Artistic design capability cultivation information management system and method based on digital twinning

Abstract

The invention relates to the technical field of computers and digital twinning, and particularly discloses an art design capability cultivation information management system and method based on digital twinning, and the system comprises a physical perception and interaction layer, a digital twinning construction and simulation layer, a capability modeling and analysis layer, and a teaching decision and intervention layer. Through real-time construction of a digital twinborn body synchronized with physical creation and based on whole process data, multi-dimensional capability quantitative analysis and personalized teaching decision making are carried out, and precise and intelligent management of an art design capability cultivation process is realized. According to the method, an intelligent closed loop from data insight to teaching actions is constructed through a teaching decision and intervention layer, and the system not only can provide an analysis report, but also can automatically match teaching strategies and recommend personalized resources, and carries out active intervention directly through environment adjustment and task pushing.

Description

Technical Field

[0001] This invention belongs to the field of computer and digital twin technology, specifically relating to an information management system and method for cultivating art design capabilities based on digital twins. Background Technology

[0002] Against the backdrop of the deep integration of information technology and education, digital competency development and assessment systems have become an important direction for the development of modern educational technology. This field aims to utilize advanced information technology to digitally record, analyze, and guide learners' knowledge acquisition, skill formation, and competency development processes, thereby improving teaching efficiency and personalization.

[0003] The cultivation of art and design skills, as a highly practical branch of education, relies heavily on real-time guidance during the creative process, multi-dimensional presentation of artworks, and dynamic assessment of skill development. Traditional training models typically focus on physical creation, with teachers providing experiential evaluations by observing the final work and combining it with limited process records.

[0004] Art and design education often combines offline physical creation with online submissions, which has significant limitations. Data from the creative process is difficult to collect and record systematically, leading to a severe lag in feedback. Digital archiving and preliminary evaluation typically only occur after the physical work is completed, making it impossible to track and intervene in key creative stages such as conception, sketching, and revisions in real time. Furthermore, the assessment of learners' abilities lacks objective, quantitative analysis based on data from the entire process, relying heavily on subjective experience and making it difficult to accurately identify individual weaknesses and development trends. In addition, teaching resources and the creative environment are disconnected, failing to construct a virtual training environment that can real-time map the physical creative state and support simulation, prediction, and interaction. This results in a lack of effective data support and simulation verification methods for planning and dynamically adjusting personalized training paths. Summary of the Invention

[0005] The purpose of this invention is to provide an information management system and method for cultivating art and design skills based on digital twins, so as to solve the technical contradictions mentioned in the background art, such as the difficulty in systematically collecting data on the art and design creation process, the lag in teaching feedback, the lack of objective quantitative basis for ability assessment, and the separation of teaching environment and resources.

[0006] To achieve the above objectives, this invention proposes an information management system for cultivating art and design skills based on digital twins. This system constructs a complete mapping and interactive closed loop from the physical creation space to the virtual twin space. Its core lies in creating a digital twin that is synchronized in real-time and interacts bidirectionally with the learner's physical creation process, and then performing multi-dimensional capability modeling and intelligent intervention based on the full-process data generated by this twin.

[0007] The system includes a physical perception and interaction layer, a digital twin construction and simulation layer, a capability modeling and analysis layer, and a teaching decision-making and intervention layer.

[0008] The physical perception and interaction layer is used to comprehensively collect multimodal data during the physical creation process and feed virtual instructions back to the physical creation environment. This layer is equipped with high-precision image acquisition devices, 3D scanning devices, pressure and posture sensors, and environmental parameter sensors. The high-precision image acquisition device captures a 2D image of the creation surface at a rate of no less than 30 frames per second to record sketch lines, brushstroke sequences of color smears, and timestamps. The 3D scanning device collects point cloud data from the 3D model or installation artwork at a preset scanning cycle, constructing a temporal snapshot of its 3D geometry. The pressure and posture sensors are integrated into the graphics tablet, stylus, and wearable device to collect real-time data on brush pressure, pen speed, acceleration, and the creator's hand and body posture. The environmental parameter sensors continuously monitor the light intensity, color temperature, and ambient sound decibels of the creation space. This layer also includes a physical feedback execution unit, which receives instructions from the teaching decision and intervention layer and dynamically adjusts the lighting conditions, reference image projection, and tabletop angle of the physical creation environment through a controllable light source array, projector, and adjustable work surface.

