An AIGC-driven animation generation method for a virtual image scene

Abstract

The application discloses a kind of AIGC driving animation generation methods for virtual image scene, specifically related to three-dimensional character animation generation and rendering technical field, for solving the problem that the animation data generated by existing AIGC cannot directly drive high-quality rendering effect due to lack of physical and semantic parameters;By obtaining the initial skeletal animation data generated by AIGC and generating instruction text, analyze the animation data to calculate the physical property parameter set, and obtain the expected physical parameter threshold range based on instruction text analysis for abnormal identification;Then, according to the time sequence variation characteristics of physical properties, the animation is divided into multiple action stages and the rendering effect strategy for each stage is matched;Finally, for each animation frame, combined with its stage strategy, specific physical parameters and abnormal identification, generate visual enhancement effect control parameters adapted to downstream rendering engine, so as to realize the automation and intelligentization coupling of animation generation and rendering link.

Description

Technical Field

[0001] This invention relates to the field of 3D character animation generation and rendering technology, and more specifically, to an AIGC-driven animation generation method for virtual image scenes. Background Technology

[0002] In the production process of virtual video content, the generation of 3D character animation and the final rendering of the visuals are two core and closely linked stages. Artificial intelligence-generated content (AIGC) technology can be introduced to automatically drive character animation, aiming to improve production efficiency. This type of AIGC-driven method can automatically generate skeletal animation sequences for characters based on high-level instructions such as text.

[0003] However, existing AIGC-based animation generation methods primarily focus on mimicking or synthesizing the form and rhythm of character movements. The animation data generated by this approach is essentially a collection of traditional motion trajectories and posture information. This leads to an inherent data-level gap between the AIGC animation generation stage and the downstream high-quality rendering stage that relies on physical and semantic parameters. Since the generation process does not cover the internal representation of the physical properties and potential visual effects inherent in the action, its output cannot directly and effectively drive many complex effects that enhance realism in the rendering pipeline. This causes the automated process to be interrupted at the critical final stage, severely restricting the end-to-end application value of AIGC technology in high-quality virtual image production. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides an AIGC-driven animation generation method for virtual image scenes to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: An AIGC-driven animation generation method for virtual image scenes includes: S1: Obtain the initial skeletal animation data and corresponding generation instruction text generated by AIGC; S2: Analyze the motion trajectory of predefined bone nodes in the initial skeletal animation data and calculate the set of physical attribute parameters associated with each animation frame; S3: Parse the generated instruction text and determine the expected physical parameter threshold range that matches the generated instruction text based on the preset mapping relationship; S4: Compare the set of physical property parameters with the expected physical parameter threshold range to identify abnormal physical parameters in the set of physical property parameters; S5: Based on the temporal variation characteristics of the physical property parameter set, the initial skeletal animation data is divided into multiple continuous action stages, and a corresponding stage-level rendering effect strategy is matched for each action stage. S6: For each animation frame, based on the stage-level rendering effect strategy of its respective action stage, and combined with the specific physical attribute parameter values ​​and abnormal physical parameters of the animation frame, generate visual enhancement effect control parameters adapted to the downstream rendering engine.

[0006] Furthermore, S1 includes: Receive initial skeletal animation data from the animation generation interface; Receive the generated instruction text from the instruction input interface; The initial skeletal animation data and the generated instruction text are associated and bound together based on the same task identifier that accompanies the initial skeletal animation data and the generated instruction text.

[0007] Furthermore, S2 includes: Extract the position and rotation data of predefined bone nodes in consecutive animation frames from the initial skeletal animation data; Based on the changes in position data between consecutive animation frames, calculate the linear velocity and linear acceleration of the skeletal nodes; The angular velocity of the bone nodes is calculated based on the changes in rotation data between consecutive animation frames; For each animation frame, the calculated linear velocity, linear acceleration, and angular velocity are assembled into a set of physical property parameters for that animation frame.

[0008] Furthermore, S3 includes: Extract keywords describing actions from the generated instruction text; The keywords are matched with a preset action type rule base to determine the target action type corresponding to the generated instruction text; Based on a preset mapping table between action types and physical parameter threshold ranges, query the expected physical parameter threshold range associated with the target action type.

[0009] Furthermore, the preset mapping relationship is established in the following way: collect training animation data containing multiple action types and their associated action description text; extract physical attribute parameters from the training animation data and statistically analyze their distribution; associate the action description text with the corresponding physical attribute parameter distribution to construct a mapping table between action type and physical parameter threshold range.

[0010] Furthermore, S4 includes: Compare each parameter value in the set of physical property parameters with the corresponding parameter threshold range in the expected physical parameter threshold range; Determine whether the parameter value falls within the corresponding parameter threshold range; Parameter values ​​that do not fall within the corresponding parameter threshold range are identified as abnormal physical parameters.

[0011] Furthermore, S5 includes: Extract energy features characterizing motion intensity from a set of physical property parameters and generate energy change curves; The boundary frames of the action phase are determined based on the local extreme points in the energy change curve and the preset energy threshold. The initial skeletal animation data time series is divided into multiple consecutive time periods based on the boundary frames, and each time period is defined as an action phase; Based on the statistical characteristics of the physical attribute parameter set within each action stage, the corresponding stage-level rendering effect strategy is matched from the preset stage-level rendering effect strategy library.

[0012] Furthermore, each strategy in the stage-level rendering effect strategy library includes: an effect type identifier, an effect intensity baseline range, and the target bone node set to which the corresponding strategy applies; matching the corresponding stage-level rendering effect strategy includes: calculating the statistical characteristics of the physical attribute parameter set within the action stage on the target bone node set; matching the statistical characteristics with the intensity baseline range of each strategy in the strategy library, and selecting the stage-level rendering effect strategy with the highest matching degree.

[0013] Furthermore, S6 includes: For the currently processed animation frame, obtain the stage-level rendering effect strategy corresponding to the action stage to which the animation frame belongs; Based on the effect intensity baseline range defined in the stage-level rendering effect strategy and the specific physical property parameter values ​​of the current animation frame, the preliminary effect intensity value of the current animation frame is obtained through interpolation calculation. Determine whether the specific physical property parameter values ​​of the current animation frame are marked as abnormal physical parameters; If it is marked as an abnormal physical parameter, the initial effect intensity value is adjusted according to the preset abnormal parameter correction rules to obtain the final effect intensity value; Generate visual enhancement effect control parameters that include effect type identifier and intensity parameter value based on the final effect intensity value.

