Virtual digital human generation method

Abstract

The invention relates to virtual reality, in particular to a virtual digital human generation method, which comprises the following steps: acquiring a front image of a target head; generating an initial three-dimensional model corresponding to the front image of the target head based on a preset substrate three-dimensional model; determining a series of facial actions according to the propaganda language, and generating a corresponding intermediate three-dimensional model based on the initial three-dimensional model; performing interpolation processing on the initial three-dimensional model and the middle three-dimensional model by adopting an improved spherical linear interpolation method to obtain multiple groups of interpolation three-dimensional coordinates; a corresponding transition three-dimensional model is generated according to the multiple sets of interpolation three-dimensional coordinates, and a dynamic model sequence is generated by combining the initial three-dimensional model and the middle three-dimensional model; performing three-dimensional rendering on the dynamic model sequence by using a preset three-dimensional renderer to generate a video sequence; according to the technical scheme, the defects that in the prior art, it is difficult to ensure that the face actions of the virtual digital human are matched with publicity languages, and the face actions of the virtual digital human are difficult to change stably can be effectively overcome.

Description

Technical Field

[0001] The invention relates to virtual reality, and in particular to a method for generating a virtual digital human. Background Art

[0002] With the rapid development of science and technology, virtual digital human technology has become a new favorite for corporate image promotion. Virtual digital human, as a virtual character image built by the most advanced computer technology, can simulate human appearance, voice, movement and behavior, and can even self-learn through artificial intelligence technology, providing a more real and intelligent experience.

[0003] The development of virtual digital human technology has not only enriched the ways of corporate image promotion, but also brought unprecedented marketing opportunities to enterprises. Virtual digital humans are widely used in corporate image promotion, including brand spokespersons, virtual customer service, virtual lecturers, etc. As brand spokespersons, virtual digital humans can create a more charismatic brand image for enterprises and attract consumers' attention.

[0004] Although virtual digital human technology has shown great potential in corporate image promotion, it still faces some technical challenges. For example, how to ensure that the facial movements of virtual digital humans are compatible with the promotional language, how to make the facial movements of virtual digital humans change more smoothly, and effectively improve the realism. In the future, virtual digital human technology will play a more important role in corporate image promotion and bring greater commercial value to enterprises. Summary of the invention

[0005] 1. Technical issues to be solved

[0006] In view of the above-mentioned shortcomings of the prior art, the present invention provides a method for generating a virtual digital human, which can effectively overcome the defects of the prior art that it is difficult to ensure that the facial movements of the virtual digital human are compatible with the promotional language, and that the facial movements of the virtual digital human are difficult to change smoothly.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A method for generating a virtual digital human comprises the following steps:

[0010] S1, obtain the front image of the target head;

[0011] S2, generating an initial three-dimensional model corresponding to the frontal image of the target head based on a preset base three-dimensional model;

[0012] S3, determining a series of facial movements according to the promotional language, and generating a corresponding intermediate three-dimensional model based on the initial three-dimensional model;

[0013] S4, using an improved spherical linear interpolation method to interpolate the initial three-dimensional model and the intermediate three-dimensional model to obtain multiple sets of interpolated three-dimensional coordinates;

[0014] S5, generating corresponding transitional three-dimensional models according to the multiple sets of interpolated three-dimensional coordinates, and generating a dynamic model sequence by combining the initial three-dimensional model and the intermediate three-dimensional model;

[0015] S6, using a preset 3D renderer to perform 3D rendering on the dynamic model sequence to generate a video sequence;

[0016] Among them, in the improved spherical linear interpolation method, on the one hand, an angle threshold is introduced and a nonlinear interpolation method is used to deal with the situation where the angle approaches 0. On the other hand, the parameters of spherical linear interpolation are optimized to avoid calculating complex trigonometric functions and improve calculation efficiency.

[0017] Preferably, in S2, generating an initial three-dimensional model corresponding to the frontal image of the target head based on a preset base three-dimensional model includes:

[0018] S21, generating contour three-dimensional model data of the target head front image according to the target head front image and the base three-dimensional model;

[0019] S22, performing alignment processing on the organs in the target head front image to generate three-dimensional model data of the organs in the target head front image;

[0020] S23, fusing the contour three-dimensional model data and the organ three-dimensional model data to generate an initial three-dimensional model corresponding to the target head frontal image.