[0009] The digital twin construction and simulation layer receives and fuses multimodal data streams from the physical perception and interaction layers, constructs and drives a digital twin that is strictly synchronized with the physical creation process, and provides a virtual simulation environment. This layer includes a twin data fusion engine, a real-time rendering engine, and a parametric simulation engine. The twin data fusion engine, based on a unified time base, spatiotemporally aligns and correlates 2D image sequences, 3D point cloud sequences, brushstroke sensor data, and environmental parameters to generate a structured twin temporal state dataset. The real-time rendering engine, based on this temporal state dataset, renders the brushstroke layer evolution animation of a 2D artwork or the geometric shape growth animation of a 3D artwork in real-time in virtual space, with a refresh latency of no more than 100 milliseconds. The parametric simulation engine, based on the current state of the twin, allows teaching systems or learners to perform non-destructive parameter modifications and effect previews of the artwork in a virtual environment. Parameter modifications include, but are not limited to, overall replacement of the color system, displacement and scaling of compositional elements, structural topology optimization of the 3D model, and replacement of material textures.

[0010] The capability modeling and analysis layer is used for in-depth analysis of the entire process data recorded by the digital twin, constructing a multi-dimensional capability quantification model of the learner, and conducting developmental assessments and bottleneck diagnoses. This layer includes a feature extraction module, a multi-dimensional capability quantification module, and a development trajectory prediction module. The feature extraction module extracts predefined creative process feature vectors from the temporal state data of the digital twin. These features include the sketch modification frequency, line smoothness index, and color experiment diversity entropy value in the conception stage; the brushstroke density distribution, local refinement index, and global composition balance index in the refinement stage; and the modification backtracking depth and optimization iteration number in the adjustment stage.

[0011] The multidimensional ability quantification module inputs the extracted feature vectors into a set of pre-trained ability assessment models. These models employ a deep neural network architecture and output standardized scores for learners across five core dimensions: creative conception, modeling expression, color application, spatial composition, and technical implementation. Each dimension's score ranges from 0 to 100. The development trajectory prediction module, based on learners' historical multidimensional ability score sequences, uses a time-series prediction algorithm to predict their development trend curves across each ability dimension over the next three teaching cycles, identifying bottleneck dimensions where growth has stagnated or may decline.

[0012] The instructional decision-making and intervention layer generates personalized instructional decisions and triggers corresponding intervention actions based on the output of the competency modeling and analysis layer. This layer includes a strategy matching engine, a resource recommendation engine, and an intervention instruction generator. The strategy matching engine matches learners' current competency gaps and predicted developmental bottlenecks with the intervention strategy map in the instructional knowledge base. This strategy map defines a series of instructional strategies, practice tasks, and feedback points corresponding to different competency issues. The resource recommendation engine dynamically selects and assembles personalized learning resource packages from the digital resource library based on the matched instructional strategies. These resource packages may include targeted instructional video clips, classic case analyses, virtual simulation practice tasks, and reference material libraries. The intervention instruction generator is responsible for transforming decisions into executable instructions. These instructions are divided into three categories: the first category is feedback instructions, which present learners with their competency radar chart, process replay comparative analysis, and specific improvement suggestions through the user interface; the second category is environmental adjustment instructions, which are issued to the physical feedback execution unit of the physical perception and interaction layer to change the creative environment; and the third category is task push instructions, which directly inject new practice tasks into the learner's creative process.

[0013] Furthermore, the parametric simulation engine in the digital twin construction and simulation layer operates as follows: upon receiving a simulation request, the engine first creates a copy instance of the digital twin in memory. Operations on this copy instance are completely independent of the real-time synchronized master twin. The engine provides a set of parametric controls, allowing users to adjust specific parameters in the copy instance, such as shifting the overall hue of a painting by 60 degrees or increasing the local mesh subdivision level of a 3D model by 2 levels. After adjustment, the real-time rendering engine immediately performs visual rendering of the modified copy instance, allowing users to intuitively compare the differences in effects before and after the modification. All simulation operations and results are recorded, but they are not written back or affect the data stream of the master twin corresponding to the ongoing physical creation process.