[0014] Furthermore, the abnormal parameter correction rule is as follows: when the specific physical attribute parameter value of the animation frame is marked as an abnormal physical parameter, the corresponding correction coefficient will be selected from the preset correction coefficient table according to the parameter type to which the abnormal physical parameter belongs; the preliminary effect intensity value will be multiplied by the correction coefficient to obtain the adjusted effect intensity value as the final effect intensity value.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. By introducing a complete physical property analysis and rendering strategy mapping process, the data gap between AIGC-generated animation and high-quality rendering pipeline is effectively bridged. The system automatically calculates the physical property parameters contained in the original skeletal animation sequence and compares and verifies them with the expected physical performance derived from the semantics of text instructions. This enables the system to identify physical anomalies in the generated animation that may not conform to the high-level intent. Based on the temporal changes in motion energy, the system intelligently divides continuous animation into action stages with different intensity characteristics and matches the most suitable rendering effect theme strategy for each stage. This gives the AIGC output, which was originally just motion data, a layered and structured physical and semantic annotation.

[0016] 2. For each frame of animation, control parameters are generated to be precisely adapted to the downstream rendering engine. Taking into account the rendering theme of the stage, the physical state of the current frame, and the anomaly verification results, the parameters are dynamically calculated through interpolation and correction mechanisms. Finally, a set of visual effect-driven data that is closely coupled with the physical logic and artistic intent of the animation content itself is output. This directly improves the usability of AIGC-generated animations in professional virtual image production processes, realizes end-to-end automated generation from text instructions to final renderable animation assets, and significantly improves the efficiency and quality of content production. Attached Figure Description

[0017] Figure 1 This is a flowchart of an AIGC-driven animation generation method for virtual image scenes according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example: Figure 1 This invention presents an AIGC-driven animation generation method for virtual image scenes, comprising: S1: Obtain the initial skeletal animation data and corresponding generation instruction text generated by AIGC; S2: Analyze the motion trajectory of predefined bone nodes in the initial skeletal animation data and calculate the set of physical attribute parameters associated with each animation frame; S3: Parse the generated instruction text and determine the expected physical parameter threshold range that matches the generated instruction text based on the preset mapping relationship; S4: Compare the set of physical property parameters with the expected physical parameter threshold range to identify abnormal physical parameters in the set of physical property parameters; S5: Based on the temporal variation characteristics of the physical property parameter set, the initial skeletal animation data is divided into multiple continuous action stages, and a corresponding stage-level rendering effect strategy is matched for each action stage. S6: For each animation frame, based on the stage-level rendering effect strategy of its respective action stage, and combined with the specific physical attribute parameter values ​​and abnormal physical parameters of the animation frame, generate visual enhancement effect control parameters adapted to the downstream rendering engine.

[0020] S1: Obtain the initial skeletal animation data generated by AIGC and the corresponding generation instruction text. The specific implementation is as follows: When implementing the step of acquiring the initial skeletal animation data and corresponding generation instruction text generated by AI-generated content technology, the operation of receiving the initial skeletal animation data from the animation generation interface is performed. The animation generation interface is a predefined program call point configured to listen for network ports or message queue channels that output animation data from external AI-generated content technology systems. After the external system completes animation generation, it encapsulates the generated animation data into data packets and sends them to the network port listened to by the animation generation interface via a network protocol, or writes the data packets to a specific message queue subscribed to by the animation generation interface. The received initial skeletal animation data is expressed in terms of content as a sequence of poses of the virtual character skeleton at consecutive points in time. The specific format of the initial skeletal animation data is structured data containing hierarchical bone node definitions and keyframe sequences. The receiving process includes a verification operation for the arriving data packets, which calculates the checksum of the data packet and compares it with the checksum transmitted in the data packet. For data packets with matching checksums, a decapsulation operation is performed to parse out the structured skeletal animation sequence, and the parsed skeletal animation sequence is stored in memory for access in subsequent steps.

[0021] This process receives the generated instruction text from the instruction input interface. The instruction input interface is a program call point for receiving text input; it can be a graphical user interface (GUI) component that provides text input functionality, combining the characters typed by the user into a string. The generated instruction text is a natural language string describing the expected animation content. The receiving process includes reading the complete string of user input from the GUI component's buffer and performing preprocessing on the read string, including removing leading and trailing whitespace characters.

[0022] The process involves associating and binding the initial skeletal animation data and the generated instruction text based on the same task identifier accompanying them. This same task identifier is a unique code generated when an animation generation task is initiated, used to identify all associated data for that task. Upon task initiation, the task scheduling module generates a unique string as the same task identifier. This task scheduling module sends the same task identifier along with the user-submitted generated instruction text to an external AI-generated content technology system. When returning the initial skeletal animation data, the external AI-generated content technology system fills in the same unique string in the header field of the data packet corresponding to the same task identifier. The animation generation interface extracts the string from the header field of the received data packet, denoted as identifier A. The instruction input interface extracts the string from the metadata field of the sent request, denoted as identifier B, when receiving text. The association and binding operation compares the character sequences of identifier A and identifier B. If the character sequences of identifier A and identifier B are completely identical, a new record is created in the association database, with identifier A as the record key, and the corresponding initial skeletal animation data and generated instruction text stored as a pair of data under this record key. If the character sequences of identifier A and identifier B are inconsistent, the received initial skeletal animation data or generation instruction text is discarded. Through association binding operations, it is ensured that the initial skeletal animation data output by each animation generation task accurately matches the generation instruction text that triggered the task.

