapping on the first image information and the second image information at the same angle;

[0057] Map the pattern into each second image information according to the coordinate points corresponding to the pattern at each angle to obtain the mapped second image information;

[0058] Reflect the mapped second image information onto the preset clothing model to obtain the final clothing model.

[0059] In another possible implementation, when the video determination module determines the pleat videos of the clothing at each joint from the video information, it specifically is used for:

[0060] Determine a preset range corresponding to each joint based on the position of the joint;

[0061] Perform feature recognition on the video information to obtain the feature points of each joint;

[0062] Use the feature points as the center points to obtain an intercepted area according to the preset range, perform feature tracking on each joint, and intercept the video information according to the intercepted area during the feature tracking process to obtain the pleat videos of the clothing at each joint.

[0063] In another possible implementation, when the processing module processes the candidate clothing model based on the pleat videos and the mapping video to obtain the final clothing model of the target object during real-person mapping, it specifically is used for:

[0064] Determine the first pleat features at each bending angle from the pleat videos corresponding to each joint;

[0065] Perform rotation simulation on the first pleat features at each bending angle to obtain the second pleat features when the pleat features at each bending angle are rotated to each angle. Each second pleat feature corresponds to a bending angle and a rotation angle;

[0066] Perform action recognition on each frame of the mapping video to obtain the bending angles formed at each joint in each frame of the video and the relative rotation angles of the limbs on both sides of the joint;

[0067] Based on the bending angles formed at each joint in each frame of the video and the relative rotation angles of the limbs on both sides of the joint, search for the target pleat features corresponding to each joint from the second pleat features of each joint;

[0068] Map the target pleat features corresponding to each joint to the corresponding positions on the candidate clothing model to obtain the final clothing model.

[0069] In another possible implementation, the AI-based real-person digital avatar mapping system further includes:

[0070] An image determination module, configured to determine the third image information at each joint in each frame of the mapping video. Each third image information corresponds to the bending angle formed at the corresponding joint and the relative rotation angle of the limbs on both sides of the joint;

[0071] A search module for searching for associated second fold features whose bending angle and rotation angle are consistent with the third image information;

[0072] A first calculation module for calculating a first similarity between the associated second fold feature and the third image information;

[0073] An adjustment module for adjusting the associated second fold feature based on the third image information when the first similarity does not reach a preset similarity threshold.

[0074] In another possible implementation manner, when the adjustment module adjusts the associated second fold feature based on the third image information, it specifically is used for:

[0075] Performing texture extraction on the third image information and the second fold feature respectively to obtain a first texture map of the third image information and a second texture map of the second fold feature;

[0076] Overlapping and mapping the first texture map and the second texture map to determine the non-overlapping texture;

[0077] Determining the gray value of the area corresponding to the non-overlapping texture from the third image information;

[0078] Adjusting the associated second fold feature according to the gray value and the non-overlapping texture to obtain an adjusted second fold feature.

[0079] In another possible implementation manner, the AI-based real-person digital avatar mapping system further includes:

[0080] An edge detection module for performing edge detection on each frame of the mapping video to determine fourth image information of the clothing;

[0081] A second calculation module for calculating a second similarity between the fourth image information and the final clothing model corresponding to each frame of the video;

[0082] A third calculation module for averaging the second similarities of each frame of the mapping video to obtain an average value of the second similarities;

[0083] An output module for outputting the average value of the second similarities.

[0084] In a third aspect, the present application provides an electronic device, adopting the following technical solution:

[0085] An electronic device, which includes:

[0086] At least one processor;

[0087] A memory;

[0088] At least one application program, where the at least one application program is stored in a memory and configured to be executed by at least one processor, and the at least one is configured to: execute an AI-based real-person digital avatar mapping method shown in any possible implementation manner of the first aspect.

[0089] In a fourth aspect, the present application provides a computer-readable storage medium, adopting the following technical solution:

[0090] A computer-readable storage medium, when the computer program is executed in a computer, causes the computer to execute an AI-based real-person digital avatar mapping method described in any item of the first aspect.

[0091] In summary, the present application includes at least one of the following beneficial technical effects:

[0092] Obtaining the first image information of the target object from multiple angles facilitates subsequent feature extraction of the first image information to obtain the types of clothing worn by the target object. Then, find the predicted clothing model corresponding to the clothing type from a preset database, and extract the image on the target object's clothing from the first image information from multiple angles, and map the pattern onto the preset clothing model to obtain a candidate clothing model that is consistent with the clothing worn by the target object in reality. Obtain the video information of the target object performing activities according to a preset action. When the target object performs activities according to the preset action, the clothing at each joint will have deformations such as wrinkles. However, in the virtual world, the deformations such as wrinkles at the joints cannot accurately simulate the same effect as in reality. Therefore, determine the wrinkle videos of the clothing at each joint from the video information. The wrinkle videos record the specific situations of the wrinkles that appear when each joint moves. Therefore, obtain the mapping video of the target object during real-person mapping, and process the candidate clothing model according to the wrinkle videos and the mapping video to obtain the final clothing model of the target object during real-person virtual space mapping, that is, determine the manifestation forms of the wrinkles at each joint from the wrinkle videos according to the specific situations of the target object's actions and activities in the mapping video, and then obtain the final clothing model. The final clothing model is closer to the joints of the clothing worn by the target object in reality at each joint, thus making the virtual mapping effect better. Description of the Drawings

[0093] Figure 1 is a flowchart of an AI-based real-person digital avatar mapping method according to an embodiment of the present application.

