A shot script automatic generation method based on a mapping operator

Abstract

The application relates to the technical field of machine learning, and discloses a mapping operator-based automatic generation method of a split-screen script, which adopts a power attention-Transformer model to perform hierarchical analysis on a split-screen script text; a visual composition scheme is generated through a semantic visual mapping operator; wherein an improved Bezier curve method of a four-parameter Bernstein base is introduced into an action trajectory to realize flexible reconstruction and key frame extraction of complex actions, and emotion style parameters can self-adaptively adjust picture colors, light and line styles, so that a narrative atmosphere is consistent with a director's intention; finally, a complete split-screen script sequence is rendered; the application can realize efficient, accurate and style-unified split-screen script generation in film and television, animation and advertisement production, and significantly improves creation efficiency and picture expressiveness.

Description

Technical Field

[0001] This invention relates to the field of machine learning technology, and in particular to an automatic generation method for storyboard artwork based on mapping operators. Background Technology

[0002] With the increasing demands for production efficiency and quality in the film, animation, and digital media industries, more and more intelligent storyboard generation systems are being proposed and applied to actual projects. These systems can typically automatically generate basic shots or draft shots based on the input script text, improving the automation level of storyboard creation to some extent. However, existing intelligent systems still have significant shortcomings in complex scene environments: on the one hand, for scripts containing multi-character interactions, large-scale scene transitions, or highly dynamic actions, the generated shots often lack detailed depiction of environmental elements, character relationships, and action rhythm, resulting in simplistic visual expression and a lack of narrative coherence; on the other hand, in terms of emotional atmosphere or style expression, existing systems mostly rely on general templates or fixed parameters, making it difficult to adapt to the lighting, color, and rendering needs of different script contexts, resulting in deviations between the generated effects and the director's or designer's creative intentions. Therefore, how to achieve accurate mapping and controllable expression of complex semantic structures and multi-dimensional environmental factors while maintaining automated generation efficiency has become a key issue that current intelligent storyboard generation technology urgently needs to overcome. Summary of the Invention

[0003] This invention provides an automatic storyboard generation method based on mapping operators, aiming to solve the problems of insufficient narrative coherence, rough action depiction, and lack of emotional atmosphere expression in existing storyboard generation systems under complex scenes. This method uses a power-attention-Transformer model to perform layered analysis of the script text, accurately capturing environmental elements, character relationships, and emotional context in multi-character interactions, large scene transitions, and complex action sequences. It then utilizes semantic visual mapping operators to transform the analysis results into executable visual composition parameters such as camera language, character layout, action trajectories, and emotional style. This enables the system to handle group actions in war films, multi-scene transitions in urban dramas, or high-dynamic action scenes in animation. When creating a scene, the system can automatically generate storyboards that are narratively coherent, realistically animated, and stylistically consistent. The motion trajectory reconstruction method based on a four-parameter Bernstein basis allows for flexible control of the amplitude and rhythm of movements in action fight scenes or dance performance scripts. Meanwhile, the generation of emotional style parameters can adaptively adjust color and lighting effects in suspense, romance, or epic genres, ensuring consistency between the visual atmosphere and the script text. Through these designs, the invention not only improves the automation level and generation quality of storyboard production but also significantly shortens the production cycle, reduces manual labor, and enhances the consistency between the finished artwork and the director's intentions in application scenarios such as pre-production creation of film and television projects, animation storyboard drawing, and rapid production of short commercials.

[0004] This invention provides a method for automatically generating storyboard drawings based on mapping operators, characterized by the following steps:

[0005] Step S1: Constructing the semantic structure: Obtain the storyboard text sequence, construct the power attention-Transformer model, and perform hierarchical parsing of the storyboard text sequence through the power attention-Transformer model to establish the scene semantic structure; The construction method of the power attention-Transformer model is as follows: construct the Transformer model, introduce power attention to replace the Softmax attention of the Transformer model, and construct the power attention-Transformer model.