[0023] S2: Analyze the motion trajectory of predefined bone nodes in the initial skeletal animation data, and calculate the set of physical attribute parameters associated with each animation frame. Specifically, this is implemented as follows: When analyzing the motion trajectories of predefined bone nodes in the initial skeletal animation data and calculating the set of physical attribute parameters associated with each animation frame, the operation of extracting the position and rotation data of the predefined bone nodes in consecutive animation frames from the initial skeletal animation data is performed. Predefined bone nodes refer to a set of key bone nodes pre-selected in the virtual character's skeletal hierarchy for motion analysis. Key bone nodes may include the character's root bone node, pelvic bone node, left and right ankle bone nodes, left and right wrist bone nodes, and head bone node. The extraction operation involves traversing each frame of animation data contained in the initial skeletal animation data. For the currently traversed animation frame, for each predefined bone node, the position and rotation data of that bone node are read from the currently traversed animation frame data. The position data is a numerical value expressing the coordinates of the bone node in three-dimensional space. The position data contains three floating-point components, representing the coordinate values ​​in the X-axis, Y-axis, and Z-axis directions of the three-dimensional coordinate system, respectively. The unit of the coordinate values ​​is meters. Rotation data represents the orientation of a bone node in three-dimensional space. Rotation data can be represented as a quaternion, consisting of four floating-point components; or as Euler angles, which are three angle values ​​representing rotation around the X, Y, and Z axes of the three-dimensional coordinate system, measured in degrees. The position and rotation data extracted for each predefined bone node are arranged in chronological order according to the animation frames, forming a position and rotation data sequence corresponding to each predefined bone node.

[0024] Based on the changes in position data between consecutive animation frames, the linear velocity and linear acceleration of bone nodes are calculated. The operation to calculate linear velocity involves iterating through the position data sequence corresponding to each predefined bone node. For the currently iterated animation frame, the position coordinates of the bone node in the current frame are obtained using the current frame's index value and recorded as the current frame's position coordinates. The position coordinates of the same bone node in the immediately preceding animation frame are also obtained and recorded as the previous frame's position coordinates. The differences between the current frame's position coordinates and the previous frame's position coordinates on each coordinate axis component are calculated to obtain the displacement vector from the previous frame to the current frame. The frame time interval of the animation is obtained; the frame time interval is the time difference between two adjacent animation frames, and the unit of the frame time interval is seconds. The frame time interval can be read from the metadata of the initial bone animation data; for example, when the animation frame rate is 30 frames per second, the frame time interval is 1 divided by 30 seconds. Each component of the displacement vector is divided by the frame time interval, and the resulting vector is the linear velocity of the bone node in the current animation frame. The linear velocity is a three-dimensional vector, and the unit of linear velocity is meters per second. The operation of calculating linear acceleration involves further iterating through the predefined bone nodes and calculating the linear velocity sequence for each predefined bone node. For the currently traversed animation frame, the linear velocity of the bone node in the current animation frame is obtained using the index value of the current frame, and denoted as the linear velocity of the current frame; the linear velocity of the same bone node in the immediately preceding animation frame is also obtained, and denoted as the linear velocity of the preceding frame. The difference between the linear velocity of the current frame and the linear velocity of the preceding frame is calculated for each component, resulting in a velocity change vector from the preceding frame to the current frame. Dividing each component of the velocity change vector by the frame time interval yields the linear acceleration of the bone node in the current animation frame. Linear acceleration is a three-dimensional vector, and its unit is meters per second squared.

[0025] The angular velocity of a bone node is calculated based on the changes in rotation data between consecutive animation frames. The operation for calculating angular velocity involves iterating through the rotation data sequence corresponding to each predefined bone node, assuming the rotation data is represented in Euler angles. For the currently iterated animation frame, the Euler angles of the bone node in the current frame are obtained using the current frame's index. These Euler angles include the angle values ​​around the X-axis, Y-axis, and Z-axis. The Euler angles of the same bone node in the immediately preceding animation frame are also obtained, denoted as the previous frame's X-axis, Y-axis, and Z-axis angles. The change in angle around the X-axis is calculated as the current frame's X-axis angle minus the previous frame's X-axis angle; the change in angle around the Y-axis is calculated as the current frame's Y-axis angle minus the previous frame's Y-axis angle; and the change in angle around the Z-axis is calculated as the current frame's Z-axis angle minus the previous frame's Z-axis angle. An angle difference threshold is set to adjust for periodic angle jumps. This threshold is based on the periodic representation range of Euler angles. Euler angles in 3D animation are typically represented as a range of -180 degrees to 180 degrees or 0 to 360 degrees. When the absolute value of the angle change calculated between two consecutive frames exceeds 180 degrees, it indicates a possible jump caused by angle wrapping rather than actual physical rotation. Therefore, setting the angle difference threshold to 180 degrees triggers adjustment logic, correcting the angle change to a reasonable range of -180 to 180 degrees, thus accurately reflecting the true rotation direction and amplitude of the skeletal nodes. When the absolute value of the calculated change in angle around the X-axis exceeds the angle difference threshold, the change in angle around the X-axis is adjusted. The adjustment is achieved by subtracting the absolute value of the change in angle around the X-axis from 360 degrees to obtain the absolute value of the adjustment. Then, based on the original sign of the change in angle around the X-axis, the absolute value of the adjustment is assigned the opposite sign to obtain the adjusted change in angle around the X-axis. The same judgment and adjustment operation based on the angle difference threshold is performed on the changes in angle around the Y-axis and Z-axis. The adjusted changes in angle around the X-axis, Y-axis, and Z-axis are divided by the frame time interval, yielding the angular velocity components around the X-axis, Y-axis, and Z-axis, respectively, in degrees per second. These angular velocity components together constitute a three-dimensional angular velocity vector representing the rotational motion of the bone node in the current animation frame.

[0026] For each animation frame, the calculated linear velocity, linear acceleration, and angular velocity are assembled into a set of physical attribute parameters for that animation frame. The assembly operation involves creating a data container as the set of physical attribute parameters for each animation frame in the initial skeletal animation data. All predefined bone nodes are traversed. For the currently traversed predefined bone node, the linear velocity vector belonging to the current animation frame is extracted from the linear velocity sequence calculated for that predefined bone node; the linear acceleration vector belonging to the current animation frame is extracted from the linear acceleration sequence calculated for that predefined bone node; and the 3D angular velocity vector belonging to the current animation frame is extracted from the 3D angular velocity vector sequence calculated for that predefined bone node. The extracted linear velocity vector, linear acceleration vector, and 3D angular velocity vector are stored as key-value pairs in the physical attribute parameter set of the current animation frame. The key is a unique identifier for the current predefined bone node, and the value is a data combination containing the linear velocity vector, linear acceleration vector, and 3D angular velocity vector. For example, for a pelvic bone node identified as "Hip", a record is created in the physical attribute parameter set. The key of the record is the string "Hip", and the value is a data combination containing three members: the first member stores a linear velocity vector, the second member stores a linear acceleration vector, and the third member stores a three-dimensional angular velocity vector. After completing the above storage operation for all predefined bone nodes in the current animation frame, the assembly of the physical attribute parameter set for that animation frame is complete. The assembly operation is repeated for each animation frame in the initial skeletal animation data, ultimately obtaining a sequence of physical attribute parameter sets corresponding one-to-one for each animation frame. The sequence of physical attribute parameter sets is the output of step S2.