[0094] Figure 2 is a structural diagram of an AI-based real-person digital avatar mapping system according to an embodiment of the present application.

[0095] Figure 3 is a structural diagram of an electronic device according to an embodiment of the present application. Detailed implementation mode

[0096] The following further elaborates on this application in conjunction with the accompanying drawings.

[0097] After reading this specification, those skilled in the art can make modifications to this embodiment without creative contributions as needed, but as long as they are within the scope of the claims of this application, they are protected by the patent law.

[0098] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, rather than all of them. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of this application.

[0099] In addition, the term "and / or" in this article is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after, unless otherwise specified.

[0100] The following further describes the embodiments of this application in detail in conjunction with the drawings of the specification.

[0101] The embodiments of this application provide a method for mapping a real-person digital avatar based on AI, which is executed by an electronic device. The electronic device can be a server or a terminal device. Among them, the server can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal device can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto. The terminal device and the server can be directly or indirectly connected through wired or wireless communication methods, and this application does not limit this here. As Figure 1 shown, the method includes step S101, step S102, step S103, step S104, step S105, step S106, and step S107, where

[0102] S101, obtain first image information of the target object from multiple angles.

[0103] For the embodiments of the present application, the electronic device is connected to the camera device through a wire. Multiple preset voice commands can be stored in the electronic device, and each preset voice command represents a guidance for instructing the target object to rotate a certain angle in place, such as "rotate 45° clockwise in place", "rotate 90° clockwise in place", etc. The electronic device outputs the preset voice command, and the target object rotates under the guidance of the preset voice command, so that the camera device can collect the first image information of the target object at multiple angles, and further enable the electronic device to obtain the first image information at multiple angles. The multiple first image information records the relevant information of the clothes worn by the target object, such as patterns, etc.

[0104] S102, perform feature extraction on the first image information to obtain the type of clothing worn by the target object.

[0105] For the embodiments of the present application, the electronic device inputs the first image information at multiple angles into a trained network model for clothing type recognition, such as identifying a hooded sweatshirt, jeans, etc. worn by the target object. Specifically, the network model can be a convolutional neural network model, a recurrent neural network model, or other types of network models, which are not limited here.

[0106] S103, find the preset clothing model of the clothing type from the preset database, and extract the clothing map and the pattern on the clothing of the target object from the first image information at multiple angles.

[0107] For the embodiments of the present application, the preset database stores preset clothing models of different types of clothing, and each preset clothing model is pre-modeled and stored by relevant personnel. Each preset clothing model corresponds to a clothing type, so the electronic device can find the preset clothing model in the preset database according to the determined clothing type of the target object. In order to make the clothing model in the virtual world more consistent with the clothing in reality, the electronic device extracts the clothing map and pattern of the target object's clothing from the first image information at multiple angles. Specifically, the electronic device can perform edge detection on the first image information at multiple angles. First, perform denoising processing on the first image information, then perform gray-scale transformation on the denoised first image information to obtain a gray-scale image, then determine the edge between the clothing and the environment by determining the positions where the gray-scale values in the gray-scale image undergo steps, then determine the image within the edge range as the clothing map, and then perform edge detection on the foreign object map to obtain the pattern in the clothing map.

[0108] S104, map the clothing map and the pattern to the corresponding preset clothing model to obtain a candidate clothing model.

[0109] For the embodiments of the present application, after the electronic device determines the clothing texture map and pattern of the target object's clothing, it can map the clothing texture map and pattern onto the corresponding preset clothing model to obtain the candidate clothing model.

[0110] S105. Obtain video information of the target object performing activities according to preset actions.

[0111] For the embodiments of the present application, the camera device is also used to collect video information when the target object performs activities according to preset actions. Multiple preset action videos can also be stored in the electronic device. The electronic device controls the display device to output and display the preset action videos, so that the target object makes corresponding actions according to the preset action videos, such as raising hands, bending down, rotating arms, squatting and other actions. The camera device collects the video information when the target object performs the above activities and the electronic device can obtain it.

[0112] S106. Determine the wrinkle videos of the clothing at each joint from the video information.

[0113] For the embodiments of the present application, after the electronic device obtains the video information, it determines the wrinkle videos of the clothing at each joint from the video information, which is convenient for further rendering and adjustment of the candidate clothing model according to the wrinkle videos later and reduces the data volume.

[0114] S107. Obtain the mapping video when the target object performs real-person mapping, and process the candidate clothing model based on the wrinkle videos and the mapping video to obtain the final clothing model when the target object performs real-person mapping.