[0006] Step S2: Semantic Visual Mapping: The semantic visual mapping transformation operator is used to map the scene semantic structure into a visual composition scheme; the semantic visual mapping transformation operator consists of a shot language parameter generation unit, a character layout parameter generation unit, an action visualization parameter generation unit, an emotional style parameter generation unit, and a combined output unit;

[0007] Step S3: Visual Rendering Generation: Transform the visual composition scheme into specific storyboard frames and generate a complete storyboard sequence.

[0008] Furthermore, step S1 specifically includes the following steps:

[0009] Step S11: Segment and encode the storyboard text sequence, mapping each word element to a word vector representation to form an initial word vector matrix;

[0010] Step S12: Input the initial word vector matrix into the Transformer model to generate the context feature representation matrix of the storyboard text sequence. The Transformer model includes a self-attention sub-layer and a feedforward neural network sub-layer.

[0011] Step S13: Optimize the attention weight calculation mechanism and context dependency modeling capability of the self-attention sub-layer through power-order attention, improve the generation effect of the context feature representation matrix, and obtain the enhanced context feature representation matrix;

[0012] Step S14: Use the enhanced contextual feature representation matrix to perform hierarchical semantic modeling on the storyboard text sequence to obtain scene-level representation layer, character-level representation layer, action-level representation layer and dialogue / emotion-level representation layer, and obtain scene environment, character information, action sequence and emotion tags;

[0013] Step S15: Integrate scene environment, character information, action sequence and emotional tags to construct scene semantic structure.

[0014] Furthermore, step S13 specifically includes the following steps:

[0015] Step S131: Perform a linear transformation on the context feature representation matrix to generate the Query matrix, Key matrix, and Value matrix respectively; and use a power attention mapping function to map the Query matrix and Key matrix to generate a power mapping result; the power attention mapping function is any one of the following mapping functions: TPOW mapping, SPOW mapping, or TSPOW mapping.

[0016] Step S132: Based on the power mapping result, calculate the power attention weights of the self-attention sub-layer and weight the Value matrix to obtain the weighted context representation matrix;

[0017] Step S133: Combine the weighted context representation matrix and the power mapping result to perform context information fusion and output the enhanced context feature representation matrix.

[0018] Furthermore, step S2 specifically includes the following steps:

[0019] Step S21: The lens language parameter generation unit calculates the lens type and camera angle based on the scene environment in the scene semantic structure, and outputs the lens language parameters;

[0020] Step S22: The character layout parameter generation unit allocates the screen area according to the golden ratio principle based on the character information in the scene semantic structure, calculates the character position and orientation using the polar coordinate layout formula, and outputs the character layout parameters.

[0021] Step S23: The motion visualization parameter generation unit uses a motion trajectory reconstruction method based on a four-parameter Bernstein basis to convert the motion in the motion sequence into motion trajectory, then extracts key motion frames and outputs motion visualization parameters.

[0022] Step S24: The emotion style parameter generation unit outputs emotion style parameters based on the emotion tags in the scene semantic structure;

[0023] Step S25: The combined output unit stores the camera language parameters, character layout parameters, motion visualization parameters, and emotional style parameters as a visual composition scheme.

[0024] Furthermore, step S23 specifically includes the following steps:

[0025] Step S231: Parse the action sequence to generate continuous action instructions, and discretize the continuous action instructions into action feature points;

[0026] Step S232: Based on the action feature points, construct the action control point set, introduce the four-parameter Bernstein basis function, and add smoothness control parameters, offset control parameters, scale control parameters and curvature control parameters to the Bernstein basis to adjust the curve shape of the Bézier curve in multiple dimensions. While maintaining the geometric properties of the Bézier curve such as convex hull and endpoint interpolation, the flexibility and controllability of the curve in complex action sequence modeling are improved. Optimize the Bézier curve, construct the improved Bézier curve, and fit and interpolate the action control point set according to the improved Bézier curve to calculate the continuous trajectory of the action sequence.