[0027] S3: Parse the generated instruction text and determine the expected physical parameter threshold range that matches the generated instruction text based on a preset mapping relationship. Specifically, the implementation is as follows: In the steps of parsing the generated instruction text and determining the expected physical parameter threshold range matching the generated instruction text based on a preset mapping relationship, the operation of extracting keywords describing the action behavior from the generated instruction text is performed. The generated instruction text is a natural language string input by the user or provided by the upstream system. The keyword extraction operation performs word segmentation on the generated instruction text, which is the process of dividing a continuous text string into language vocabulary units. For Chinese generated instruction text, a dictionary-based maximum forward matching algorithm is used for word segmentation. The dictionary is a predefined set containing common action vocabulary. The maximum forward matching algorithm starts from the beginning character of the generated instruction text and attempts to match the longest possible word in the dictionary in turn. The successfully matched word is output as a word segmentation result, and then the matching continues from the remaining part of the text until the entire generated instruction text is processed. For English generated instruction text, word segmentation is performed based on spaces and punctuation marks. After word segmentation, a word segmentation sequence composed of multiple vocabulary units is obtained. Then, the word segmentation sequence is filtered. The filtering operation removes stop words from the word segmentation sequence. Stop words are common words that do not carry substantial action semantics. The stop word list includes, for example, prepositions and conjunctions. The remaining lexical units after filtering are initially identified as candidate keywords. Then, these candidate keywords are validated and screened based on a predefined action keyword library, which is a collection of verbs, adverbs, and adjectives describing movement states related to character actions. Each candidate keyword is compared with entries in the action keyword library, and those that exist in the library are retained. These retained lexical units are the final extracted keywords describing action behaviors.

[0028] The system performs an operation to match keywords against a pre-defined action type rule base to determine the target action type corresponding to the generated instruction text. The pre-defined action type rule base is a database storing various standard action type definitions and their corresponding triggering rules. Each standard action type represents a character movement pattern with clear semantics and typical physical manifestations in a virtual scene. The triggering rules for each standard action type consist of one or more logical judgment conditions. The matching process iterates through all standard action type entries in the pre-defined action type rule base. For the currently iterated standard action type, its triggering rules are checked. A common form of triggering rule requires that a set of required keywords must all appear in the extracted keyword set. Another form of triggering rule includes optional keywords and excluded keywords. The system calculates a matching score for each standard action type. The matching score is calculated based on the degree to which the extracted keywords satisfy their triggering rules. One way to calculate the matching score is to assign a weight value to each satisfied rule condition and sum them. After calculation, the standard action type with the highest matching score is selected as the target action type corresponding to the generated instruction text. If the highest matching score is lower than a preset matching confidence threshold, such as a 60% confidence threshold, it is determined that a specific target action type cannot be matched. In this case, a default general action type can be used or an error handling process can be triggered.

[0029] The process involves querying a predefined mapping table of action types and physical parameter threshold ranges to find the expected physical parameter threshold range associated with the target action type. This mapping table is a data structure that records the reasonable numerical ranges of various physical parameters corresponding to each standard action type. The query operation uses the defined target action type as the index key to search within the mapping table. Logically, the mapping table can be viewed as a two-dimensional table, where each row corresponds to a standard action type and each column corresponds to a specific physical parameter. Physical parameters include the magnitude of linear velocity, linear acceleration, and angular velocity of a specific skeletal node, calculated in step S2. The expected physical parameter threshold range is typically represented by a numerical interval containing a lower limit and an upper limit. The query result is a set containing the expected threshold ranges of all predefined physical parameters related to the target action type; this set is the expected physical parameter threshold range.

[0030] The pre-defined mapping relationship is established as follows: First, training animation data containing various motion types and their associated motion description texts are collected. The training animation data consists of a large number of existing, high-quality 3D skeletal animation sequences with accurate motion labels. Each training animation data sample is associated with one or more motion description texts that accurately describe its motion content. Sufficiently diverse samples need to be collected to cover the various standard motion types expected to be identified.

[0031] Physical attribute parameters are extracted from the training animation data, and their distribution is statistically analyzed. For each collected training animation data sample, the same processing method as in step S2 is used to calculate a set of physical attribute parameters associated with each animation frame. Then, for each sample, statistical analysis is performed on all values ​​of a specific physical attribute parameter calculated over the entire animation time series. A common statistical method is to calculate the statistical distribution characteristics of the physical attribute parameter in the sample, such as calculating the mean and standard deviation of the parameter across all frames. Another statistical method is to directly record the maximum and minimum values ​​of the parameter in the sample. By statistically analyzing multiple training samples belonging to the same standard motion type, the distribution of a certain physical parameter under that motion type can be obtained.

[0032] This step associates action description text with the corresponding physical attribute parameter distribution to construct a mapping table between action types and physical parameter threshold ranges. The core of this step is grouping training samples based on action description text. Specifically, action description text is mapped to corresponding standard action types through keyword extraction and action type matching. All training samples mapped to the same standard action type are grouped together. For each group of samples, for each predefined physical parameter to be monitored, a reasonable expected threshold range is determined based on the distribution characteristics of that parameter statistically analyzed from all samples within the group. One method for determining the threshold range is to use the average value of the statistical values ​​in the group plus or minus a certain number of standard deviations as the range boundary. For example, for the linear velocity of the root bone node in the running action type, the total average and total standard deviation of the average velocities of all samples in the group are calculated, and then the expected physical parameter threshold range is set from the total average minus twice the total standard deviation to the total average plus twice the total standard deviation. Another method for determining the threshold range is to directly take the minimum and maximum values ​​of the statistical values ​​of that parameter in the group of samples as the threshold range boundary. Finally, a threshold range is calculated and stored for each standard action type and each combination of physical parameters, thus filling the complete mapping table between action types and physical parameter threshold ranges. This mapping table is loaded into memory during system initialization for use by step S3 at runtime. The preset mapping relationship established in the above manner transforms the action intention described in natural language into quantifiable and specific physical motion parameter constraints.