[0014] Furthermore, the feature extraction module in the capability modeling and analysis layer calculates the line smoothness index as follows: First, the module extracts continuous coordinate points from the stroke sensing data sequence and calculates the rate of change of direction angle between adjacent points. Then, a sliding time window is used to smooth the rate of change of direction angle sequence to filter out high-frequency jitter noise. The line smoothness index is defined as the reciprocal of the standard deviation of the smoothed sequence; a lower standard deviation indicates a smoother and more gradual change in line direction, and a higher corresponding smoothness index value. This index is calculated segmented by time window and ultimately incorporated into the creation process feature vector.

[0015] Furthermore, the strategy matching engine in the teaching decision-making and intervention layer uses a weighted rule network for its matching logic. This network uses the difference between the learner's ability dimension score and the target score as input weights, with larger differences resulting in higher weights for those dimensions. Each intervention strategy in the strategy graph is labeled with its primary and secondary ability dimension tags. The matching process calculates the matching degree between each strategy's tag and the learner's high-weight ability dimensions, prioritizing intervention strategies with a matching degree exceeding a preset threshold of 85%. If multiple strategies have similar matching degrees, they are further ranked and optimized by combining the learner's historical strategy execution performance data.

[0016] Furthermore, the system operates within a layered collaborative framework, which includes a data synchronization layer, a business logic layer, and an interactive presentation layer. The data synchronization layer is responsible for high-frequency, low-latency data communication between the physical perception and interaction layer and the digital twin construction and simulation layer, ensuring synchronization between the twins. The business logic layer encapsulates the core algorithms and rules of the capability modeling and analysis layer and the teaching decision-making and intervention layer, performing asynchronous computation and decision generation. The interactive presentation layer is responsible for uniformly presenting the digital twin's visual interface, analysis result reports, and teaching intervention content to the terminals of teachers and learners.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention, through the collaboration of the physical perception and interaction layer and the digital twin construction and simulation layer, has for the first time achieved lossless digital mirroring and real-time synchronization of the entire creative process of art and design entities. Every stroke, every modification, and every change in three-dimensional form during the creative process is transformed into structured time-series data and recorded by the twin, completely solving the fundamental problems of data loss and feedback lag in the creative process under the traditional model, and providing an unprecedented full data foundation for subsequent accurate analysis.

[0018] 2. This invention transforms subjective, empirical assessments into objective, quantitative analyses based on feature vectors throughout the entire process through a capability modeling and analysis layer. It utilizes a deep neural network model to extract features strongly correlated with core capability dimensions from massive amounts of process data and outputs standardized capability scores and development trajectory predictions. This allows the diagnosis of learners' capability status to evolve from fuzzy judgments to precise measurements and can proactively identify developmental bottlenecks, providing reliable data insights for personalized teaching.

[0019] 3. This invention constructs an intelligent closed loop from data insight to teaching action through the teaching decision-making and intervention layer. The system not only provides analysis reports but also automatically matches teaching strategies, recommends personalized resources, and proactively intervenes directly through environmental adjustments and task pushes. This data-driven real-time interaction and intervention mechanism transforms the digital twin from a passive recording and display tool into an intelligent coach that can actively participate in the teaching process and dynamically optimize the training path, significantly improving the accuracy and efficiency of art and design ability cultivation. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention; Figure 2 This is a schematic diagram of the core principle framework for the construction and real-time synchronization of digital twins in this invention; Figure 3 This is a schematic diagram of the multi-level interaction relationship and data flow between the physical perception and interaction layer and the digital twin construction and simulation layer in this invention; Figure 4 This is a logical flow diagram of the capability modeling and analysis layer in this invention; Figure 5 This is a logical flowchart of the teaching decision-making and intervention layer in this invention. Detailed Implementation

[0021] The overall technical architecture of the digital twin-based art design ability training information management system proposed in this invention is shown in the attached figure. Figures 1 to 5As shown, the system comprises four functional layers: the physical perception and interaction layer, the digital twin construction and simulation layer, the capability modeling and analysis layer, and the teaching decision-making and intervention layer. These layers are tightly coupled through high-bandwidth, low-latency data channels, forming a complete closed loop from physical creation to virtual twin modeling, and then to capability quantification assessment and intelligent teaching intervention. The entire system operates within a layered, collaborative software framework, which includes a data synchronization layer, a business logic layer, and an interactive presentation layer. These layers are responsible for the real-time transmission of the underlying data stream, the asynchronous computation of the mid-level algorithm model, and the unified output of the upper-level human-computer interaction content, respectively.