[0115] For the embodiments of the present application, when the target object starts to perform virtual mapping in the metaverse, the electronic device obtains the video of the target object collected by the camera device, that is, the mapping video. The mapping video records the specific situation of the target object performing activities. Therefore, the electronic device can determine the activity situation of each joint from the mapping video, and then determine the wrinkles that need to be formed at each joint in the virtual world according to the wrinkle videos. Then, map the wrinkles that need to be formed onto the candidate clothing model to obtain the final clothing model that is consistent with the clothing performance in reality. The final clothing model is closer to the joints of the target object wearing clothing in reality at each joint, so that the effect of virtual mapping is better.

[0116] In a possible implementation manner of the embodiments of the present application, in step S104, mapping the clothing texture map and pattern onto the corresponding preset clothing model to obtain the candidate clothing model specifically includes step S1041 (not shown in the figure), step S1042 (not shown in the figure), step S1043 (not shown in the figure), and step S1044 (not shown in the figure), where

[0117] S1041. Map the clothing texture map corresponding to each angle into a preset coordinate system to determine the coordinate points of each position on the pattern.

[0118] For the embodiments of the present application, the electronic device maps the clothing texture maps of each angle into a preset coordinate system respectively. The preset coordinate system is a plane rectangular coordinate system. Then, the coordinate points of the clothing texture maps of each angle are determined. Then, the electronic device maps the pattern of each angle into the preset coordinate system to obtain the coordinate points of each position on the pattern.

[0119] S1042. Obtain the second image information of the preset clothing model at multiple angles, and perform overlapping mapping on the first image information and the second image information at the same angle.

[0120] For the embodiments of the present application, the electronic device rotates the preset clothing model through relevant simulation software and takes screenshots to obtain the second image information of the preset clothing model at multiple angles. The second image information is also the clothing texture maps of the preset clothing model at multiple angles. The multiple angles of the second image information are the same as those of the first image information. Then, the electronic device performs overlapping mapping on the first image information and the second image information at the same angle to facilitate more accurately determining the position of the pattern on the preset clothing model subsequently.

[0121] S1043. Map the pattern into each second image information according to the coordinate points corresponding to the pattern at each angle to obtain the mapped second image information.

[0122] For the embodiments of the present application, the electronic device maps the pattern into the second image information of the corresponding angle according to the coordinate points corresponding to the pattern at each angle to obtain the mapped second image information of the pattern, that is, the mapped clothing texture map.

[0123] S1044. Reflect the mapped second image information onto the preset clothing model to obtain the final clothing model.

[0124] For the embodiments of the present application, the electronic device reflects the mapped second image information onto the preset clothing model to obtain the final clothing model. It is more accurate to obtain the final preset clothing model by determining the coordinate points of the pattern and performing operations such as mapping the first image information and the second image information.

[0125] A possible implementation manner of the embodiments of the present application. In step S106, determining the pleat videos of the clothing at each joint from the video information specifically includes step S1061, step S1062, and step S1063, where

[0126] S1061. Determine the preset range corresponding to each joint based on the position of the joint.

[0127] For the embodiments of the present application, the electronic device determines a preset range corresponding to each joint according to the size and position of the joint. The preset range is rectangular. For example, the volume and area of the elbow joint are smaller than those of the knee joint. Therefore, the preset range of the elbow joint is smaller than that of the knee joint. The preset range corresponding to each joint is pre-stored in the electronic device. Therefore, after the electronic device determines the position of the joint, such as the elbow joint, it can find the preset range corresponding to the elbow joint.

[0128] S1062. Perform feature recognition on the video information to obtain the feature points of each joint.

[0129] For the embodiments of the present application, the electronic device inputs the video information into a trained network model for feature recognition to obtain the feature points of each joint. For example, the bending points of the elbow joint and the knee joint are both feature points. The network model can be a convolutional neural network model, a recurrent neural network model, or other network models.

[0130] S1063. Use the feature points as the center points to obtain an intercepted area according to the preset range, perform feature tracking on each joint, and intercept the video information according to the intercepted area during the feature tracking process to obtain the wrinkled video of the clothing at each joint.

[0131] For the embodiments of the present application, the electronic device determines the center point of the preset range with the feature points, thereby forming a video interception area. The electronic device performs feature tracking on the feature points of each joint and intercepts the video information according to the intercepted area during the tracking process to obtain the wrinkled video of the clothing at each joint.

[0132] In a possible implementation manner of the embodiments of the present application, in step S107, the candidate clothing model is processed based on the wrinkled video and the mapping video to obtain the final clothing model of the target object during real-person mapping, specifically including step S1071 (not shown in the figure), step S1072 (not shown in the figure), step S1073 (not shown in the figure), step S1074 (not shown in the figure), and step S1075 (not shown in the figure), where

[0133] S1071. Determine the first wrinkle features at each bending angle from the wrinkled video corresponding to each joint.

[0134] For the embodiments of the present application, the electronic device determines the vertex of the bending angle at the joint bending part in the wrinkled video corresponding to each joint, then simplifies the limbs on both sides of the bending part into straight lines and connects the vertices to obtain an angle. Since the wrinkled video is the dynamic performance of the target object during movement, the electronic device determines the angle value of the obtained angle, thereby determining the display effect of the wrinkles at each joint at each bending angle, that is, the first wrinkle feature.