[0027] Step S233: Obtain the spatial path in the form of an action curve based on the continuous trajectory of the action sequence;

[0028] Step S234: Calculate the motion amplitude based on the motion feature points, calculate the trajectory curvature change rate based on the spatial path in the form of the motion curve, perform differential sampling on the continuous trajectory of the motion sequence based on the joint discrimination of motion amplitude and trajectory curvature change rate, generate differential motion trajectory, and extract key motion frames.

[0029] Step S235: Fuse and transform the differential motion trajectory with the key action frames to generate motion visualization parameters.

[0030] By adopting the above solution, the beneficial effects achieved by the present invention are as follows:

[0031] This invention introduces a power-attention-Transformer model to achieve deep analysis of script text and capture long-distance dependencies, improving semantic understanding capabilities in multi-character interactions, large scene transitions, and complex narrative contexts. It also solves the problem of insufficient depiction of scene environment, character relationships, and emotional atmosphere in existing storyboard generation methods. With the help of this model, this invention can generate storyboards with more coherent narratives and more complete logic, thereby enhancing the accuracy and completeness of the visuals in the mapping of script intent.

[0032] This invention achieves automatic transformation from semantic structure to visual composition through a semantic visual mapping operator, effectively improving the parametric expression capabilities of camera language, character layout, action trajectory, and emotional style. It solves the limitations of previous systems that generated monotonous images and failed to reflect the director's creative intentions. In specific applications, the system can automatically generate storyboard schemes that conform to visual aesthetics and narrative logic for different scene requirements such as war films, urban dramas, or animations, significantly enhancing the practical value of storyboards in terms of scene adaptability and expressiveness.

[0033] This invention achieves further breakthroughs in action and style expression: by using a motion trajectory reconstruction method based on a four-parameter Bernstein basis, it enables flexible control of complex actions and accurate extraction of keyframes, improving the trajectory reproduction accuracy and rhythm in high-dynamic action scenes such as fighting and dancing; by adaptively generating emotional style parameters, it solves the lack of flexibility in the expression of color, lighting, and lines in existing systems, enhancing the ability of storyboards to create atmosphere in suspense, romance, or epic scripts; the above designs not only improve the efficiency and quality of storyboard generation, but also shorten the pre-production cycle of film, animation, and advertising production, reduce labor input costs, and provide creators with a more efficient, reliable, and expressive creative tool. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the semantic visual mapping transformation operator proposed in step S2 of Embodiment 1;

[0035] Figure 2 This is a trend chart of dimensional changes for the power mapping method proposed in Examples 3, 4, and 5.

[0036] Figure 2 The figure shows the trend of the mapping dimension of three power mapping methods, TPOW, SPOW and TSPOW, when the input dimension d=64 and tile=8 increases with the power order p. In the figure, the blue dotted line represents the TPOW mapping dimension, the orange square line represents the SPOW mapping dimension, and the green triangle line represents the TSPOW mapping dimension; the horizontal axis (X-axis) is the power order p; the vertical axis (Y-axis) is the mapping dimension (logarithmic scale). Detailed Implementation

[0037] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0038] Example 1, according to Figure 1 This invention provides a method for automatically generating storyboard drawings based on mapping operators, characterized by the following steps:

[0039] Step S1: Constructing the semantic structure: Obtain the storyboard text sequence, construct the power attention-transformer model, and use the power attention-transformer model to perform hierarchical parsing of the storyboard text sequence to extract scene environment, character list, action sequence and emotional tags, and establish the scene semantic structure; The construction method of the power attention-transformer model is as follows: construct the Transformer model, introduce power attention to replace the Softmax attention of the Transformer model, and construct the power attention-transformer model;

[0040] Step S2: Semantic Visual Mapping: The semantic visual mapping transformation operator is used to map the scene semantic structure into a visual composition scheme. The visual composition scheme includes camera language parameters, character layout parameters, motion visualization parameters, and emotional style parameters. The semantic visual mapping transformation operator consists of a camera language parameter generation unit, a character layout parameter generation unit, a motion visualization parameter generation unit, an emotional style parameter generation unit, and a combined output unit.