[0022] During the system startup phase, learners enter a physical creation space equipped with a full set of sensing devices. This space serves as the deployment platform for the physical perception and interaction layer. Its core task is to achieve lossless, holographic acquisition of all quantifiable behaviors and environmental variables during the artistic creation process, and to have the ability to feed virtual commands back to the physical world for execution. Specifically, this layer is equipped with high-precision image acquisition devices, 3D scanning devices, pressure and posture sensing devices, and environmental parameter sensing devices. The high-precision image acquisition device uses an industrial-grade CMOS sensor to continuously capture two-dimensional images of the creation surface at a fixed frame rate of 30 frames per second. Each frame is embedded with a timestamp accurate to the millisecond level and transmitted to the digital twin construction and simulation layer via gigabit Ethernet after being compressed using H.265 encoding. This image sequence is used to reconstruct the line generation order, color layering, and brushstroke coverage relationships during the sketching process, serving as the fundamental data source for constructing the evolution trajectory of the two-dimensional artwork.

[0023] The 3D scanning device employs structured light or laser triangulation principles to collect point cloud data from the learner's 3D model or installation art piece at a preset 5-second scanning cycle. Each scan generates point cloud data containing no fewer than 1 million spatial coordinate points, each with accompanying normal vector and reflection intensity information. After denoising, registration, and meshing, the point cloud data forms a 3D geometric snapshot of the artwork at a specific moment. These snapshots are arranged chronologically to constitute a temporal dataset of the artwork's morphological evolution, used to drive the dynamic growth animation of the 3D model in virtual space.

[0024] Pressure and attitude sensors are integrated into the surface of the graphics tablet, the stylus, and the inertial measurement unit (IMU) wristband worn by the learner. The graphics tablet records the X and Y coordinates, Z-axis pressure value, and tilt angle of the pen tip on the plane at a sampling frequency of 200 times per second; the stylus's built-in six-axis gyroscope and accelerometer output the spatial attitude quaternion of the pen body at a rate of 100 Hz; the IMU wristband simultaneously collects the joint angles and motion accelerations of the wrist, forearm, and even shoulder. The above multi-source sensor data is converged to the local edge computing node via Bluetooth 5.0 protocol, where timestamp alignment and coordinate system transformation are performed to ultimately form a high-dimensional temporal signal describing the creator's microscopic operational behavior.

[0025] The environmental parameter sensing devices include a illuminometer, a color temperature sensor, and a sound level meter, which monitor the light intensity (in lux), light source color temperature (in Kelvin), and ambient noise level (in decibels) in the creative space once per second, respectively. These environmental variables have been shown to have a significant impact on the emotional state and visual judgment of artistic creation, and therefore have been incorporated into the twin's state description system as an important dimension of contextual perception.

[0026] The physical perception and interaction layer also includes a physical feedback execution unit, which consists of a controllable LED light source array, a high-lumen short-throw projector, and an electrically adjustable tilt worktable. Upon receiving environmental adjustment instructions from the teaching decision-making and intervention layer, this unit can dynamically change the lighting conditions of the work area (e.g., switching the color temperature from 3000K to 5600K to simulate sunlight), project reference images or compositional guidelines onto the worktable, or adjust the worktable's tilt angle to accommodate different creative postures. All actions are accompanied by a status feedback mechanism to ensure the reliability and traceability of instruction execution.

[0027] Please refer to the attached document. Figure 3 The multimodal raw data streams collected by the physical perception and interaction layer are pushed to the digital twin construction and simulation layer in real time. The core function of this layer is to construct a digital twin that is strictly synchronized with the physical creation process and has the same state, and to provide non-destructive virtual simulation capabilities. This layer consists of three subsystems: a twin data fusion engine, a real-time rendering engine, and a parametric simulation engine.

[0028] The twin data fusion engine first establishes a globally unified time reference, anchored at the frame start time of the high-precision image acquisition device. Through hardware synchronization signals or software interpolation algorithms, data streams from different sensors are mapped to the same time axis. For example, the twin state at a given time t is defined by the following data: the 2D image frame at time t, the most recent 3D scan point cloud (time ≤ t and closest to t), the average pen pressure and pose data within a window of t ± 50 milliseconds, and the environmental parameter readings at time t. The engine structurally encapsulates this heterogeneous data, generating a twin time-series state data object containing multiple fields. This object is serialized in JSON format, with each field corresponding to a data modality and accompanied by metadata describing its source, accuracy, and confidence level.

