Virtual character construction system and method based on multi-modal information fusion

Abstract

The invention discloses a virtual character construction system and method based on multi-modal information fusion, and relates to the technical field of virtual character construction, and the method comprises the steps: evaluating the image construction quality of a user image in image data; obtaining historical image data of a historical sample user in the construction model, obtaining target image data of the user, and analyzing a reference value of the historical sample user for virtual role construction of the user; the method comprises the steps of obtaining target image data, constructing a virtual role of a user to obtain virtual role data, obtaining historical virtual role data of a target historical sample user, evaluating the matching degree between the virtual role data and the user, and optimizing the virtual role data to obtain target virtual role data; and obtaining posture data of the user, rendering the virtual role of the user in combination with the target virtual role data of the user, and managing the virtual role constructed by the user.

Description

Technical Field

[0001] The present invention relates to the technical field of virtual character construction, and in particular to a virtual character construction system and method based on multimodal information fusion. Background Art

[0002] Multimodal information fusion refers to the technology of integrating and processing multiple information from different sensors and different types of data sources. With the continuous development of science and technology, multimodal information fusion technology has been increasingly widely used in the construction of virtual characters. The construction of virtual characters based on multimodal information fusion mainly includes but is not limited to the following benefits: 1. Enhanced sense of reality. The behavior and communication methods of virtual characters constructed by multimodal information fusion technology more realistically simulate humans, giving users a more realistic experience; 2. Enhanced expressive ability. Different modal information can complement and enhance each other, making the expressions and movements of the constructed virtual characters more comprehensive and delicate.

[0003] In the process of building a virtual character, it is necessary to obtain the user's image data and input the image data into the model to build the user's virtual character. Although the virtual character built in this way is fast, there are some problems. First of all, whether the constructed virtual character meets the user's characteristics and the set game style requires manual judgment, and manual judgment is prone to inaccuracies. In addition, once the virtual character does not meet the user's characteristics and the set game style, the model cannot optimize and modify the constructed virtual character, which will not only make the character less realistic, but also lead to insufficient resonance and identification of the user with the character, affecting the user's activity and retention rate, making it impossible for the company to profit, and may even cause serious economic and reputational losses to the company. Summary of the invention

[0004] The purpose of the present invention is to provide a virtual character construction system and method based on multimodal information fusion to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for constructing a virtual character based on multimodal information fusion, the method comprising: Step S100: obtaining image data of the user, obtaining sample image data in the constructed model, evaluating the image construction quality of the user image in the image data, and obtaining target image data; Step S200: Obtain historical image data of historical sample users in the construction model, obtain target image data of the user, analyze the reference value of the historical sample users to the construction of the user's virtual role, and obtain the target historical sample user; Step S300: Acquire target image data, construct a virtual character of the user, obtain virtual character data, obtain historical virtual character data of a target historical sample user, evaluate the degree of fit between the virtual character data and the user, and optimize the virtual character data to obtain target virtual character data; Step S400: Acquire the user's posture data, and render the user's virtual character in combination with the user's target virtual character data, and manage the user's constructed virtual character.