[0135] S1072. Perform rotational simulation on the first fold feature at each bending angle to obtain the second fold feature when the fold feature at each bending angle rotates to each angle.

[0136] Among them, each second fold feature corresponds to a bending angle and a rotation angle.

[0137] For the embodiments of this application, when the target object moves in reality, during the bending process of each joint, there may also be relative rotational movement between the limbs on both sides of the bending part, and the relative rotational movement of the two sides of the limbs will have deformation effects such as stretching on the folds at the current angle. Therefore, the electronic device performs rotational simulation on the first fold feature at each bending angle through a convolutional neural network or related analysis software to obtain the second fold feature when the fold feature at each bending angle rotates to each angle. Through rotational simulation, the folds can be made more realistic.

[0138] S1073. Perform action recognition on each frame of the mapped video to obtain the bending angle formed at each joint and the relative rotation angle of the limbs on both sides of the joint in each frame of the video.

[0139] For the embodiments of this application, the electronic device extracts each frame of the video from the mapped video, and then extracts the posture of the target object in each frame of the video. Specifically, the electronic device inputs each frame of the video into a trained network model for action recognition to obtain the bending angle formed at each joint and the relative rotation angle on both sides of the joint in each frame of the video.

[0140] S1074. Based on the bending angle formed at each joint and the relative rotation angle of the limbs on both sides of the joint in each frame of the video, search for the target fold feature corresponding to each joint from the second fold features of each joint.

[0141] For the embodiments of this application, after the electronic device determines the bending angle and relative rotation angle formed at each joint, the consistent target fold feature can be searched from the second fold features of each joint according to these two angle values.

[0142] S1075. Map the target fold feature corresponding to each joint to the corresponding position on the candidate clothing model to obtain the final clothing model.

[0143] For the embodiments of this application, after the electronic device determines the target fold feature of each joint, mapping the target fold feature to the corresponding position on the candidate clothing model can obtain the final clothing model corresponding to each frame of the video, thereby achieving a more realistic mapping between the metaverse reality and the virtual.

[0144] In a possible implementation manner of the embodiment of the present application, after step S107, steps Sa (not shown in the figure), step Sb (not shown in the figure), step Sc (not shown in the figure), and step Sd (not shown in the figure) are further included, where

[0145] Sa, determine the third image information at each joint in each frame of the mapped video, and each third image information corresponds to the bending angle formed at the corresponding joint and the relative rotation angle of the limbs on both sides of the joint.

[0146] For the embodiment of the present application, the electronic device can intercept the images of the specific manifestations in reality at each joint from each frame of the mapped video according to the intercepted area recorded in step S1063, that is, the third image information. Then the electronic device performs feature recognition on the third image information to obtain the bending angle at each joint and the relative rotation angle of the limbs on both sides of the joint. The manner in which the electronic device determines the bending angle and the relative rotation angle can be determined by referring to the manner recorded in step S1071 and step S1072.

[0147] Sb, search for the associated second fold feature whose bending angle and rotation angle are consistent with the third image information.

[0148] For the embodiment of the present application, the electronic device searches for the consistent associated second fold feature from the second fold features corresponding to each joint according to the bending angle and the relative rotation angle corresponding to the third image information at each joint.

[0149] Sc, calculate the first similarity between the associated second fold feature and the third image information.

[0150] For the embodiment of the present application, the electronic device can calculate the cosine distance value between the associated second fold feature and the third image information, and represent the first similarity by the cosine distance value, that is, represent the picture as a vector, and represent the similarity between two pictures by calculating the cosine distance between the vectors. The first similarity can also be calculated through a histogram, or calculated in other ways, or the electronic device inputs the second fold feature and the third image information into a trained network model for similarity calculation to obtain the first similarity. This is not limited herein.

[0151] Sd, if the first similarity does not reach the preset similarity threshold, adjust the associated second fold feature based on the third image information.

[0152] For the embodiments of the present application, the preset similarity threshold serves as a demarcation point for whether the associated second fold feature is relatively close to the third image information. The electronic device compares the determined first similarity with the preset similarity threshold. If the first similarity does not reach the preset similarity threshold, it indicates that the simulated associated second feature has a large difference from the joint bending performance in reality, and the simulation effect is poor. Therefore, the associated second fold feature is adjusted according to the third image information representing the joint bending in reality, so that the second fold feature used in subsequent virtual simulations of each joint is more realistic.

[0153] In a possible implementation manner of the embodiments of the present application, adjusting the associated second fold feature based on the third image information in step Sd specifically includes step S1 (not shown in the figure), step S2 (not shown in the figure), step S3 (not shown in the figure), and step S4 (not shown in the figure), where

[0154] S1, perform texture extraction on the third image information and the second fold feature respectively to obtain a first texture map of the third image information and a second texture map of the second fold feature.