[0029] The real-time rendering engine receives the structured state data objects mentioned above and reconstructs the creation scene in a virtual 3D space based on their content. For 2D paintings, the engine parses the image frame sequence into multiple transparent layers, recognizing each brushstroke as an independent vector path or bitmap region and assigning it temporal attributes. Through interpolation algorithms, the engine can generate smooth brushstroke evolution animations at any point in time, with an animation refresh rate of no less than 10 frames per second and end-to-end rendering latency controlled within 100 milliseconds. For 3D sculptures or installations, the engine converts temporal point cloud data into a dynamically changing mesh model, uses GPU-accelerated surface reconstruction algorithms (such as Poisson reconstruction) to generate smooth surfaces in real time, and supports observing the morphological growth process of the work from any perspective.

[0030] The parametric simulation engine provides an isolated virtual experimental field. When a learner or teacher initiates an "effect preview" request, the engine copies the complete state of the current master twin in memory, creating a copy instance. This copy is completely decoupled from the master twin; any modifications only affect the copy. The engine exposes a set of parametric control interfaces, allowing users to adjust key properties of their work. For example, in the color dimension, hue offset ΔH (range -180 to +180 degrees), saturation scaling factor S (range 0.1 to 2.0), and lightness offset L (range -50 to +50) can be specified; in the composition dimension, translation vectors (dx, dy), scaling ratios k (range 0.5 to 2.0), and rotation angles can be applied to selected elements. (Range -180 to +180 degrees); In the 3D model dimension, the local mesh subdivision level can be adjusted (incremental ±1 to ±3 levels) or a topology optimization algorithm can be applied to reduce the number of faces. After each parameter adjustment, the real-time rendering engine immediately performs a visual rendering of the modified copy and displays the original and modified states side by side on the interface for intuitive comparison by the user. All simulation operation logs (including parameter values, operation time, and user ID) are persistently stored, but never written back to the main twin data stream, ensuring the integrity and non-interference of the entity creation process.

[0031] Please refer to the attached document. Figure 2 The construction and synchronization mechanism of the digital twin relies on a double-buffered state update strategy. The primary buffer stores the current state, which is strictly synchronized with the physical world, while the secondary buffer receives and temporarily stores newly arriving sensor data. When the data accumulated in the secondary buffer reaches a complete time slice (e.g., a 100-millisecond window), the system triggers an atomic state switch, promoting the secondary buffer to the primary buffer and simultaneously clearing the original primary buffer to become the new secondary buffer. This mechanism effectively avoids data races and rendering tearing, ensuring the consistency and timeliness of the digital twin's state.

[0032] The structured twin time-series state dataset output from the digital twin construction and simulation layer is asynchronously pushed to the capability modeling and analysis layer. The core task of this layer is to extract features highly correlated with artistic design capabilities from massive amounts of process data and construct a quantifiable, multi-dimensional capability model. (See attached image) Figure 4 As shown, this layer includes a feature extraction module, a multi-dimensional capability quantification module, and a development trajectory prediction module.

[0033] The feature extraction module first analyzes the twin's temporal state data in stages. The entire creative process is divided into three logical stages: conception, refinement, and adjustment, based on abrupt changes in brushstroke activity density or prolonged pauses. In the conception stage, the module focuses on analyzing the generation and modification behavior of sketch lines. The system identifies all continuously drawn line segments and calculates the line smoothness index for each segment. The calculation process for this index is as follows: extracting a continuous coordinate point sequence from the brushstroke sensor data. ,in Calculate the included angle between the vectors formed by three adjacent points, and then obtain the direction angle sequence. .right The sequence was smoothed using a sliding Hanning window with a length of 500 milliseconds to obtain a smoothed sequence. The line smoothness index F is defined as:

[0034] in for Standard deviation of the sequence A very small positive number (0.001) is used to prevent division by zero errors. The higher the F value, the smoother the change in line direction and the smoother the stroke. This indicator is calculated independently for each line segment, and a weighted average (weighted by the line segment length) is taken at the end of the design phase as the smoothness characteristic of that phase.