[0006] Furthermore, step S100 includes: Step S101: acquiring image data of a user in the current cycle, the image data including each user image input into the construction model when the user is constructing a virtual character; Step S102: Acquire sample image data for constructing a model for image evaluation, where the sample image data includes each sample image in the constructed model; Step S103: performing grayscale conversion on the user image of the user, constructing a two-dimensional coordinate system with one pixel as a unit length, and obtaining the coordinates of each pixel in the user image; Calculate the image sharpness value of each user image of the user, where the image sharpness value of the a-th user image of the user is B a : , Among them, K is the preset Laplace convolution kernel; I a (α, β) is the grayscale value of the pixel with abscissa α and ordinate β in the a-th user image; n is the maximum value of the abscissa of the pixel in the a-th user image; m is the maximum value of the ordinate of the pixel in the a-th user image; Step S104: Evaluate the image construction quality of each user image in the image data, wherein the image construction quality of the d-th user image in the image data is evaluated, and the specific process is as follows: Get the average grayscale value μ of the pixels in the dth user image d , obtain the image sharpness of each sample image, and calculate the quality evaluation value T of the dth user image d : , Where j is the total number of sample images; B´ i is the image sharpness of the i-th sample image; B d is the image sharpness of the d-th user image; μ´ i is the average gray value of the pixels in the i-th sample image; γ1 and γ2 are respectively the preset first evaluation coefficient and the second evaluation coefficient; When Td Less than or equal to the preset quality evaluation threshold T △ , then the construction quality of the dth user image is judged to be unqualified. When the quality evaluation value T d >T △ , determine that the construction quality of the d-th user image is qualified, and record the d-th user image as the user's target user image; Step S105: Acquire and aggregate the target user images of the user to obtain the target image data of the user.

[0007] Further, step S200 includes: Step S201: acquiring historical image data of historical sample users in the constructed model, wherein the historical sample users are historical users whose constructed virtual characters have been verified and whose virtual character type is the same as that of the user, and the historical image data includes historical user images of the historical sample users; Step S202: construct a two-dimensional coordinate system, perform grayscale conversion on the historical user image to obtain image E (x, y), perform Gaussian filtering on the historical user image to obtain image R (x, y), and the specific filtering formula is: , Among them, σ is the standard deviation preset in the construction model; Step S203: Obtain each historical sample user in the construction model, and analyze the reference value of each historical sample user to the construction of the user's virtual role, wherein the reference value of the e-th historical sample user to the construction of the user's virtual role is analyzed. The specific process is: Obtain the historical user image of the e-th historical sample user after Gaussian filtering, and use a preset key point detection algorithm to obtain each key joint point in the historical user image of the e-th historical sample user; Step S204: Based on the historical user images of the e-th historical sample user, resize and Gaussian filter each target user image in the target image data, and calculate the construction reference value F of the e-th historical sample user to the user e : , Among them, δ is the total number of target user images; q is the total number of key joint points; D z ε,ε+1 is the distance between the εth key joint point and the ε+1th key joint point in the user’s zth target user image; D ε,ε+1 is the distance between the εth key joint point and the ε+1th key joint point in the historical user image of the eth historical sample user; Step S205: When constructing the reference value Fe If it is greater than a preset construction reference threshold, it is determined that the e-th historical sample user has reference value for the construction of the user's virtual character, and the e-th historical sample user is recorded as the target historical sample user of the user; The reason for performing Gaussian filtering on the historical user image in the above steps is that Gaussian filtering can effectively remove random noise in the image, making the image smoother, and Gaussian filtering can ensure the stability of image extraction. By analyzing the image after Gaussian filtering, the accuracy of the construction reference value of the historical sample user for the user can be guaranteed, so as to provide data support for the optimization of the virtual character constructed by the user in the following text, and thus make the virtual character constructed by the construction model more in line with the user himself.

[0008] Further, step S300 includes: Step S301: Obtain the target image data of the user, use the construction model to construct the virtual character of the user, obtain virtual character data, where the virtual character data is the 3D model of the virtual character of the user constructed by the construction model, and obtain the bounding box of the 3D model of the virtual character; Obtain the historical virtual character data of the target historical sample user of the user. The historical virtual character data includes the 3D model of the virtual character of the target historical sample user, and evaluate the degree of fit between the virtual character data and the user. The specific process is as follows: Obtain the bounding boxes of the 3D models of the virtual characters of each target historical sample user of the user, and calculate the feature fit value U between the virtual character data and the user: , where L, W, and H are the length, width, and height of the bounding box of the 3D model of the virtual character in the virtual character data; L v , W v , H v are the length, width, and height of the bounding box of the 3D model of the virtual character of the v-th target historical sample character of the user respectively; p is the total number of each target historical sample user; Step S302: When the feature fit value U is greater than the preset feature fit threshold U´, it is determined that the virtual character data fits the user, and the virtual character data is recorded as the target virtual character data. When U < U´, it is determined that the virtual character data does not fit the user. Based on the historical virtual character data of each target historical sample user, adjust the various model parameters in the 3D model of the virtual character of the virtual character data to optimize the 3D model of the virtual character of the user until U ≥ U´, and save the optimized 3D model of the virtual character to obtain the target virtual character data.