[0041] Step S3: Visual Rendering Generation: Transform the visual composition scheme into specific storyboard frames and generate a complete storyboard sequence.

[0042] Example 2, this example is based on Example 1. In this example, step S1 specifically includes the following steps:

[0043] Step S11: Segment and encode the storyboard text sequence, mapping each word element to a word vector representation to form an initial word vector matrix;

[0044] Step S12: Input the initial word vector matrix into the Transformer model to generate the context feature representation matrix of the storyboard text sequence. The Transformer model includes a self-attention sub-layer and a feedforward neural network sub-layer, and achieves the layer-by-layer extraction and transmission of multi-layer semantic features under the action of residual connection and normalization mechanism.

[0045] Step S13: Replace traditional Softmax attention with power-law attention to optimize the attention weight calculation mechanism and context dependency modeling capability of the self-attention sub-layer, improve the generation effect of the context feature representation matrix, and obtain an enhanced context feature representation matrix; power-law attention performs polynomial expansion on the similarity between query and key through power-law mapping operator, avoiding the normalization operation of exponential function, which can expand the scale of state representation under long sequence conditions, improve the ability to capture long-distance dependencies, and maintain a high weight-state balance in the process of weight calculation and state update, thereby achieving efficient modeling of long context semantics;

[0046] Step S14: Use the enhanced contextual feature representation matrix to perform hierarchical semantic modeling on the storyboard text sequence to obtain scene-level representation layer, character-level representation layer, action-level representation layer and dialogue / emotion-level representation layer, and obtain scene environment, character information, action sequence and emotion tags;

[0047] At the scene-level representation layer, time, location, lighting, and background elements in the text are parsed to form the scene environment;

[0048] In the role-level representation layer, the list of roles and attribute information appearing in the text are identified to obtain role information;

[0049] At the action-level representation layer, extract the character's action sequence;

[0050] At the dialogue / emotion level representation layer, the dialogue content of the characters and their corresponding emotional tags are captured in combination with the context.

[0051] Step S15: Integrate scene environment, character information, action sequence and emotional tags to construct scene semantic structure.

[0052] Example 3, according to Figure 2 This embodiment is based on Embodiment 2. In this embodiment, step S13 specifically includes the following steps:

[0053] Step S131: Perform a linear transformation on the context feature representation matrix to generate the Query matrix, Key matrix, and Value matrix respectively; and use the power attention mapping function to map the Query matrix and Key matrix to generate the power mapping result; the power attention mapping function selected is TPOW mapping;

[0054] The power mapping result is a new matrix obtained after the query-key similarity matrix is ​​processed by the power mapping function. It can change the attention distribution, making the model more biased towards highlighting strong correlations or enhancing weak correlations.

[0055] TPOW mapping: The mapping is implemented using Tensor Power of Word (TPOW) expansion, defined as follows:

[0056] , ;

[0057] in, Indicates the power order. This represents the input vector, with dimension . ; , These represent the Query and Key vectors, respectively. Represents the power mapping function, Represents the input vector conduct The mapping result obtained by the power expansion of the tensor of order 1 has a mapping dimension of 1. It includes the components of the vector in All product combinations in the higher-order space can be used to explicitly model the interaction relationships between higher-order features; This represents the dimension of the vector space after the TPOW mapping; This represents the transpose of the query vector. This represents the inner product of the Query and Key in the power-law mapping space; The inner product is calculated in the original space and taken. Power;

[0058] The advantage is that it is simple in form and easy to calculate, but it has the problem of high dimensional redundancy.