[0021] Preferably, generating contour three-dimensional model data of the target head front image according to the target head front image and the base three-dimensional model in S21 includes:

[0022] S211, extracting a first feature point in the front image of the target head, and determining a second feature point corresponding to the first feature point in the base three-dimensional model;

[0023] S212: Generate contour three-dimensional model data of the target head frontal image based on the first feature point and the second feature point.

[0024] Preferably, generating the contour three-dimensional model data of the target head frontal image according to the first feature point and the second feature point in S212 includes:

[0025] S2121, aligning the base three-dimensional model with the target head front image according to the first feature point and the second feature point to obtain an aligned base three-dimensional model;

[0026] S2122, using radial basis functions to perform interpolation processing on the aligned base three-dimensional model to generate contour three-dimensional model data of the target head frontal image.

[0027] Preferably, in S3, a series of facial movements are determined according to the promotional language, and a corresponding intermediate three-dimensional model is generated based on the initial three-dimensional model, including:

[0028] A series of facial movements are determined according to the language text and language emotion of the propaganda language, and the mapping of each facial movement is completed on the initial three-dimensional model to generate a corresponding intermediate three-dimensional model.

[0029] Preferably, in S4, an improved spherical linear interpolation method is used to interpolate the initial three-dimensional model and the intermediate three-dimensional model to obtain multiple sets of interpolated three-dimensional coordinates, including:

[0030] S41, preprocessing the initial three-dimensional model and the intermediate three-dimensional model;

[0031] S42, extracting a set of feature points from each three-dimensional model, and using a feature point matching algorithm to determine the corresponding relationship between the feature points of two adjacent three-dimensional models;

[0032] S43, mapping the matched feature points of each three-dimensional model onto a high-dimensional unit sphere to convert the feature point coordinates into a corresponding set of intermediate unit vectors;

[0033] S44, selecting a series of interpolation points in the spherical space between two adjacent three-dimensional models, and obtaining multiple groups of transition unit vectors;

[0034] S45, based on the intermediate unit vector and the transition unit vector, using an improved spherical linear interpolation method to perform interpolation calculation on two adjacent three-dimensional models to obtain multiple groups of interpolation unit vectors;

[0035] S46, converting multiple groups of interpolation unit vectors into corresponding interpolation three-dimensional coordinates;

[0036] S47, repeat S44 to S46 until the interpolation calculation of the last two intermediate three-dimensional models is completed to obtain multiple sets of interpolated three-dimensional coordinates.

[0037] Preferably, preprocessing the initial three-dimensional model and the intermediate three-dimensional model in S41 includes:

[0038] The initial 3D model and the intermediate 3D model are preprocessed including noise removal, surface smoothing and size standardization.

[0039] Preferably, in S45, based on the intermediate unit vector and the transition unit vector, an improved spherical linear interpolation method is used to perform interpolation calculation on two adjacent three-dimensional models to obtain multiple groups of interpolation unit vectors, including:

[0040] S451, selecting two adjacent groups of intermediate unit vectors and the first group of transition unit vectors therebetween, and using the corresponding unit vectors in the first group of intermediate unit vectors and the first group of transition unit vectors as the starting unit vector and the target unit vector, respectively;

[0041] S452, using an improved spherical linear interpolation formula to perform interpolation calculation on the starting unit vector and the target unit vector to obtain an interpolation unit vector;

[0042] S453, repeat S451 to S452 until the interpolation calculation of the last pair of unit vectors in the corresponding relationship between the previous group of intermediate unit vectors and the first group of transition unit vectors is completed;

[0043] S454, repeat S451 to S453 until the interpolation calculation of the last group of transition unit vectors and the last pair of corresponding unit vectors in the next group of intermediate unit vectors is completed, and multiple groups of interpolation unit vectors are obtained.

[0044] Preferably, in S452, an improved spherical linear interpolation formula is used to perform interpolation calculation on the starting unit vector and the target unit vector to obtain an interpolated unit vector, including:

[0045] The spherical linear interpolation formula is improved by introducing an angle threshold and using a nonlinear interpolation method to deal with the situation where the angle approaches 0, and by optimizing the parameters of the spherical linear interpolation to avoid calculating complex trigonometric functions to improve the calculation efficiency:

[0046]

[0047] Among them, Slerp(q 0 ,q 1 ) is the calculated interpolation unit vector, q 0 is the starting unit vector, q 1 is the target unit vector, θ is the starting unit vector q 0 With the target unit vector q 1 , ε is the angle threshold, t is the interpolation parameter, 0≤t≤1.