[0009] Further, step S400 includes: Step S401: Acquire user's posture data, where the posture data includes video information collected by a camera when the user is performing motion capture; Step S402: Based on the posture data and in combination with the user's target virtual character data, the user's virtual character is rendered, and the user's virtual character is color corrected using image editing software to manage the user's constructed virtual character.

[0010] In order to better implement the above method, a virtual character construction system based on multimodal information fusion is also proposed, the system includes a target image data module, a reference evaluation module, a data optimization module, and an intelligent management module; A target image data module is used to evaluate the image construction quality of the user image in the image data to obtain target image data; A reference evaluation module is used to obtain the target image data of the user, analyze the reference value of the historical sample users to the user's virtual character construction, and obtain the target historical sample users; A data optimization module is used to evaluate the degree of fit between the virtual character data and the user, and optimize the virtual character data to obtain target virtual character data; The intelligent management module is used to render the user's virtual character and manage the user's constructed virtual character.

[0011] Further, the target image data module includes a quality evaluation value unit and a target image data unit; A quality evaluation value unit, used to calculate a quality evaluation value of a user's user image; The target image data unit is used to evaluate the image construction quality of the user image according to the quality evaluation value of the user image to obtain the target image data.

[0012] Further, the reference evaluation module includes a reference value construction unit and a reference evaluation unit; A construction reference value unit, used to calculate the construction reference value of the historical sample user to the user; The reference analysis unit is used to analyze the reference value of the historical sample users for the user's virtual role construction according to the construction reference value, and obtain the target historical sample users.

[0013] Furthermore, the data optimization module includes a virtual character construction unit and a data optimization unit; A virtual character construction unit is used to construct a virtual character of a user and obtain virtual character data; The data optimization unit is used to optimize the virtual character data to obtain target virtual character data.

[0014] Further, the intelligent management module includes an intelligent management unit; The intelligent management unit is used to acquire the user's posture data, and render the user's virtual character in combination with the user's target virtual character data, and manage the user's constructed virtual character.