[0155] For the embodiments of the present application, the electronic device performs preprocessing operations on the third image information and the second fold feature, such as grayscale conversion, denoising, etc. Extract texture features from the preprocessed images. For example, use spatial filters such as Gabor filters and Laws filters to extract the texture information of the images. Based on statistical methods (such as gray-level co-occurrence matrix GLCM, gray-level gradient histogram GLDH), frequency-based methods (such as Fourier transform, wavelet transform), etc. Quantify the extracted texture features and encode them in vector or matrix form as needed. For example, histograms or vectors can be used to represent the texture features. Classify or recognize the images according to the extracted texture features. Various machine learning algorithms (such as support vector machine SVM, K-nearest neighbor algorithm KNN, neural network, etc.) can be used to implement texture classification and recognition tasks. After completing texture extraction and classification, post-processing (such as smoothing, threshold segmentation, etc.) can be performed on the results as needed, and the results can be displayed in a visual manner.

[0156] S2, perform overlapping mapping on the first texture map and the second texture map to determine the non-overlapping texture.

[0157] For the embodiments of the present application, the electronic device overlaps and maps the first texture map and the second texture map into a preset coordinate system, thereby determining the coordinates of each pixel point, and then determining the pixel points with non-coincident coordinate points. The pixel points with non-coincident coordinate points form the non-overlapping texture. The non-overlapping texture is the area where the folds are inconsistent.

[0158] S3, determine the gray value of the area corresponding to the non-overlapping texture from the third image information.

[0159] For the embodiments of the present application, the electronic device performs grayscale transformation on the preprocessed third image information to obtain a grayscale image, and then the electronic device maps the non-overlapping texture into the grayscale image to obtain the grayscale value of the area corresponding to the non-overlapping texture, and the magnitude of the grayscale value characterizes the light and shadow display effect at the wrinkle.

[0160] S4. Adjust the associated second wrinkle feature according to the grayscale value and the non-overlapping texture to obtain the adjusted second wrinkle feature.

[0161] For the embodiments of the present application, the electronic device replaces the grayscale value of the area corresponding to the associated second feature with the grayscale value of the non-overlapping area, and then replaces the grayscale value of the area corresponding to the associated second feature with the non-overlapping texture to obtain the adjusted second wrinkle feature. The associated second wrinkle feature is adjusted by the third image information characterizing the real wrinkle in reality, so as to optimize the second wrinkle feature and make the simulation of the second wrinkle feature more accurate.

[0162] In a possible implementation manner of the embodiments of the present application, after step S107, there are further included step S108 (not shown in the figure), step S109 (not shown in the figure), step S110 (not shown in the figure), and step S111 (not shown in the figure), where

[0163] S108. Determine each frame of the mapping video and perform edge detection to obtain the fourth image information of the clothing.

[0164] For the embodiments of the present application, the electronic device performs edge detection on each frame of the video to obtain the fourth image information of the clothing in each frame. Specifically, the electronic device performs denoising processing on each frame of the video and then performs grayscale transformation on the denoised frame of the video to obtain a grayscale image, and divides the grayscale image at the place where the grayscale value has a step to obtain the fourth image information of the clothing area.

[0165] S109. Calculate the second similarity between the fourth image information and the final clothing model corresponding to each frame of the video.

[0166] For the embodiments of the present application, the electronic device determines the final clothing model corresponding to the same time point in the virtual simulation according to the time point of the fourth image information, and performs operations such as taking a screenshot of the final clothing model at this moment to obtain a picture of the final clothing model at this moment. Then, the electronic device calculates the second similarity according to the method described in step S1. The higher the second similarity, the closer the simulation is to the performance of the clothing in reality.

[0167] S110. Perform an averaging process on the second similarities of each frame of the mapping video to obtain an average value of the second similarities.

[0168] For the embodiments of the present application, after the electronic device determines the second similarity of each frame of the mapped video during the entire mapping process, it calculates the average value of the second similarity using the average value calculation formula, and uses the average value of the second similarity to characterize the overall proximity level of the virtual simulation to the clothing performance in reality. The larger the average value of the second similarity, the more realistic the virtual simulation during the entire process.

[0169] S111, output the average value of the second similarity.

[0170] For the embodiments of the present application, after the electronic device determines the average value of the second similarity, it can send the average value of the second similarity to the terminal device of the target object or the terminal device of the management personnel in the form of a text message, so as to facilitate the target object and the management personnel to intuitively know the level of the virtual simulation effect in a timely manner, and further facilitate subsequent improvement.

[0171] The above embodiments introduce a method for mapping a real-person digital avatar based on AI from the perspective of the method flow. The following embodiments introduce a system 20 for mapping a real-person digital avatar based on AI from the perspective of virtual modules or virtual units. For details, see the following embodiments.

[0172] The embodiments of the present application provide a system 20 for mapping a real-person digital avatar based on AI, as Figure 2 shown, a system 20 for mapping a real-person digital avatar based on AI may specifically include:

[0173] An image acquisition module 201, configured to acquire first image information of the target object from multiple angles;

[0174] A first extraction module 202, configured to perform feature extraction on the first image information to obtain the types of clothing worn by the target object;

[0175] A second extraction module 203, configured to find a preset clothing model of the clothing type from a preset database, and extract clothing texture maps and patterns on the target object's clothing from the first image information from multiple angles;

[0176] A mapping module 204, configured to map the clothing texture maps and patterns to the corresponding preset clothing model to obtain a candidate clothing model;

[0177] A video acquisition module 205, configured to acquire video information of the target object performing activities according to a preset action;

[0178] A video determination module 206, configured to determine the pleat video of the clothing at each joint from the video information;

[0179] The processing module 207 is configured to obtain the mapping video of the target object during real-person mapping, and process the candidate clothing model based on the wrinkle video and the mapping video to obtain the final clothing model of the target object during real-person mapping.