[0033] S4: Compare the set of physical property parameters with the expected physical parameter threshold range to identify abnormal physical parameters in the set of physical property parameters. Specifically, this is implemented as follows: In the step of comparing the set of physical attribute parameters with the expected physical parameter threshold range and identifying abnormal physical parameters in the set of physical attribute parameters, the operation of comparing each parameter value in the set of physical attribute parameters with the corresponding parameter threshold range in the expected physical parameter threshold range is performed. The set of physical attribute parameters is a data structure generated in step S2. The set of physical attribute parameters contains a series of physical parameter values ​​for each animation frame. Each physical parameter value is associated with a specific physical quantity type and a bone node identifier. The expected physical parameter threshold range is a data structure generated in step S3. The expected physical parameter threshold range contains a series of threshold ranges for different combinations of physical quantity types and bone nodes. Each threshold range is defined by a lower limit and an upper limit. The comparison operation requires traversing all entries in the set of physical attribute parameters. For the currently traversed entry in the set of physical attribute parameters, its associated physical quantity type and bone node identifier are read from the entry. The read physical quantity type is, for example, the root bone linear velocity, and the read bone node identifier is, for example, Hip. Then, the read physical quantity type and bone node identifier are used as a composite lookup key to perform a query in the expected physical parameter threshold range data structure. The query operation searches for an entry within the expected physical parameter threshold range whose physical quantity type and skeletal node identifier perfectly match the composite lookup key. If a matching entry is found, its stored parameter threshold range is extracted. The extracted threshold range contains a defined lower limit and an upper limit. Then, the stored parameter values ​​are read from the currently traversed set of physical attribute parameters. The core of the comparison operation is performing a numerical relationship determination. This determination checks whether the read parameter value is less than or equal to the upper limit of the threshold range, and simultaneously checks whether it is greater than or equal to the lower limit. This determination is accomplished through logical AND operations within the program.

[0034] The operation determines whether the parameter value falls within the corresponding parameter threshold range. This operation is a direct result of the numerical relationship determination in the aforementioned comparison operation. Based on the logical result of the comparison operation, if the parameter value simultaneously satisfies the condition of being less than or equal to the upper limit value and the condition of being greater than or equal to the lower limit value, the logical determination result is true. A true logical determination result means that the parameter value falls within the corresponding parameter threshold range. If the parameter value does not satisfy the condition of being less than or equal to the upper limit value or the condition of being greater than or equal to the lower limit value (e.g., the parameter value is greater than the upper limit value or less than the lower limit value), the logical determination result is false. A false logical determination result means that the parameter value does not fall within the corresponding parameter threshold range. The judgment operation is represented in the program as a conditional branch statement. The conditional branch statement leads to different processing paths based on whether the logical determination result is true or false. Boundary case handling is an inherent part of the judgment logic. When the parameter value is exactly equal to the lower limit value of the parameter threshold range, according to the greater than or equal to operator used in the judgment logic, this situation is determined to fall within the parameter threshold range. When a parameter value is exactly equal to the upper limit of the parameter threshold range, it is determined to fall within the parameter threshold range according to the less than or equal to operator used in the judgment logic. Equal to the boundary indicates that the parameter value is on the acceptable boundary.

[0035] The process involves identifying parameter values ​​that do not fall within the corresponding parameter threshold range as abnormal physical parameters. This operation follows the result of the judgment operation. When the logical judgment result of the judgment operation is false, the parameter value is determined to be outside the corresponding parameter threshold range, triggering the abnormal determination process. The specific action of determining abnormal physical parameters is to create a new data record. The data record is used to identify and store an abnormal event. The currently processed physical attribute parameter value itself is stored as a field in the abnormal data record. The metadata associated with the parameter value is also stored in the abnormal data record. The metadata includes the animation frame number to which the parameter value belongs. The metadata includes the physical quantity type corresponding to the parameter value. The metadata includes the bone node identifier corresponding to the parameter value. The lower and upper limits of the parameter threshold range used in this comparison are also stored as reference information in the abnormal data record. This abnormal data record represents a determined instance of an abnormal physical parameter. After completing the judgment of a parameter value and recording a possible abnormality, the process continues to traverse the next parameter value in the physical attribute parameter set. The process repeatedly executes the comparison operation, judgment operation, and abnormal determination operation. After all parameter values ​​in the physical attribute parameter set have been traversed and processed, all generated abnormal data records are collected to form an abnormal physical parameter list. This abnormal physical parameter list is the output of step S4. The abnormal physical parameter list systematically identifies which specific physical motion parameters in the animation sequence generated by AI-generated content technology do not conform to the reasonable range expected based on textual intent. The abnormal physical parameter list provides precise input for subsequent steps to adjust the rendering effects. The entire identification process relies on clear numerical comparison rules and deterministic logical judgment. No fuzzy or heuristic reasoning is required. The entire process ensures the consistency and interpretability of the results.

[0036] S5: Based on the temporal variation characteristics of the physical property parameter set, the initial skeletal animation data is divided into multiple consecutive action stages, and a corresponding stage-level rendering effect strategy is matched for each action stage. The specific implementation is as follows: In the step of dividing the initial skeletal animation data into multiple consecutive action stages based on the temporal variation characteristics of the physical attribute parameter set and matching the corresponding stage-level rendering effect strategy for each action stage, the operation of extracting energy features representing motion intensity from the physical attribute parameter set and generating energy change curves is performed. The physical attribute parameter set is the set of physical parameter data associated with each animation frame, output from step S2. The operation of extracting energy features requires defining an energy calculation formula for quantifying the overall motion intensity. A typical energy calculation formula is to sum the magnitudes of the sum of linear velocities of all predefined bone nodes in the physical attribute parameter set for each animation frame. The magnitude of the sum of linear velocities refers to the magnitude of the linear velocity vector of the bone node, which is calculated by taking the square root of the sum of the squares of the X-axis, Y-axis, and Z-axis components of the linear velocity vector. For the current animation frame, all predefined bone nodes are traversed, the magnitude of the sum of linear velocities of each bone node is calculated sequentially, and then the magnitudes of the sum of linear velocities of all bone nodes are added together. The result is the motion energy value of the current animation frame. Another energy calculation formula can include the magnitude of the sum of linear accelerations in the calculation. Repeating the above calculation for all animation frames yields a sequence of motion energy values ​​that corresponds one-to-one with the temporal order of the animation frames. This sequence of motion energy values ​​is the energy characteristic sequence representing the motion intensity. Generating an energy change curve is the process of visualizing or numerically representing this motion energy value sequence with time as the horizontal axis and motion energy value as the vertical axis. Essentially, it is an ordered list of values, where each element represents the motion intensity at a specific point in time.