[0015] Compared with the prior art, the beneficial effects of the present invention are: the present invention realizes the intelligent construction of the user's virtual character, and before the user's virtual character is constructed, the image quality of the user's user image is evaluated by using the sample image data in the construction model, so that the image quality of the user image of the virtual character is guaranteed, and the virtual character model constructed by the user is optimized and adjusted through the user's target historical sample users, so that the constructed virtual character model is more suitable for the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a method flow chart of a virtual character construction method based on multimodal information fusion of the present invention; Figure 2 It is a module schematic diagram of a virtual character construction system based on multimodal information fusion of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution, a method for constructing a virtual character based on multimodal information fusion, the method comprising: Step S100: obtaining image data of the user, obtaining sample image data in the constructed model, evaluating the image construction quality of the user image in the image data, and obtaining target image data; Wherein, step S100 includes: Step S101: acquiring image data of a user in the current cycle, the image data including each user image input into the construction model when the user is constructing a virtual character; Step S102: Acquire sample image data for constructing a model for image evaluation, where the sample image data includes each sample image in the constructed model; Step S103: performing grayscale conversion on the user image of the user, constructing a two-dimensional coordinate system with one pixel as a unit length, and obtaining the coordinates of each pixel in the user image; Calculate the image sharpness value of each user image of the user, where the image sharpness value of the a-th user image of the user is B a : , Among them, K is the preset Laplace convolution kernel; I a (α, β) is the grayscale value of the pixel with abscissa α and ordinate β in the a-th user image; n is the maximum value of the abscissa of the pixel in the a-th user image; m is the maximum value of the ordinate of the pixel in the a-th user image; Step S104: Evaluate the image construction quality of each user image in the image data, wherein the image construction quality of the d-th user image in the image data is evaluated, and the specific process is as follows: Get the average grayscale value μ of the pixels in the dth user image d , obtain the image sharpness of each sample image, and calculate the quality evaluation value T of the dth user image d : , Where j is the total number of sample images; B´ i is the image sharpness of the i-th sample image; B d is the image sharpness of the d-th user image; μ´ i is the average gray value of the pixels in the i-th sample image; γ1 and γ2 are respectively the preset first evaluation coefficient and the second evaluation coefficient; When T d Less than or equal to the preset quality evaluation threshold T △ , then the construction quality of the dth user image is judged to be unqualified. When the quality evaluation value T d >T △ , determine that the construction quality of the d-th user image is qualified, and record the d-th user image as the user's target user image; Step S105: Acquire and aggregate target user images of the user to obtain target image data of the user; Step S200: Obtain historical image data of historical sample users in the construction model, obtain target image data of the user, analyze the reference value of the historical sample users to the construction of the user's virtual role, and obtain the target historical sample user; Wherein, step S200 includes: Step S201: acquiring historical image data of historical sample users in the constructed model, wherein the historical sample users are historical users whose constructed virtual characters have been verified and whose virtual character type is the same as that of the user, and the historical image data includes historical user images of the historical sample users; Step S202: construct a two-dimensional coordinate system, perform grayscale conversion on the historical user image to obtain image E (x, y), perform Gaussian filtering on the historical user image to obtain image R (x, y), and the specific filtering formula is: , Among them, σ is the standard deviation preset in the construction model; Step S203: Obtain each historical sample user in the construction model, and analyze the reference value of each historical sample user to the construction of the user's virtual role, wherein the reference value of the e-th historical sample user to the construction of the user's virtual role is analyzed. The specific process is: Obtain the historical user image of the e-th historical sample user after Gaussian filtering, and use a preset key point detection algorithm to obtain each key joint point in the historical user image of the e-th historical sample user; Step S204: Based on the historical user images of the e-th historical sample user, resize and Gaussian filter each target user image in the target image data, and calculate the construction reference value F of the e-th historical sample user to the user e : , Among them, δ is the total number of target user images; q is the total number of key joint points; D z ε,ε+1 is the distance between the εth key joint point and the ε+1th key joint point in the user’s zth target user image; D ε,ε+1 is the distance between the εth key joint point and the ε+1th key joint point in the historical user image of the eth historical sample user; Step S205: When constructing the reference value F e If the value is greater than the preset construction reference threshold, it is determined that the e-th historical sample user has reference value for the user's virtual character construction, and the e-th historical sample user is recorded as the user's target historical sample user; Step S300: Acquire target image data, construct