[0059] Step S132: Based on the power-order mapping result, calculate the power-order attention weights of the self-attention sub-layer and weight the Value matrix to obtain the weighted context representation matrix. The formula used is as follows:

[0060] ;

[0061] in, This indicates that under the power-law attention mechanism, for the first input sequence... At each position, the power-law similarity between the Query and all historical keys is calculated, and the corresponding value vectors are weighted and combined to obtain the final weighted context representation matrix. Indicates the first A key vector at each position, No. Value vectors at each position; This represents the inner product similarity between the Query and the Key, measuring the similarity of the first and second parts. The position and the first The strength of the correlation between the positions; This represents the weights after applying a power-law mapping to the inner product similarity;

[0062] Step S133: Combine the weighted context representation matrix and the power mapping result to perform context information fusion and output the enhanced context feature representation matrix.

[0063] Step S2 specifically includes the following steps:

[0064] Step S21: The lens language parameter generation unit calculates the lens type and camera angle based on the scene environment in the scene semantic structure, and outputs the lens language parameters. The lens language parameters include lens category (long shot, medium shot, close shot), lens angle (overhead shot, low shot, eye-level shot) and movement mode (push, pull, pan, tilt).

[0065] Step S22: The character layout parameter generation unit allocates the screen area according to the golden ratio principle based on the character information in the scene semantic structure, calculates the character position and orientation using the polar coordinate layout formula, and outputs the character layout parameters, including the character's coordinate position, orientation, and foreground and background layers.

[0066] Step S23: The motion visualization parameter generation unit uses a motion trajectory reconstruction method based on a four-parameter Bernstein basis to convert the motion in the motion sequence into a motion trajectory, then extracts key motion frames, and outputs motion visualization parameters. The motion visualization parameters include the character's motion path, key frame timing, and motion amplitude, which are used to drive the dynamic performance of subsequent storyboard drawings. The construction method of the motion trajectory reconstruction method based on a four-parameter Bernstein basis is as follows: by introducing four types of control parameters—smoothing, offset, scale, and curvature—into the traditional Bernstein basis function, the multi-dimensional controllability model of the motion control point set is performed. While maintaining the convex hull property and endpoint interpolation property of the Bézier curve, the fine reconstruction and flexible adjustment of the motion sequence trajectory are achieved.

[0067] Step S24: The emotional style parameter generation unit outputs emotional style parameters based on the emotional tags in the scene semantic structure. The emotional style parameters include the color tone of the picture, the distribution of light intensity, and the line rendering style.

[0068] Step S25: The combined output unit stores the camera language parameters, character layout parameters, motion visualization parameters, and emotional style parameters as a visual composition scheme.

[0069] Step S23 specifically includes the following steps:

[0070] Step S231: Parse the action sequence to generate continuous action instructions, and discretize the continuous action instructions into action feature points. The action feature points include timestamps, spatial locations, and action amplitude information.

[0071] Step S232: Based on the action feature points, construct the action control point set, introduce the four-parameter Bernstein basis function, and add smoothness control parameters, offset control parameters, scale control parameters, and curvature control parameters to the Bernstein basis to adjust the shape of the Bézier curve in multiple dimensions. While maintaining the geometric properties of the Bézier curve, such as convex hull and endpoint interpolation, improve the flexibility and controllability of the curve in complex action sequence modeling, optimize the Bézier curve, construct an improved Bézier curve, and fit and interpolate the action control point set according to the improved Bézier curve to calculate the continuous trajectory of the action sequence. The formula used is as follows:

[0072] Formula for Bézier curves on a four-parameter Bernstein basis:

[0073] , z∈[0,1] ;

[0074] in, This indicates the smoothing control parameter (which affects the smoothness of the curve). This indicates the offset control parameter (the offset of the control curve relative to the convex hull). This indicates the scale control parameter (which controls local stretching or compression). This represents the curvature control parameter, which adjusts the degree of curvature of the curve; These represent curve parameters and control the curve's progress. Indicates the order of the curve; This represents an improved Bézier curve based on four parameters. Indicates the index of the control point. Indicates the first A number of control points are used to determine the geometric position of the curve; Represents the four-parameter Bernstein basis functions. This represents the curvature correction function, used to describe the additional adjustment capability of the basis functions in the direction of curvature change;

[0075] Step S233: Based on the continuous trajectory of the action sequence, obtain the spatial path in the form of an action curve. The spatial path in the form of an action curve has stronger shape controllability and robustness compared to the traditional Bézier trajectory.