[0037] The operation determines the boundary frames of the action phase based on local extrema in the energy change curve and a preset energy threshold. A local extremum is a point on the energy change curve whose motion energy value is greater than or less than the motion energy values ​​of its preceding and following points. The algorithm for identifying local extrema iterates through every point in the energy change curve except the first and last points. For the currently traversed point, its motion energy value is obtained, along with the motion energy values ​​of its preceding and following adjacent points. If the current point's motion energy value is greater than both of these adjacent points, it is identified as a local maximum. If the current point's motion energy value is less than both, it is identified as a local minimum. The preset energy threshold is a threshold used to filter out invalid or minor fluctuations in motion energy values. The preset energy threshold can be set by analyzing the statistical characteristics of the entire energy change curve. For example, a preset energy threshold can be set to 30% of the average motion energy value across the entire energy change curve. Another method for setting a preset energy threshold is based on statistical analysis of energy distribution in historical animation data, using a fixed empirical value. The process of determining the boundary frames of an action phase involves filtering out points from all identified local extrema where the absolute value of the motion energy value exceeds the preset energy threshold. These filtered points mark moments when the motion intensity changes significantly. The animation frame numbers corresponding to these points are then designated as candidate boundary frames. Typically, the start or end of an action phase is marked by a significant increase or decrease in motion intensity; therefore, these candidate boundary frames divide the animation sequence into multiple segments with different intensity characteristics. To ensure the integrity of the phases, the first and last frames of the entire animation sequence are also forcibly added to the boundary frame list.

[0038] The initial skeletal animation data's time sequence is divided into multiple consecutive time segments based on boundary frames. Each time segment is defined as an action phase. A boundary frame is a list of animation frame numbers arranged chronologically. The segmentation operation divides the initial skeletal animation data's time sequence into multiple intervals according to the order of the boundary frames. Specifically, the first time segment is the interval from the first boundary frame to the second boundary frame (excluding the second boundary frame itself), which is left-closed and right-open. The second time segment is then the interval from the second boundary frame to the third boundary frame, and so on until the last boundary frame. Each time segment contains a set of consecutive animation frames. Each time segment is defined as an independent action phase. Each action phase has a unique phase identifier and records its start and end frame numbers. All animation frames in the initial skeletal animation data are assigned to one and only one action phase, thus completing the process of discretizing continuous animation data into several semantically meaningful action units.

[0039] The process involves matching a stage-level rendering effect strategy to each motion phase. The stage-level rendering effect strategy library is a predefined database storing various stage-level rendering effect strategies. Each strategy in the library contains three core elements: an effect type identifier, an effect intensity baseline range, and the set of target bone nodes to which the strategy applies. The effect type identifier is a string used to uniquely name a visual enhancement effect, such as a slight muscle tremor effect or a sweat effect. The effect intensity baseline range is a numerical range defining the recommended lower and upper limits of the effect's intensity under normal circumstances. The set of target bone nodes to which the strategy applies is a list indicating which bone nodes the effect primarily applies to; for example, a strategy for a tired breathing effect might include chest and abdominal bone nodes in its target bone node set. Matching the corresponding stage-level rendering effect strategy involves two sub-steps. The first sub-step is to calculate the statistical characteristics of the set of physical property parameters within the motion phase on the set of target bone nodes. For a given motion phase, based on the set of target bone nodes in the strategy, physical parameter data relevant only to these target bone nodes is filtered from the set of physical property parameters of all animation frames covered by that motion phase. Then, for a specific type of physical quantity, the statistical characteristics of that physical quantity are calculated across all frames of the action phase. Statistical characteristics can be the mean, maximum, minimum, or standard deviation. For example, the average linear velocity of the target chest and abdominal bone nodes during the fatigue breathing action phase is calculated. The second sub-step is to match the statistical characteristics with the intensity baseline range of each strategy in the strategy library, selecting the stage-level rendering effect strategy with the highest matching degree. One method for calculating the matching degree is to evaluate how well the statistical characteristic value falls within the intensity baseline range of the strategy. If the statistical characteristic value is completely within the intensity baseline range, the matching degree is the highest. If the statistical characteristic value deviates from the intensity baseline range, the minimum distance from its boundary to the range is calculated as the mismatch degree; the smaller the mismatch degree, the higher the matching degree. Another method for calculating the matching degree is to set an ideal interval for the center value of the intensity baseline range and calculate the relative error between the statistical characteristic value and the center value. All strategies in the stage-level rendering effect strategy library are traversed, and a matching degree score is calculated for the current action phase with each strategy. The strategy with the highest matching degree score is selected as the stage-level rendering effect strategy corresponding to the action phase. If the highest matching score is below an acceptable minimum matching threshold, such as 50%, the action phase may be assigned a default general strategy or marked as having no specific strategy. The minimum matching threshold can be verified experimentally or set empirically, for example, by testing a batch of known correctly matching animation phases and strategy pairs, statistically analyzing their matching score distribution, and setting the threshold at a low quantile of the distribution.Ultimately, each action phase is associated with a stage-level rendering strategy, which provides high-level guidelines for generating frame-by-frame rendering control parameters in subsequent steps. This process transforms continuous, physically-based animation data analysis into discrete action phase plans with explicit rendering semantics, assigning them appropriate visual enhancement themes. This lays a structured foundation for generating realistic visual effects that conform to motion semantics.