a virtual character of the user, obtain virtual character data, obtain historical virtual character data of a target historical sample user, evaluate the degree of fit between the virtual character data and the user, and optimize the virtual character data to obtain target virtual character data; Wherein, step S300 includes: Step S301: obtaining target image data of the user, constructing a virtual character of the user using a construction model, obtaining virtual character data, the virtual character data being a 3D model of the virtual character of the user constructed by the construction model, and obtaining a bounding box of the 3D model of the virtual character; Obtain the historical virtual character data of the target historical sample users of the user. The historical virtual character data includes the 3D model of the virtual character of the target historical sample users. Evaluate the degree of fit between the virtual character data and the user. The specific process is as follows: Obtain the bounding boxes of the 3D models of the virtual characters of each target historical sample user of the user, and calculate the feature fit value U between the virtual character data and the user: , where L, W, and H are the length, width, and height of the bounding box of the 3D model of the virtual character in the virtual character data; L v , W v , H v are the length, width, and height of the bounding box of the 3D model of the virtual character of the v-th target historical sample character of the user; p is the total number of all target historical sample users; For example, the length, width, and height L, W, and H of the bounding box of the 3D model of the virtual character in the virtual character data are 10, 4, and 8 respectively; p is 2; the length, width, and height L1, W1, and H1 of the bounding box of the 3D model of the virtual character of the first target historical sample character of the user are 12, 8, and 10 respectively; the length, width, and height L2, W2, and H2 of the bounding box of the 3D model of the virtual character of the second target historical sample character of the user are 14, 10, and 6 respectively; , Step S302: When the feature fit value U is greater than the preset feature fit threshold U´, determine that the virtual character data fits the user, and record the virtual character data as the target virtual character data. When U < U´, determine that the virtual character data does not fit the user. Based on the historical virtual character data of each target historical sample user, adjust the various model parameters in the 3D model of the virtual character of the virtual character data, optimize the 3D model of the virtual character of the user until U ≥ U´, and save the optimized 3D model of the virtual character to obtain the target virtual character data; For example, the various model parameters include geometric parameters, transformation parameters, etc.; Step S400: Obtain the pose data of the user, and combine it with the target virtual character data of the user to render the virtual character of the user and manage the constructed virtual character of the user; Among them, step S400 includes: Step S401: Obtain the pose data of the user. The pose data includes the video information collected by the camera when the user performs motion capture; Step S402: Based on the pose data, combine it with the target virtual character data of the user to render the virtual character of the user, and use image editing software to perform color correction on the virtual character of the user and manage the constructed virtual character of the user; In order to better implement the above method, a virtual character construction system based on multimodal information fusion is also proposed, the system includes a target image data module, a reference evaluation module, a data optimization module, and an intelligent management module; A target image data module is used to evaluate the image construction quality of the user image in the image data to obtain target image data; A reference evaluation module is used to obtain the target image data of the user, analyze the reference value of the historical sample users to the user's virtual character construction, and obtain the target historical sample users; A data optimization module is used to evaluate the degree of fit between the virtual character data and the user, and optimize the virtual character data to obtain target virtual character data; An intelligent management module, used to render the user's virtual character and manage the user's constructed virtual character; Wherein, the target image data module includes a quality evaluation value unit and a target image data unit; A quality evaluation value unit, used to calculate a quality evaluation value of a user's user image; A target image data unit, used to evaluate the image construction quality of the user image according to the quality evaluation value of the user image, and obtain target image data; Wherein, the reference evaluation module includes a reference value construction unit and a reference evaluation unit; A construction reference value unit, used to calculate the construction reference value of the historical sample user to the user; A reference analysis unit, used to analyze the reference value of the historical sample users for the user's virtual role construction according to the construction reference value, and obtain the target historical sample users; Among them, the data optimization module includes a virtual character construction unit and a data optimization unit; A virtual character construction unit is used to construct a virtual character of a user and obtain virtual character data; A data optimization unit, used for optimizing the virtual character data to obtain target virtual character data; Wherein, the intelligent management module includes an intelligent management unit; The intelligent management unit is used to acquire the user's posture data, and render the user's virtual character in combination with the user's target virtual character data, and manage the user's constructed virtual character.