[0035] In addition, the conceptualization stage also calculates features such as sketch modification frequency (the number of lines deleted or redrawn per unit time) and color experiment diversity entropy (Shannon entropy calculated based on the hue distribution used for the first time). The refinement stage focuses on the depiction of details in the work, extracting features such as brushstroke density distribution (the number of brushstrokes per unit area), local fineness index (the richness of detail measured by the standard deviation of image gradient amplitude), and global compositional balance index (calculated based on the offset distance between the work's center of gravity and the center of the image, and the variance of pixel distribution in the four quadrants). The adjustment stage focuses on optimization behavior, extracting features such as modification backtracking depth (the average number of steps to revert from the current state to a historical state) and optimization iteration count (the number of rounds of continuous fine-tuning operations). Finally, all features are combined into a 128-dimensional feature vector of the creative process.

[0036] The multidimensional ability quantification module receives the feature vector and inputs it into a set of pre-trained deep neural network models. This set of models contains five independent sub-networks, corresponding to the five core dimensions of creative conception ability, form expression ability, color application ability, spatial composition ability, and technical implementation ability. Each sub-network adopts a three-layer fully connected structure, with 256, 128, and 64 neurons in the hidden layers, respectively. The activation function is ReLU, and the output layer uses the Sigmoid function to map the result to a standardized score range of 0 to 100. The model training data comes from a labeled dataset of thousands of art students, with labels obtained by averaging scores independently given by three senior teachers. The inference process is completed under GPU acceleration, with a single evaluation taking no more than 200 milliseconds. The output is a quintuple score vector. ,in to These correspond to the quantitative scores of the five capability dimensions mentioned above.

[0037] The development trajectory prediction module performs time-series modeling based on learners' historical ability score sequences. The system maintains an ability development profile for each learner, recording their five-dimensional scores after each creative task. This module employs a Long Short-Term Memory (LSTM) network architecture, taking the score sequence of the past six teaching cycles as input and outputting predicted scores for the next three cycles. The LSTM network contains two layers of hidden units, each with 128 neurons, and is trained end-to-end on historical data using a backpropagation algorithm. The prediction results include not only point estimates but also 95% confidence intervals to assess the uncertainty of the prediction. The module also automatically detects the growth slope of scores in each dimension. If the growth rate of a dimension is less than 1% for two consecutive cycles, it is marked as "stagnant growth"; if the predicted score decline exceeds 5% in future cycles, it is marked as "potential decline risk," and that dimension is identified as a development bottleneck.

[0038] The output of the competency modeling and analysis layer is then transmitted to the instructional decision-making and intervention layer. (See attached...) Figure 5 As shown, this layer consists of a strategy matching engine, a resource recommendation engine, and an intervention instruction generator, and is responsible for transforming data analysis insights into specific teaching actions.

[0039] The strategy matching engine maintains a teaching knowledge base that stores an intervention strategy map. This map contains hundreds of intervention strategies validated by education experts. Each strategy is labeled with a primary ability dimension (e.g., "color application ability") and a secondary dimension (e.g., "creative conception ability"), and is associated with specific teaching methods (e.g., "limited color practice"), practice tasks (e.g., "monochrome gradation deduction"), and feedback points (e.g., "paying attention to warm and cool contrast"). The matching process uses a weighted rule network: first, it calculates the difference between the learner's current ability dimensions and the target baseline (usually 85 points). ,like If the value is greater than 0, then that dimension receives a weight. Otherwise, the weight is 0. Then, for each policy in the policy graph, the matching degree between its label and the high-weight dimensions is calculated. :

[0040] The system prioritizes strategies with an M-value greater than 0.85. If multiple highly matching strategies exist, the system further queries the learner's historical strategy execution records and selects the strategy with the highest past success rate for ranking and optimization.

[0041] The resource recommendation engine retrieves relevant content from a digital resource library based on the matched intervention strategy. This library includes instructional videos (graded by knowledge point and difficulty), classic case studies (with multi-dimensional analytical tags), virtual simulation task templates (configurable by parameters), and a material library (containing high-definition textures, color swatches, and composition reference images). The engine employs a content-based filtering algorithm, combining the resource types required by the strategy with the learner's preference history (such as the duration of frequently watched videos and the style of materials commonly used), to dynamically assemble personalized learning resource packages. These resource packages are pushed to the interactive presentation layer via an API interface for learners to access at any time.