[0076] Step S234: Calculate the motion amplitude based on the motion feature points, calculate the trajectory curvature change rate based on the spatial path in the form of the motion curve, perform differential sampling on the continuous trajectory of the motion sequence based on the joint discrimination of motion amplitude and trajectory curvature change rate, generate differential motion trajectory, and extract key motion frames; key motion frames are used to represent the main turning points, amplitude peaks and rhythm nodes of the motion.

[0077] Step S235: Fuse and transform the differential motion trajectory with the key action frames to generate motion visualization parameters for subsequent storyboard composition and rendering.

[0078] Example 4, according to Figure 2 This embodiment is based on embodiment two. The difference between this embodiment and embodiment three is that the power-order attention mapping function uses SPOW mapping.

[0079] SPOW mapping: The mapping is implemented using Symmetric Power of Word (SPOW) expansion, defined as follows:

[0080] , ;

[0081] in, Represents the input vector conduct The mapping result obtained from the power expansion of the symmetric tensor. Denotes the binomial coefficient, indicating the use of The number of times the variables are composed is The number of independent monomials contained in a symmetric polynomial; Let represent the dimension of the vector space after the SPOW mapping, which is equal to the binomial coefficients mentioned above. This mapping significantly reduces the expansion dimension compared to TPOW by eliminating redundant terms caused by different permutations, and ensures that the inner product calculation in the mapped space is equal to the dot product in the original space. Strict equivalence of powers;

[0082] This approach significantly reduces the dimensionality of the mapping results by constraining the multiple indexes to be non-descending and eliminating redundant combinations, thus avoiding the high redundancy problem of TPOW. Its advantages are higher storage and computation efficiency, while its disadvantages are a more complex structure and slightly less hardware-friendly performance.

[0083] Example 5, according to Figure 2 This embodiment is based on embodiment two. The difference between this embodiment and embodiment three is that the power-order attention mapping function uses TSPOW mapping.

[0084] TSPOW mapping: The mapping is implemented using Tiled Symmetric Power of Word (TSPOW) expansion, dividing the input vector into several tiles. Within each tile, SPOW is used to maintain the non-decreasing index constraint, while a GPU-friendly parallel structure is employed between tiles.

[0085] , ;

[0086] in, Represents the input vector conduct The mapping result obtained by the block-symmetric tensor power expansion is that the input vector is first divided into several sub-blocks, and symmetric tensor power expansion is used within the sub-blocks to remove redundancy in the arrangement. The sub-blocks are expanded by a parallel and friendly combination method. This represents the dimension size of a single tile (sub-tile). This represents the dimension of the high-dimensional feature vector obtained by the TSPOW mapping;

[0087] This approach combines the dimensionality compression advantage of SPOW with the hardware friendliness of TPOW, improving execution efficiency on GPUs while ensuring computational equivalence.

[0088] The three mapping methods in Examples 3, 4, and 5 all satisfy the power-law kernel equivalence relation, but they differ in terms of mapping dimension, storage overhead, and hardware adaptability: TPOW is simple in form but has high dimensional redundancy, SPOW has strong dimensional compression capability but relatively high computational complexity, and TSPOW balances dimensional compression and GPU parallel execution efficiency while ensuring mapping equivalence.

[0089] The present invention and its embodiments have been described above. This description is not restrictive. The accompanying drawings are only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present invention, such design should fall within the protection scope of the present invention.