[0019] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A method for constructing a virtual character based on multimodal information fusion, characterized in that: The method comprises: Step S100: obtaining image data of a user, obtaining sample image data in a constructed model, evaluating the image construction quality of the user image in the image data, and obtaining target image data; Step S200: acquiring historical image data of historical sample users in the constructed model, acquiring target image data of the user, analyzing the reference value of the historical sample users to the construction of the user's virtual role, and obtaining the target historical sample users; Step S300: acquiring the target image data, constructing a virtual character of the user to obtain virtual character data, acquiring historical virtual character data of the target historical sample user, evaluating the degree of fit between the virtual character data and the user, and optimizing the virtual character data to obtain target virtual character data; Step S400: Acquire the posture data of the user, and render the virtual character of the user in combination with the target virtual character data of the user, and manage the constructed virtual character of the user.

2. The method for constructing a virtual character based on multimodal information fusion according to claim 1, characterized in that: The step S100 includes: Step S101: acquiring image data of a user in a current cycle, wherein the image data includes each user image input into a construction model when the user constructs a virtual character; Step S102: Acquire sample image data for image evaluation of the constructed model, wherein the sample image data includes each sample image in the constructed model; Step S103: performing grayscale conversion on the user image of the user, constructing a two-dimensional coordinate system with one pixel as a unit length, and obtaining the coordinates of each pixel in the user image; Calculate the image sharpness value of each user image of the user, wherein the image sharpness B of the a-th user image of the user a : ， Among them, K is the preset Laplace convolution kernel; I a (α, β) is the grayscale value of the pixel with abscissa α and ordinate β in the a-th user image; n is the maximum value of the abscissa of the pixel in the a-th user image; m is the maximum value of the ordinate of the pixel in the a-th user image; Step S104: evaluating the image construction quality of each user image in the image data, wherein the image construction quality of the d-th user image in the image data is evaluated, and the specific process is as follows: Get the average grayscale value μ of the pixels in the dth user image d , obtain the image sharpness of each sample image, and calculate the quality evaluation value T of the dth user image d : ， Wherein, j is the total number of sample images; B´ i is the image sharpness of the i-th sample image; B d is the image sharpness of the d-th user image; μ´ i is the average gray value of the pixels in the i-th sample image; γ1 and γ2 are respectively the preset first evaluation coefficient and the second evaluation coefficient; When the T d Less than or equal to the preset quality evaluation threshold T △ , then the construction quality of the dth user image is judged to be unqualified, when the quality evaluation value T d >T △ , determining that the construction quality of the d-th user image is qualified, and recording the d-th user image as the target user image of the user; Step S105: Acquire and aggregate target user images of the user to obtain target image data of the user.

3. The method for constructing a virtual character based on multimodal information fusion according to claim 2, characterized in that: The step S200 includes: Step S201: acquiring historical image data of historical sample users in the constructed model, wherein the historical sample users are historical users whose constructed virtual characters have been verified and whose virtual character types are the same as those of the users, and the historical image data include historical user images of the historical sample users; Step S202: construct a two-dimensional coordinate system, perform grayscale conversion on the historical user image to obtain image E (x, y), perform Gaussian filtering on the historical user image to obtain image R (x, y), and the specific filtering formula is: ， Among them, σ is the standard deviation preset in the construction model; Step S203: Acquire each historical sample user in the construction model, and analyze the reference value of each historical sample user to the construction of the user's virtual role, wherein the reference value of the e-th historical sample user to the construction of the user's virtual role is analyzed. The specific process is: Obtain the historical user image of the e-th historical sample user after Gaussian filtering, and use a preset key point detection algorithm to obtain each key joint point in the historical user image of the e-th historical sample user; Step S204: Based on the historical user images of the e-th historical sample user, resize and Gaussian filter each target user image in the target image data, and calculate the construction reference value F of the e-th historical sample user for the user e : ， Wherein, δ is the total number of the target user images; q is the total number of the key joint points; D z ε,ε+1 is the distance between the εth key joint point and the ε+1th key joint point in the zth target user image of the user; D ε,ε+1 is the value of the distance between the εth key joint point and the ε+1th key joint point in the historical user image of the eth historical sample user; Step S205: When the construction reference value F e If the value of the e-th historical sample user is greater than a preset construction reference threshold, it is determined that the e-th historical sample user has reference value for the construction of the user's virtual role, and the e-th historical sample user is recorded as the target historical sample user of the user.