[0180] An embodiment of the present application discloses an AI-based real-person digital avatar mapping system 20. Among them, the image acquisition module 201 acquires the first image information of the target object from multiple angles to facilitate the subsequent feature extraction of the first image information by the first extraction module 202 to obtain the clothing type worn by the target object, and then searches for the predicted clothing model corresponding to the clothing type from the preset database. And the second extraction module 203 extracts the images on the target object's clothing from the first image information from multiple angles. The mapping module 204 maps the pattern onto the preset clothing model to obtain a candidate clothing model that is consistent with the clothing worn by the target object in reality. The video acquisition module 205 acquires the video information of the target object performing activities according to the preset actions. When the target object performs activities according to the preset actions, wrinkles and other deformations will occur in the clothing at each joint. However, in the virtual world, the wrinkles and other deformations at the joints cannot accurately simulate the same effect as in reality. Therefore, the video determination module 206 determines the wrinkle video of the clothing at each joint from the video information. The wrinkle video records the specific situation of the wrinkles that appear when each joint moves. Therefore, the processing module 207 acquires the mapping video of the target object during real-person mapping, and processes the candidate clothing model according to the wrinkle video and the mapping video to obtain the final clothing model of the target object during real-person virtual space mapping, that is, according to the specific situation of the target object making actions and activities in the mapping video, determines the manifestation form of the wrinkles at each joint from the wrinkle video to obtain the final clothing model. The final clothing model is closer to the joints of the clothing worn by the target object in reality at each joint, thus making the virtual mapping effect better.

[0181] A possible implementation manner of the embodiment of the present application is that when the mapping module 204 maps the clothing texture map and the pattern onto the corresponding preset clothing model to obtain the candidate clothing model, it is specifically configured to:

[0182] Map the clothing texture map corresponding to each angle into the preset coordinate system to determine the coordinate points of each position on the pattern;

[0183] Acquire the second image information of the preset clothing model from multiple angles, and overlap and map the first image information and the second image information at the same angle;

[0184] Map the pattern into each second image information according to the coordinate points corresponding to the pattern at each angle to obtain the mapped second image information;

[0185] Reflect the mapped second image information onto the preset clothing model to obtain the final clothing model.

[0186] In a possible implementation of the embodiment of the present application, when the video determination module 206 determines the wrinkled video of the clothing at each joint from the video information, it is specifically used for:

[0187] Determine a preset range corresponding to each joint based on the position of the joint;

[0188] Perform feature recognition on the video information to obtain the feature points of each joint;

[0189] Use the feature points as the center points to obtain an intercepted area according to the preset range, perform feature tracking on each joint, and intercept the video information according to the intercepted area during the feature tracking process to obtain the wrinkled video of the clothing at each joint.

[0190] In a possible implementation of the embodiment of the present application, when the processing module 207 processes the candidate clothing model based on the wrinkled video and the mapping video to obtain the final clothing model of the target object during real-person mapping, it is specifically used for:

[0191] Determine the first wrinkle features at each bending angle from the wrinkled video corresponding to each joint;

[0192] Perform rotation simulation on the first wrinkle features at each bending angle to obtain the second wrinkle features when the wrinkle features at each bending angle are rotated to each angle. Each second wrinkle feature corresponds to a bending angle and a rotation angle;

[0193] Perform action recognition on each frame of the mapping video to obtain the bending angle formed at each joint in each frame of the video and the relative rotation angle of the limbs on both sides of the joint;

[0194] Based on the bending angle formed at each joint in each frame of the video and the relative rotation angle of the limbs on both sides of the joint, find the target wrinkle features corresponding to each joint from the second wrinkle features of each joint;

[0195] Map the target wrinkle features corresponding to each joint to the corresponding positions on the candidate clothing model to obtain the final clothing model.

[0196] In a possible implementation of the embodiment of the present application, an AI-based real-person digital avatar mapping system 20 further includes:

[0197] An image determination module, configured to determine the third image information at each joint in each frame of the mapping video. Each third image information corresponds to the bending angle formed at the corresponding joint and the relative rotation angle of the limbs on both sides of the joint;

[0198] A search module, configured to search for the associated second wrinkle features whose bending angle and rotation angle are consistent with the third image information;

[0199] A first calculation module, configured to calculate a first similarity between the associated second fold feature and the third image information;

[0200] An adjustment module, configured to adjust the associated second fold feature based on the third image information when the first similarity does not reach a preset similarity threshold.