[0048] Preferably, in S5, corresponding transitional three-dimensional models are generated according to the multiple sets of interpolated three-dimensional coordinates, and a dynamic model sequence is generated by combining the initial three-dimensional model and the intermediate three-dimensional model, including:

[0049] Corresponding transitional three-dimensional models are generated according to multiple sets of interpolated three-dimensional coordinates, and the transitional three-dimensional models are inserted into corresponding positions between the initial three-dimensional model and the intermediate three-dimensional model, and the intermediate three-dimensional models, to generate a dynamic model sequence.

[0050] (III) Beneficial effects

[0051] Compared with the prior art, the method for generating a virtual digital human provided by the present invention has the following beneficial effects:

[0052] 1) Generate contour 3D model data of the target head front image according to the target head front image and the base 3D model, perform alignment processing on the organs in the target head front image, generate organ 3D model data of the target head front image, perform fusion processing on the contour 3D model data and the organ 3D model data to generate an initial 3D model corresponding to the target head front image, determine a series of facial movements according to the language text and language emotion of the promotional language, complete the mapping of each facial movement on the initial 3D model, and generate a corresponding intermediate 3D model, so as to ensure that the intermediate 3D model has facial movements that are compatible with the language text and language emotion of the promotional language, effectively improve the viewer's perception, help enterprises create a more charismatic brand image, and attract consumers' attention;

[0053] 2) An improved spherical linear interpolation method is used to interpolate the initial three-dimensional model and the intermediate three-dimensional model to obtain multiple sets of interpolated three-dimensional coordinates, and corresponding transitional three-dimensional models are generated according to the multiple sets of interpolated three-dimensional coordinates. A dynamic model sequence is generated by combining the initial three-dimensional model and the intermediate three-dimensional model, and a preset three-dimensional renderer is used to perform three-dimensional rendering on the dynamic model sequence to generate a video sequence. The improved spherical linear interpolation method can make the virtual digital human smoothly transform between two adjacent intermediate three-dimensional models, effectively improve the realism, further enhance the viewer's perception, play a more important role in corporate image promotion, and bring greater commercial value to the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0055] Figure 1 It is a schematic diagram of the process of the present invention;

[0056] Figure 2 A schematic diagram of the process of generating an initial three-dimensional model corresponding to a frontal image of a target head in the present invention;

[0057] Figure 3 The figure is a flow chart of generating a dynamic model sequence by using the improved spherical linear interpolation method in the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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.

[0059] A method for generating a virtual digital human, such as Figure 1 As shown, S1, obtain the front image of the target head.

[0060] S2. Based on the preset base three-dimensional model, generate an initial three-dimensional model corresponding to the frontal image of the target head, such as Figure 2 As shown, specifically including:

[0061] S21, generating contour three-dimensional model data of the target head front image according to the target head front image and the base three-dimensional model;

[0062] S22, performing alignment processing on the organs in the target head front image to generate three-dimensional model data of the organs in the target head front image;

[0063] S23, fusing the contour three-dimensional model data and the organ three-dimensional model data to generate an initial three-dimensional model corresponding to the target head frontal image.

[0064] Specifically, in S21, the contour three-dimensional model data of the target head front image is generated according to the target head front image and the base three-dimensional model, such as Figure 2 As shown, including:

[0065] S211, extracting a first feature point in the front image of the target head, and determining a second feature point corresponding to the first feature point in the base three-dimensional model;

[0066] S212: Generate contour three-dimensional model data of the target head frontal image based on the first feature point and the second feature point.

[0067] Specifically, in S212, the contour three-dimensional model data of the target head frontal image is generated according to the first feature point and the second feature point, such as Figure 2 As shown, including:

[0068] S2121, aligning the base three-dimensional model with the target head front image according to the first feature point and the second feature point to obtain an aligned base three-dimensional model;

[0069] S2122, using radial basis functions to perform interpolation processing on the aligned base three-dimensional model to generate contour three-dimensional model data of the target head frontal image.