4. The method for constructing a virtual character based on multimodal information fusion according to claim 3, characterized in that: The step S300 includes: Step S301: Obtain the target image data of the user, use a construction model to construct a virtual character of the user, and obtain virtual character data. The virtual character data is a 3D model of the virtual character of the user constructed by the construction model, and obtain the bounding box of the 3D model of the virtual character; Obtain the historical virtual character data of the target historical sample user of the user. The historical virtual character data includes the 3D model of the virtual character of the target historical sample user, and evaluate the degree of fit between the virtual character data and the user. The specific process is as follows: Obtain the bounding boxes of the 3D models of the virtual characters of each target historical sample user of the user, and calculate the feature fit value U between the virtual character data and the user: ， Wherein, L, W, and H are respectively the length, width, and height of the bounding box of the 3D model of the virtual character in the virtual character data; L v , W v , H v are the length, width and height of the bounding box of the 3D model of the virtual character of the vth target historical sample character of the user; p is the total number of the target historical sample users; Step S302: When the feature fit value U is greater than a preset feature fit threshold U´, determine that the virtual character data fits the user, and record the virtual character data as the target virtual character data. When U < U´, determine that the virtual character data does not fit the user, and based on the historical virtual character data of each target historical sample user, adjust the various model parameters in the 3D model of the virtual character of the virtual character data to optimize the 3D model of the virtual character of the user until U ≥ U´, and save the optimized 3D model of the virtual character to obtain the target virtual character data.

5. The method for constructing a virtual character based on multimodal information fusion according to claim 4, characterized in that: The step S400 includes: Step S401: Obtain the pose data of the user. The pose data includes the video information collected by the camera when the user performs motion capture; Step S402: Based on the pose data, combined with the target virtual character data of the user, render the virtual character of the user, and use image editing software to perform color correction on the virtual character of the user and manage the constructed virtual character of the user.

6. A virtual character construction system based on multimodal information fusion, used to execute a virtual character construction method based on multimodal information fusion as described in any one of claims 1 to 5, characterized in that: The system includes a target image data module, a reference evaluation module, a data optimization module, and an intelligent management module; The target image data module is used to evaluate the image construction quality of the user image in the image data to obtain target image data; The reference evaluation module is used to obtain the target image data of the user, analyze the reference value of the historical sample user for the construction of the virtual character of the user, and obtain the target historical sample user; The data optimization module is used to evaluate the degree of fit between the virtual character data and the user, and optimize the virtual character data to obtain the target virtual character data; The intelligent management module is used to render the virtual character of the user and manage the constructed virtual character of the user.

7. A virtual character construction system based on multimodal information fusion according to claim 6, characterized in that: The target image data module includes a quality evaluation value unit and a target image data unit; The quality evaluation value unit is used to calculate the quality evaluation value of the user image of the user; The target image data unit is used to evaluate the image construction quality of the user image according to the quality evaluation value of the user image to obtain the target image data.

8. The virtual character construction system based on multimodal information fusion according to claim 6, characterized in that: The reference evaluation module includes a reference value construction unit and a reference evaluation unit; The construction reference value unit is used to calculate the construction reference value of the historical sample user to the user; The reference analysis unit is used to analyze the reference value constructed by the historical sample users for the user's virtual role according to the constructed reference value, and obtain the target historical sample users.

9. The virtual character construction system based on multimodal information fusion according to claim 6, characterized in that: The data optimization module includes a virtual character construction unit and a data optimization unit; The virtual character construction unit is used to construct the virtual character of the user to obtain virtual character data; The data optimization unit is used to optimize the virtual character data to obtain target virtual character data.

10. The virtual character construction system based on multimodal information fusion according to claim 6, characterized in that: The intelligent management module includes an intelligent management unit; The intelligent management unit is used to acquire the posture data of the user, and render the virtual character of the user in combination with the target virtual character data of the user, and manage the constructed virtual character of the user.

Images