[0201] In a possible implementation manner of the embodiment of the present application, when the adjustment module adjusts the associated second fold feature based on the third image information, it is specifically configured to:

[0202] Extract textures from the third image information and the second fold feature respectively to obtain a first texture map of the third image information and a second texture map of the second fold feature;

[0203] Overlap and map the first texture map and the second texture map to determine the non-overlapping texture;

[0204] Determine the gray value of the area corresponding to the non-overlapping texture from the third image information;

[0205] Adjust the associated second fold feature according to the gray value and the non-overlapping texture to obtain the adjusted second fold feature.

[0206] In a possible implementation manner of the embodiment of the present application, an AI-based real-person digital avatar mapping system 20 further includes:

[0207] An edge detection module, configured to perform edge detection on each frame of the mapping video to obtain fourth image information of the clothing;

[0208] A second calculation module, configured to calculate a second similarity between the fourth image information and the final clothing model corresponding to each frame of the video;

[0209] A third calculation module, configured to perform an averaging process on the second similarities of each frame of the mapping video to obtain an average value of the second similarities;

[0210] An output module, configured to output the average value of the second similarities.

[0211] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the above-described AI-based real-person digital avatar mapping system 20 can refer to the corresponding process in the foregoing method embodiment, and will not be elaborated herein.

[0212] In the embodiment of the present application, an electronic device is provided, as Figure 3 shown Figure 3The electronic device 30 shown includes: a processor 301 and a memory 303. Among them, the processor 301 and the memory 303 are connected, such as connected through a bus 302. Optionally, the electronic device 30 may further include a transceiver 304. It should be noted that in practical applications, the transceiver 304 is not limited to one, and the structure of the electronic device 30 does not constitute a limitation to the embodiments of the present application.

[0213] The processor 301 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in connection with the disclosure of the present application. The processor 301 may also be a combination that implements computing functions, such as a combination including one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0214] The bus 302 may include a path for transmitting information between the above components. The bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 302 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0215] The memory 303 can be a ROM (Read Only Memory), or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory), or other types of dynamic storage devices that can store information and instructions. It can also be an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0216] The memory 303 is used to store the application program code for executing the solution of this application, and is controlled by the processor 301 for execution. The processor 301 is used to execute the application program code stored in the memory 303 to implement the content shown in the foregoing method embodiments.

[0217] Among them, the electronic device includes but is not limited to: mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), PMPs (Portable Multimedia Players), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. It can also be a server, etc. Figure 3 The shown electronic device is only an example and should not impose any restrictions on the functions and usage scope of the embodiments of this application.

[0218] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When it runs on a computer, the computer can execute the corresponding content in the foregoing method embodiment. Compared with the related art, in the embodiment of the present application, obtaining the first image information of the target object from multiple angles facilitates subsequent feature extraction of the first image information to obtain the type of clothing worn by the target object. Then, a predicted clothing model corresponding to the clothing type is found from a preset database, and an image on the target object's clothing is extracted from the first image information from multiple angles, and the pattern is mapped onto the preset clothing model to obtain a candidate clothing model that is consistent with the clothing worn by the target object in reality. Video information of the target object performing activities according to a preset action is obtained. When the target object performs activities according to the preset action, wrinkles and other deformations will occur at the joints of the clothing, but in the virtual world, the wrinkles and other deformations at the joints cannot accurately simulate the same effect as in reality. Therefore, the wrinkle video of the clothing at each joint is determined from the video information. The wrinkle video records the specific situation of the wrinkles that appear during the activities at each joint. Therefore, the mapping video of the target object during real-person mapping is obtained, and the candidate clothing model is processed according to the wrinkle video and the mapping video to obtain the final clothing model of the target object during real-person virtual space mapping, that is, according to the specific situation of the target object making actions and activities in the mapping video, the manifestation form of the wrinkles at each joint is determined from the wrinkle video to obtain the final clothing model. The final clothing model is closer to the joints of the clothing worn by the target object in reality at each joint, so that the virtual mapping effect is better.

[0219] It should be understood that although the steps in the flowchart of the accompanying drawings are shown in sequence according to the indication of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps is not strictly limited in order, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

[0220] The above are only some embodiments of the present application. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A real-person digital avatar mapping method based on AI, characterized in that: include: Acquire first image information of a target object at multiple angles; Extracting features of the first image information to obtain the type of clothing worn by the target object; Finding a preset clothing model of the clothing type from a preset database, and extracting a clothing map of the clothing of the target object and a pattern on the clothing from the first image information of the multiple angles; Mapping the clothing map and pattern onto a corresponding preset clothing model to obtain a clothing model to be selected; Acquire video information of the target object performing activities according to preset actions; Determining the wrinkle video of the clothing at each joint from the video information; Acquire a mapping video of the target object when performing real-person mapping, and process the clothing model to be selected based on the wrinkle video and the mapping video to obtain a final clothing model of the target object when performing real-person mapping; The step of processing the clothing model to be selected based on the wrinkle video and the mapping video to obtain a final clothing model of the target object when performing real-person mapping includes: Determine the first wrinkle feature at each bending angle from the wrinkle video corresponding to each joint; Performing a rotation simulation on the first wrinkle feature at each bending angle to obtain a second wrinkle feature when the wrinkle feature at each bending angle is rotated to each angle, each second wrinkle feature corresponding to a bending angle and a rotation angle; Performing action recognition on each frame of the mapping video to obtain the bending angle formed at each joint in each frame and the relative rotation angle of the limbs on both sides of the joint; Searching for a target wrinkle feature corresponding to each joint from the second wrinkle feature of each joint based on the bending angle formed at each joint in each frame and the relative rotation angle of the limbs on both sides of the joint; The target wrinkle features corresponding to the joints are mapped to corresponding positions on the to-be-selected clothing model to obtain the final clothing model.