[0070] S3. Determine a series of facial movements according to the promotional language, and generate a corresponding intermediate three-dimensional model based on the initial three-dimensional model, specifically including:

[0071] A series of facial movements are determined according to the language text and language emotion of the propaganda language, and the mapping of each facial movement is completed on the initial three-dimensional model to generate a corresponding intermediate three-dimensional model.

[0072] The above technical scheme generates contour three-dimensional model data of the target head front image based on the target head front image and the base three-dimensional model, aligns the organs in the target head front image to generate organ three-dimensional model data of the target head front image, fuses the contour three-dimensional model data and the organ three-dimensional model data to generate an initial three-dimensional model corresponding to the target head front image, determines a series of facial movements based on the language text and language emotion of the promotional language, completes the mapping of each facial movement on the initial three-dimensional model, and generates a corresponding intermediate three-dimensional model, thereby ensuring that the intermediate three-dimensional model has facial movements that are compatible with the language text and language emotion of the promotional language, effectively improving the viewer's viewing experience, helping companies create a more charismatic brand image, and attracting consumers' attention.

[0073] S4, using the improved spherical linear interpolation method to interpolate the initial 3D model and the intermediate 3D model to obtain multiple sets of interpolated 3D coordinates, such as Figure 3 As shown, specifically including:

[0074] S41, preprocessing the initial three-dimensional model and the intermediate three-dimensional model;

[0075] S42, extracting a set of feature points from each three-dimensional model, and using a feature point matching algorithm to determine the corresponding relationship between the feature points of two adjacent three-dimensional models;

[0076] S43, mapping the matched feature points of each three-dimensional model onto a high-dimensional unit sphere to convert the feature point coordinates into a corresponding set of intermediate unit vectors;

[0077] S44, selecting a series of interpolation points in the spherical space between two adjacent three-dimensional models, and obtaining multiple groups of transition unit vectors;

[0078] S45, based on the intermediate unit vector and the transition unit vector, using an improved spherical linear interpolation method to perform interpolation calculation on two adjacent three-dimensional models to obtain multiple groups of interpolation unit vectors;

[0079] S46, converting multiple groups of interpolation unit vectors into corresponding interpolation three-dimensional coordinates;

[0080] S47, repeat S44 to S46 until the interpolation calculation of the last two intermediate three-dimensional models is completed to obtain multiple sets of interpolated three-dimensional coordinates.

[0081] In the technical solution of the present application, for the improved spherical linear interpolation method: on the one hand, by introducing an angle threshold and adopting a nonlinear interpolation method to deal with the situation where the angle approaches 0; on the other hand, by optimizing the parameters of spherical linear interpolation to avoid calculating complex trigonometric functions, the calculation efficiency is improved.

[0082] 1) In S41, the initial 3D model and the intermediate 3D model are preprocessed, such as Figure 3 As shown, including:

[0083] The initial 3D model and the intermediate 3D model are preprocessed including noise removal, surface smoothing and size standardization.

[0084] 2) In S45, based on the intermediate unit vector and the transition unit vector, an improved spherical linear interpolation method is used to perform interpolation calculation on two adjacent three-dimensional models to obtain multiple groups of interpolation unit vectors, including:

[0085] S451, selecting two adjacent groups of intermediate unit vectors and the first group of transition unit vectors therebetween, and using the corresponding unit vectors in the first group of intermediate unit vectors and the first group of transition unit vectors as the starting unit vector and the target unit vector, respectively;

[0086] S452, using an improved spherical linear interpolation formula to perform interpolation calculation on the starting unit vector and the target unit vector to obtain an interpolation unit vector;

[0087] S453, repeat S451 to S452 until the interpolation calculation of the last pair of unit vectors in the corresponding relationship between the previous group of intermediate unit vectors and the first group of transition unit vectors is completed;

[0088] S454, repeat S451 to S453 until the interpolation calculation of the last group of transition unit vectors and the last pair of corresponding unit vectors in the next group of intermediate unit vectors is completed, and multiple groups of interpolation unit vectors are obtained.