[0040] S6: For each animation frame, based on the stage-level rendering effect strategy of its respective action phase, and combined with the specific physical property parameter values ​​and abnormal physical parameters of the animation frame, generate visual enhancement effect control parameters adapted to the downstream rendering engine. The specific implementation is as follows: When generating visual enhancement control parameters adapted to the downstream rendering engine for each animation frame based on its stage-level rendering effect strategy for its respective action phase, combined with the animation frame's specific physical attribute parameters and abnormal physical parameters, the following steps are performed: First, the stage-level rendering effect strategy corresponding to the action phase to which the currently processed animation frame belongs is retrieved. The currently processed animation frame is identified by a unique animation frame number. The retrieval operation queries the action phase division results established in step S5. The action phase division results are a data structure that records the start and end frame numbers of each action phase, as well as a unique identifier for the stage-level rendering effect strategy matching that action phase. The query process compares the currently processed animation frame number with the start and end frame numbers of each action phase in the action phase division results. An action phase is searched such that the currently processed animation frame number is greater than or equal to the start frame number of that action phase, and the currently processed animation frame number is less than or equal to the end frame number of that action phase. When an action phase that meets the conditions is found, the unique identifier of its corresponding stage-level rendering effect strategy is read from the associated information recorded for that action phase. This unique identifier is then used as the key to retrieve information from the stage-level rendering effect policy library. The stage-level rendering effect policy library is a pre-loaded database that stores complete information on all defined stage-level rendering effect policies. The retrieval process retrieves the complete data record of the stage-level rendering effect policy that perfectly matches the unique identifier. This data record represents the stage-level rendering effect policy corresponding to the action stage of the current animation frame.

[0041] The process involves interpolating the effect intensity baseline range defined in the stage-level rendering effect strategy with the specific physical attribute parameter values ​​of the current animation frame to obtain a preliminary effect intensity value. The effect intensity baseline range defined in the stage-level rendering effect strategy is a numerical range containing a lower intensity limit and an upper intensity limit. The specific physical attribute parameter values ​​of the current animation frame are extracted from the physical attribute parameter set output in step S2, based on the current animation frame number and the target bone node set specified by the stage-level rendering effect strategy. Interpolation requires establishing a mapping relationship from physical attribute parameter values ​​to effect intensity values. One method for implementing this mapping relationship is linear interpolation. First, a physical quantity is determined to drive the interpolation. This physical quantity is usually the core motion feature that the stage-level rendering effect strategy focuses on. For example, for a muscle tremor effect strategy, the driving physical quantity might be the magnitude of the sum of linear accelerations of the target bone nodes. The value of this driving physical quantity is extracted from the specific physical attribute parameter values ​​of the current animation frame and denoted as the current physical quantity value. Next, a physical quantity reference range is defined. The physical quantity reference range can be determined based on the statistical characteristics of this physical quantity across all animation frames within the current action phase. For example, the minimum value of the physical quantity across all frames within the current action phase is taken as the lower reference limit, and the maximum value is taken as the upper reference limit. Then, linear interpolation is performed. The linear interpolation process involves subtracting the lower reference limit from the current physical quantity value to obtain a difference numerator. Subtracting the lower reference limit from the upper reference limit to obtain a difference denominator. Dividing the numerator by the denominator yields a normalization scaling factor between 0 and 1. If the current physical quantity value is less than the lower reference limit, the normalization scaling factor is set to 0. If the current physical quantity value is greater than the upper reference limit, the normalization scaling factor is set to 1. Finally, this normalization scaling factor is used to interpolate within the effect intensity baseline range. Specifically, the lower limit of the effect intensity baseline range is added to the normalization scaling factor multiplied by the difference between the upper and lower limits of the effect intensity baseline range. The result is the preliminary effect intensity value for the current animation frame. The preliminary effect intensity value is a specific numerical value.

[0042] The process involves determining whether the specific physical attribute parameter value of the current animation frame is marked as an abnormal physical parameter. This determination relies on the abnormal physical parameter list output in step S4. The abnormal physical parameter list is a data structure where each record contains the specific value of an abnormal physical parameter and its associated metadata, including the animation frame number, physical quantity type, and bone node identifier. The determination requires checking the specific physical attribute parameter value for the currently calculated effect intensity. First, the current physical quantity type and the current target bone node identifier are identified from the current calculation context. Then, using the currently processed animation frame number, current physical quantity type, and current target bone node identifier as a composite query key, a search is performed in the abnormal physical parameter list. The search process involves traversing each record in the abnormal physical parameter list and checking if there exists a record whose animation frame number is the same as the current animation frame number, whose physical quantity type is the same as the current physical quantity type, and whose bone node identifier is the same as the current target bone node identifier. If such a completely matching record exists, the specific physical attribute parameter value of the current animation frame is determined to be marked as an abnormal physical parameter. If no matching record is found after traversing the entire list of abnormal physical parameters, it is determined that the specific physical attribute parameter value of the current animation frame has not been marked as an abnormal physical parameter. The output of the judgment operation is a boolean value.

[0043] The process involves adjusting the initial effect intensity value according to a preset abnormal parameter correction rule if the physical parameter is marked as an abnormal physical parameter, thus obtaining the final effect intensity value. The preset abnormal parameter correction rule is defined as follows: when a specific physical attribute parameter value of an animation frame is marked as an abnormal physical parameter, a corresponding correction coefficient is selected from a preset correction coefficient table based on the parameter type of the abnormal physical parameter; the initial effect intensity value is multiplied by the correction coefficient to obtain the adjusted effect intensity value as the final effect intensity value. Implementing this rule requires a preset correction coefficient table. The preset correction coefficient table is a data structure that maps different physical quantity types to a correction coefficient value. The correction coefficient value is typically a positive number less than or equal to 1. The correction coefficient value can be set empirically. For example, by analyzing the correlation between abnormal parameters and reasonable visual effect intensity in historical animation data, the correction coefficient value can be set to 0.5. The correction coefficient value can also be derived from experimental data analysis, such as adjusting different coefficients on a test set and observing the rendering effect to select the coefficient value that makes the effect most natural. The adjustment operation is triggered when the operation output is true. First, based on the current physical quantity type, the preset correction coefficient table is consulted to obtain the correction coefficient value associated with that physical quantity type. Then, the calculated preliminary effect intensity value is multiplied by the correction coefficient value. The result of the multiplication is the adjusted effect intensity value. The adjusted effect intensity value will be used as the final effect intensity value for the current animation frame. If the operation output is false, the preliminary effect intensity value is directly used as the final effect intensity value.