2. The AI-based real-person digital avatar mapping method according to claim 1, characterized in that: The step of mapping the clothing map and pattern onto a corresponding preset clothing model to obtain a clothing model to be selected includes: Mapping the clothing map corresponding to each angle into a preset coordinate system to determine the coordinate point of each position on the pattern; Acquire second image information of the preset clothing model at the multiple angles, and overlap and map the first image information and the second image information at the same angle; Mapping the pattern to each second image information according to the coordinate point corresponding to each angle of the pattern to obtain mapped second image information; The mapped second image information is inversely mapped onto the preset clothing model to obtain the clothing model to be selected.

3. The AI-based real-person digital avatar mapping method according to claim 1, characterized in that: Determining the wrinkle video of the clothing at each joint from the video information includes: Determine a preset range corresponding to each joint based on the position of the joint; Performing feature recognition on the video information to obtain feature points of each joint; A clipping area is obtained according to a preset range with the feature point as the center point, feature tracking is performed on each joint, and in the process of feature tracking, video information is clipped according to the clipping area to obtain a video of the wrinkles of the clothing at each joint.

4. The AI-based real-person digital avatar mapping method according to claim 1, characterized in that: The method further comprises: Determine the third image information of each joint in each frame from the mapping video, each third image information corresponds to the bending angle formed at the joint and the relative rotation angle of the limbs on both sides of the joint; Find out the associated second wrinkle feature whose bending angle and rotation angle are consistent with the third image information; Calculating a first similarity between the associated second wrinkle feature and the third image information; If the first similarity does not reach a preset similarity threshold, the associated second wrinkle feature is adjusted based on the third image information.

5. The AI-based real-person digital avatar mapping method according to claim 4, characterized in that: The adjusting the associated second wrinkle feature based on the third image information includes: Performing texture extraction on the third image information and the second wrinkle feature respectively to obtain a first texture map related to the third image information and a second texture map related to the second wrinkle feature; Overlapping the first texture map and the second texture map to determine a non-overlapping texture; Determine the grayscale value of the area corresponding to the non-overlapping texture from the third image information; The associated second wrinkle feature is adjusted according to the grayscale value and the non-overlapping texture to obtain an adjusted second wrinkle feature.

6. The AI-based real-person digital avatar mapping method according to claim 1, characterized in that: The method further comprises: Determine from the mapping video to perform edge detection on each frame to obtain fourth image information of the clothing; Calculating a second similarity between the fourth image information and the final clothing model corresponding to each frame; averaging the second similarity of each frame of the mapped video to obtain a second similarity average value; The second similarity average value is output.

7. An AI-based real-person digital avatar mapping system, characterized in that: include: An image acquisition module, used to acquire first image information of a target object at multiple angles; A first extraction module, configured to extract features from the first image information to obtain the type of clothing worn by the target object; A second extraction module, configured to search a preset clothing model of the clothing type from a preset database, and extract a clothing map of the clothing of the target object and a pattern on the clothing from the first image information of the multiple angles; A mapping module, used for mapping the clothing map and pattern onto a corresponding preset clothing model to obtain a clothing model to be selected; A video acquisition module, used to acquire video information of the target object performing activities according to preset actions; A video determination module, used to determine the wrinkle video of the clothing at each joint from the video information; A processing module, used for obtaining a mapping video of a target object when performing real-person mapping, and processing a clothing model to be selected based on the wrinkle video and the mapping video to obtain a final clothing model of the target object when performing real-person mapping; It includes: Determine the first wrinkle feature at each bending angle from the wrinkle video corresponding to each joint; Performing a rotation simulation on the first wrinkle feature at each bending angle to obtain a second wrinkle feature when the wrinkle feature at each bending angle is rotated to each angle, each second wrinkle feature corresponding to a bending angle and a rotation angle; Performing action recognition on each frame of the mapping video to obtain the bending angle formed at each joint in each frame and the relative rotation angle of the limbs on both sides of the joint; Searching for a target wrinkle feature corresponding to each joint from the second wrinkle feature of each joint based on the bending angle formed at each joint in each frame and the relative rotation angle of the limbs on both sides of the joint; The target wrinkle features corresponding to the joints are mapped to corresponding positions on the to-be-selected clothing model to obtain the final clothing model.

8. An electronic device, characterized in that: It includes: at least one processor; Memory; At least one application, wherein the at least one application is stored in the memory and configured to be executed by the at least one processor, and the at least one application is used to execute an AI-based real-person digital avatar mapping method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed in a computer, the computer is caused to execute the AI-based real-person digital avatar mapping method according to any one of claims 1 to 6.

Images