[0089] Specifically, in S452, the starting unit vector and the target unit vector are interpolated by using the improved spherical linear interpolation formula to obtain the interpolated unit vector, including:

[0090] The spherical linear interpolation formula is improved by introducing an angle threshold and using a nonlinear interpolation method to deal with the situation where the angle approaches 0, and by optimizing the parameters of the spherical linear interpolation to avoid calculating complex trigonometric functions to improve the calculation efficiency:

[0091]

[0092] Among them, Slerp(q 0 ,q 1 ) is the calculated interpolation unit vector, q 0 is the starting unit vector, q 1 is the target unit vector, θ is the starting unit vector q 0 With the target unit vector q 1 , ε is the angle threshold, t is the interpolation parameter, 0≤t≤1.

[0093] S5, generating corresponding transitional three-dimensional models according to multiple sets of interpolated three-dimensional coordinates, and generating a dynamic model sequence by combining the initial three-dimensional model and the intermediate three-dimensional model, such as Figure 3 As shown, specifically including:

[0094] Corresponding transitional three-dimensional models are generated according to multiple sets of interpolated three-dimensional coordinates, and the transitional three-dimensional models are inserted into corresponding positions between the initial three-dimensional model and the intermediate three-dimensional model, and the intermediate three-dimensional models, to generate a dynamic model sequence.

[0095] S6. Perform three-dimensional rendering on the dynamic model sequence using a preset three-dimensional renderer to generate a video sequence.

[0096] The above technical scheme adopts an improved spherical linear interpolation method to interpolate the initial three-dimensional model and the intermediate three-dimensional model to obtain multiple groups of interpolated three-dimensional coordinates, generates corresponding transitional three-dimensional models according to the multiple groups of interpolated three-dimensional coordinates, and generates a dynamic model sequence by combining the initial three-dimensional model and the intermediate three-dimensional model. The dynamic model sequence is three-dimensionally rendered using a preset three-dimensional renderer to generate a video sequence. The improved spherical linear interpolation method can make the virtual digital person smoothly transform between two adjacent intermediate three-dimensional models, effectively improve the realism, further enhance the viewer's perception, play a more important role in corporate image promotion, and bring greater commercial value to the enterprise.

[0097] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating a virtual digital human, characterized in that: The following steps are involved: S1, obtain the front image of the target head; S2, generating an initial three-dimensional model corresponding to the frontal image of the target head based on a preset base three-dimensional model; S3, determining a series of facial movements according to the promotional language, and generating a corresponding intermediate three-dimensional model based on the initial three-dimensional model; S4, using an improved spherical linear interpolation method to interpolate the initial three-dimensional model and the intermediate three-dimensional model to obtain multiple sets of interpolated three-dimensional coordinates; S5, generating corresponding transitional three-dimensional models according to the multiple sets of interpolated three-dimensional coordinates, and generating a dynamic model sequence by combining the initial three-dimensional model and the intermediate three-dimensional model; S6, using a preset 3D renderer to perform 3D rendering on the dynamic model sequence to generate a video sequence; Among them, in the improved spherical linear interpolation method, on the one hand, an angle threshold is introduced and a nonlinear interpolation method is used to deal with the situation where the angle approaches 0. On the other hand, the parameters of spherical linear interpolation are optimized to avoid calculating complex trigonometric functions and improve calculation efficiency.

2. The method for generating a virtual digital human according to claim 1, characterized in that: In S2, based on the preset base three-dimensional model, an initial three-dimensional model corresponding to the frontal image of the target head is generated, including: S21, generating contour three-dimensional model data of the target head front image according to the target head front image and the base three-dimensional model; S22, performing alignment processing on the organs in the target head front image to generate three-dimensional model data of the organs in the target head front image; S23, fusing the contour three-dimensional model data and the organ three-dimensional model data to generate an initial three-dimensional model corresponding to the target head frontal image.

3. The method for generating a virtual digital human according to claim 2, characterized in that: In S21, the contour three-dimensional model data of the front image of the target head is generated according to the front image of the target head and the base three-dimensional model, including: S211, extracting a first feature point in the front image of the target head, and determining a second feature point corresponding to the first feature point in the base three-dimensional model; S212: Generate contour three-dimensional model data of the target head frontal image based on the first feature point and the second feature point.

4. The method for generating a virtual digital human according to claim 3, characterized in that: In S212, the contour three-dimensional model data of the target head front image is generated according to the first feature point and the second feature point, including: S2121, aligning the base three-dimensional model with the target head front image according to the first feature point and the second feature point to obtain an aligned base three-dimensional model; S2122, using radial basis functions to perform interpolation processing on the aligned base three-dimensional model to generate contour three-dimensional model data of the target head frontal image.