Claims

1. A method for automatically generating storyboard drawings based on mapping operators, characterized in that: Includes the following steps: Step S1: Obtain the storyboard text sequence, construct the power attention-transformer model, and perform hierarchical parsing of the storyboard text sequence through the power attention-transformer model to establish the scene semantic structure; Step S2: Use a semantic-visual mapping transformation operator to map the scene semantic structure into a visual composition scheme; The semantic visual mapping transformation operator consists of a shot language parameter generation unit, a character layout parameter generation unit, an action visualization parameter generation unit, an emotional style parameter generation unit, and a combined output unit; Step S3: Convert the visual composition scheme into storyboard frames and generate a storyboard sequence; The construction method of the power-attention-Transformer model is as follows: construct a Transformer model, introduce power-attention to replace the softmax attention of the Transformer model, and thus construct the power-attention-Transformer model. Step S2 specifically includes the following steps: Step S21: The lens language parameter generation unit calculates the lens type and camera angle based on the scene environment and outputs the lens language parameters; Step S22: The character layout parameter generation unit outputs character layout parameters based on the character information; Step S23: The motion visualization parameter generation unit uses a motion trajectory reconstruction method based on a four-parameter Bernstein basis to convert the motion in the motion sequence into motion trajectory, then extracts key motion frames and outputs motion visualization parameters. Step S24: The emotion style parameter generation unit outputs emotion style parameters based on the emotion tags; Step S25: The combined output unit stores the camera language parameters, character layout parameters, motion visualization parameters, and emotional style parameters as a visual composition scheme.

2. The method for automatically generating storyboard drawings based on mapping operators according to claim 1, characterized in that: Step S1 specifically includes the following steps: Step S11: Segment and encode the storyboard text sequence to form an initial word vector matrix; Step S12: Input the initial word vector matrix into the Transformer model to generate the context feature representation matrix of the storyboard text sequence. The Transformer model includes a self-attention sub-layer. Step S13: Optimize the attention weight calculation mechanism of the self-attention sub-layer through power-order attention to improve the generation effect of the context feature representation matrix and obtain the enhanced context feature representation matrix; Step S14: Use the enhanced contextual feature representation matrix to perform hierarchical semantic modeling on the storyboard text sequence to generate scene environment, character information, action sequence and emotional tags; Step S15: Integrate scene environment, character information, action sequence and emotional tags to construct scene semantic structure.

3. The method for automatically generating storyboard drawings based on mapping operators according to claim 2, characterized in that: Step S13 specifically includes the following steps: Step S131: Perform a linear transformation on the context feature representation matrix to generate the Query matrix, Key matrix, and Value matrix respectively; and use the power-law attention mapping function to map the Query matrix and Key matrix to generate the power-law mapping result; Step S132: Based on the power mapping result, calculate the power attention weights of the self-attention sub-layer and weight the Value matrix to obtain the weighted context representation matrix; Step S133: Combine the weighted context representation matrix and the power mapping result to perform context information fusion and output the enhanced context feature representation matrix.

4. The method for automatically generating storyboard drawings based on mapping operators according to claim 3, characterized in that: The power-order attention mapping function is any one of the following: TPOW mapping, SPOW mapping, or TSPOW mapping.

5. The method for automatically generating storyboard drawings based on mapping operators according to claim 1, characterized in that: Step S23 specifically includes the following steps: Step S231: Parse the action sequence to generate continuous action instructions, and discretize the continuous action instructions into action feature points; Step S232: Based on the action feature points, construct the action control point set, introduce the four-parameter Bernstein basis function, optimize the Bézier curve, construct the improved Bézier curve, fit and interpolate the action control point set according to the improved Bézier curve, and calculate the continuous trajectory of the action sequence. Step S233: Obtain the spatial path in the form of an action curve based on the continuous trajectory of the action sequence; Step S234: Calculate the motion amplitude based on the motion feature points, calculate the trajectory curvature change rate based on the spatial path in the form of the motion curve, perform differential sampling on the continuous trajectory of the motion sequence based on the joint discrimination of motion amplitude and trajectory curvature change rate, generate differential motion trajectory, and extract key motion frames. Step S235: Fuse and transform the differential motion trajectory with the key action frames to generate motion visualization parameters.

Images