[0044] The process involves generating visual enhancement control parameters, including an effect type identifier and an intensity parameter value, based on the final effect intensity value. The generation process begins by retrieving the effect type identifier from the stage-level rendering effect strategy. This identifier is a string. Next, the final effect intensity value is converted into an intensity parameter value that conforms to the requirements of the downstream rendering engine interface. The intensity parameter value may require scalar conversion; for example, dividing the final effect intensity value by a preset reference intensity value converts it to a scalar between 0 and 1. Finally, the effect type identifier and intensity parameter value are encapsulated into a complete data unit. This data unit can be in key-value pair format. This data unit is the visual enhancement control parameter. After generating the visual enhancement control parameter for the current animation frame, this parameter can be immediately sent to the corresponding interface of the downstream rendering engine. By repeating all the above operations for each animation frame in the initial skeletal animation data, a series of visual enhancement control parameters synchronized with the animation frames in time can be generated for the entire animation sequence.

[0045] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0046] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0047] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. Computer-readable storage media can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0048] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0049] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0050] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0051] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0052] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0053] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0054] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An AIGC-driven animation generation method for virtual image scenes, characterized in that, include: S1: Obtain the initial skeletal animation data and corresponding generation instruction text generated by AIGC; S2: Analyze the motion trajectory of predefined bone nodes in the initial skeletal animation data and calculate the set of physical attribute parameters associated with each animation frame; S3: Parse the generated instruction text and determine the expected physical parameter threshold range that matches the generated instruction text based on the preset mapping relationship; S4: Compare the set of physical attribute parameters with the expected physical parameter threshold range to identify abnormal physical parameters in the set of physical attribute parameters; S5: Based on the temporal variation characteristics of the physical property parameter set, the initial skeletal animation data is divided into multiple continuous action stages, and a corresponding stage-level rendering effect strategy is matched for each action stage. S6: For each animation frame, based on the stage-level rendering effect strategy of its respective action stage, and combined with the specific physical attribute parameter values ​​and abnormal physical parameters of the animation frame, generate visual enhancement effect control parameters adapted to the downstream rendering engine.

2. The AIGC-driven animation generation method for virtual image scenes according to claim 1, characterized in that, S1 includes: Receive initial skeletal animation data from the animation generation interface; Receive the generated instruction text from the instruction input interface; The initial skeletal animation data and the generated instruction text are associated and bound together based on the same task identifier that accompanies the initial skeletal animation data and the generated instruction text.

3. The AIGC-driven animation generation method for virtual image scenes according to claim 1, characterized in that, S2 include: Extract the position and rotation data of predefined bone nodes in consecutive animation frames from the initial skeletal animation data; Based on the changes in position data between consecutive animation frames, calculate the linear velocity and linear acceleration of the skeletal nodes; The angular velocity of the bone nodes is calculated based on the changes in rotation data between consecutive animation frames; For each animation frame, the calculated linear velocity, linear acceleration, and angular velocity are assembled into a set of physical property parameters for that animation frame.

4. The AIGC-driven animation generation method for virtual image scenes according to claim 1, characterized in that, S3 includes: Extract keywords describing actions from the generated instruction text; The keywords are matched with a preset action type rule base to determine the target action type corresponding to the generated instruction text; Based on a preset mapping table between action types and physical parameter threshold ranges, query the expected physical parameter threshold range associated with the target action type.

5. The AIGC-driven animation generation method for virtual image scenes according to claim 4, characterized in that, The preset mapping relationship is established as follows: collect training animation data containing multiple action types and their associated action description text; extract physical attribute parameters from the training animation data and statistically analyze their distribution; associate the action description text with the corresponding physical attribute parameter distribution to construct a mapping table between action type and physical parameter threshold range.

6. The AIGC-driven animation generation method for virtual image scenes according to claim 1, characterized in that, S4 include: Compare each parameter value in the set of physical property parameters with the corresponding parameter threshold range in the expected physical parameter threshold range; Determine whether the parameter value falls within the corresponding parameter threshold range; Parameter values ​​that do not fall within the corresponding parameter threshold range are identified as abnormal physical parameters.

7. The AIGC-driven animation generation method for virtual image scenes according to claim 1, characterized in that, S5 include: Extract energy features characterizing motion intensity from a set of physical property parameters and generate energy change curves; The boundary frames of the action phase are determined based on the local extreme points in the energy change curve and the preset energy threshold. The initial skeletal animation data time series is divided into multiple consecutive time periods based on the boundary frames, and each time period is defined as an action phase; Based on the statistical characteristics of the physical attribute parameter set within each action stage, the corresponding stage-level rendering effect strategy is matched from the preset stage-level rendering effect strategy library.

8. The AIGC-driven animation generation method for virtual image scenes according to claim 7, characterized in that, Each strategy in the stage-level rendering effect strategy library includes: effect type identifier, effect intensity baseline range, and the target bone node set to which the corresponding strategy applies; matching the corresponding stage-level rendering effect strategy includes: calculating the statistical characteristics of the set of physical attribute parameters in the action stage on the target bone node set; matching the statistical characteristics with the intensity baseline range of each strategy in the strategy library, and selecting the stage-level rendering effect strategy with the highest matching degree.

9. The AIGC-driven animation generation method for virtual image scenes according to claim 1, characterized in that, S6 include: For the currently processed animation frame, obtain the stage-level rendering effect strategy corresponding to the action stage to which the animation frame belongs; Based on the effect intensity baseline range defined in the stage-level rendering effect strategy and the specific physical property parameter values ​​of the current animation frame, the preliminary effect intensity value of the current animation frame is obtained through interpolation calculation. Determine whether the specific physical property parameter values ​​of the current animation frame are marked as abnormal physical parameters; If it is marked as an abnormal physical parameter, the initial effect intensity value is adjusted according to the preset abnormal parameter correction rules to obtain the final effect intensity value; Generate visual enhancement effect control parameters that include effect type identifier and intensity parameter value based on the final effect intensity value.

10. The AIGC-driven animation generation method for virtual image scenes according to claim 9, characterized in that, The abnormal parameter correction rule is as follows: when the specific physical attribute parameter value of an animation frame is marked as an abnormal physical parameter, the corresponding correction coefficient will be selected from the preset correction coefficient table according to the parameter type to which the abnormal physical parameter belongs. Multiply the initial effect intensity value by the correction factor to obtain the adjusted effect intensity value, which is then used as the final effect intensity value.

Images