5. The method for generating a virtual digital human according to claim 1, characterized in that: In S3, a series of facial actions are determined according to the promotional language, and a corresponding intermediate 3D model is generated based on the initial 3D model, including: A series of facial movements are determined according to the language text and language emotion of the propaganda language, and the mapping of each facial movement is completed on the initial three-dimensional model to generate a corresponding intermediate three-dimensional model.

6. The method for generating a virtual digital human according to claim 1, characterized in that: In S4, an improved spherical linear interpolation method is used to interpolate the initial 3D model and the intermediate 3D model to obtain multiple sets of interpolated 3D coordinates, including: S41, preprocessing the initial three-dimensional model and the intermediate three-dimensional model; S42, extracting a set of feature points from each three-dimensional model, and using a feature point matching algorithm to determine the corresponding relationship between the feature points of two adjacent three-dimensional models; S43, mapping the matched feature points of each three-dimensional model onto a high-dimensional unit sphere to convert the feature point coordinates into a corresponding set of intermediate unit vectors; S44, selecting a series of interpolation points in the spherical space between two adjacent three-dimensional models, and obtaining multiple groups of transition unit vectors; S45, based on the intermediate unit vector and the transition unit vector, using an improved spherical linear interpolation method to perform interpolation calculation on two adjacent three-dimensional models to obtain multiple groups of interpolation unit vectors; S46, converting multiple groups of interpolation unit vectors into corresponding interpolation three-dimensional coordinates; S47, repeat S44 to S46 until the interpolation calculation of the last two intermediate three-dimensional models is completed to obtain multiple sets of interpolated three-dimensional coordinates.

7. The method for generating a virtual digital human according to claim 6, characterized in that: In S41, the initial 3D model and the intermediate 3D model are preprocessed, including: The initial 3D model and the intermediate 3D model are preprocessed including noise removal, surface smoothing and size standardization.

8. The method for generating a virtual digital human according to claim 6, characterized in that: In S45, based on the intermediate unit vector and the transition unit vector, an improved spherical linear interpolation method is used to perform interpolation calculations on two adjacent three-dimensional models to obtain multiple groups of interpolation unit vectors, including: S451, selecting two adjacent groups of intermediate unit vectors and the first group of transition unit vectors therebetween, and using the corresponding unit vectors in the first group of intermediate unit vectors and the first group of transition unit vectors as the starting unit vector and the target unit vector, respectively; S452, using an improved spherical linear interpolation formula to perform interpolation calculation on the starting unit vector and the target unit vector to obtain an interpolation unit vector; S453, repeat S451 to S452 until the interpolation calculation of the last pair of unit vectors in the corresponding relationship between the previous group of intermediate unit vectors and the first group of transition unit vectors is completed; S454, repeat S451 to S453 until the interpolation calculation of the last group of transition unit vectors and the last pair of corresponding unit vectors in the next group of intermediate unit vectors is completed, and multiple groups of interpolation unit vectors are obtained.

9. The method for generating a virtual digital human according to claim 8, characterized in that: In S452, the improved spherical linear interpolation formula is used to interpolate the starting unit vector and the target unit vector to obtain the interpolation unit vector, including: The spherical linear interpolation formula is improved by introducing an angle threshold and using a nonlinear interpolation method to deal with the situation where the angle approaches 0, and by optimizing the parameters of the spherical linear interpolation to avoid calculating complex trigonometric functions to improve the calculation efficiency: Among them, Slerp(q0,q1) is the calculated interpolation unit vector, q0 is the starting unit vector, q1 is the target unit vector, θ is the angle between the starting unit vector q0 and the target unit vector q1, ε is the angle threshold, t is the interpolation parameter, 0≤t≤1.

10. The method for generating a virtual digital human according to claim 1, characterized in that: In S5, a corresponding transitional three-dimensional model is generated according to multiple sets of interpolated three-dimensional coordinates, and a dynamic model sequence is generated by combining the initial three-dimensional model and the intermediate three-dimensional model, including: Corresponding transitional three-dimensional models are generated according to multiple sets of interpolated three-dimensional coordinates, and the transitional three-dimensional models are inserted into corresponding positions between the initial three-dimensional model and the intermediate three-dimensional model, and the intermediate three-dimensional models, to generate a dynamic model sequence.

Images