[0042] The intervention instruction generator transforms the decision results into three types of executable instructions. The first type is feedback instructions, which include a radar chart of the learner's current ability (visualization of five-dimensional scores), a video comparing the current creation with their best historical work, and improvement suggestions written by the natural language generation module (e.g., "Your color attempts are rather monotonous; we suggest trying complementary color combinations to enhance visual impact"). The second type is environmental adjustment instructions, such as "Adjust the work surface color temperature to 5600K for 15 minutes." This instruction is encapsulated as a JSON message and sent via the MQTT protocol to the physical feedback execution unit of the physical perception and interaction layer. The third type is task push instructions, such as "Start the 'Color Emotion Expression' virtual simulation task." This instruction is directly injected into the learner's creative process queue and automatically pops up after the learner completes the current task.

[0043] The entire system's data flow follows a strict layered and collaborative principle. The data synchronization layer, built on Apache Kafka, constructs a distributed message queue to ensure data transmission latency from the physical perception layer to the twin construction layer is less than 50 milliseconds, with a throughput supporting 100,000 messages per second. The business logic layer adopts a microservice architecture, with each analysis module deployed as a Docker container, communicating efficiently via the gRPC protocol, and utilizing Redis to cache intermediate computation results to improve response speed. The interactive presentation layer, developed based on the WebGL and React frameworks, supports the simultaneous display of the twin visualization interface, ability analysis reports, and intervention content on multiple terminals, including teacher and learner terminals. All interface elements support real-time data binding and dynamic updates.

[0044] Through close collaboration at all levels, this system achieves digital mirroring, objective evaluation, and intelligent intervention of the entire art and design creation process. Every creative action of the learner in the physical space is accurately captured and mapped to a virtual twin. Based on the full amount of process data, the system generates a multi-dimensional ability profile and proactively pushes personalized teaching strategies and resources, thereby constructing a new paradigm of intelligent art education that is data-driven and features closed-loop feedback.

[0045] Building upon the aforementioned embodiments, this embodiment further expands the system's application scenarios and data fusion dimensions to adapt to more complex cross-media art creation teaching needs. Specifically, this embodiment adds the ability to perceive sound creation and dynamic image creation behaviors in the physical perception and interaction layer, and introduces a spatiotemporal alignment and cross-analysis mechanism for multimodal works in the digital twin construction and simulation layer.

[0046] The physical perception and interaction layer has been expanded with the deployment of an audio acquisition array and a motion capture device. The audio acquisition array consists of eight high-fidelity microphones arranged in a ring, recording the learner's audio signals during the sound art creation process at a sampling rate of 48 kHz, including vocal singing, instrumental performances, or electronic sound effects synthesis. The motion capture device employs a multi-view high-speed camera system, synchronously recording the learner's body movements, facial expressions, and interactions with props during the creation of video installations or performing arts at a rate of 60 frames per second. These new data sources are also given precise timestamps and transmitted to the digital twin construction and simulation layer via a dedicated data channel.

[0047] The digital twin construction and simulation layer correspondingly expands the functionality of the twin data fusion engine. The engine can now handle five data modalities: 2D images, 3D point clouds, pen stroke and gesture sensing, environmental parameters, audio signals, and dynamic images. The fusion process employs a multi-level temporal alignment strategy: first, using image frames as a baseline, other modal data are mapped to the same temporal grid through linear interpolation or nearest neighbor matching; for audio signals, Mel-frequency cepstral coefficients (MFCC) are used to extract features, and cross-modal alignment verification is performed with lip movements in the images to ensure data authenticity. The fused twin state object adds "audio feature vector" and "action semantic label" fields. The former includes indicators such as rhythm stability, pitch variation range, and spectral complexity, while the latter is extracted from dynamic images using a pre-trained action recognition model (such as one based on a 3D convolutional neural network), such as "gesture emphasis," "body rotation," and "facial expression exaggeration."

[0048] The capability modeling and analysis layer has also been upgraded accordingly. The feature extraction module can now extract new process features from audio and motion data, such as the improvisation index in sound creation (based on Markov transition entropy of pitch sequences) and the motion coherence index in video creation (based on the smoothness of joint trajectories). The multidimensional capability quantification module adds two new dimensions: "vocal performance capability" and "dynamic narrative capability," and the corresponding deep neural network subnetworks are retrained on the expanded dataset. The development trajectory prediction module incorporates these two new dimensions into the time series prediction scope, forming a seven-dimensional capability development model.

[0049] The strategy map for instructional decision-making and intervention has been expanded in tandem, adding intervention strategies for sound and motion picture creation, such as "rhythm stability training" and "body language and emotional expression matching exercises." The resource recommendation engine has also been integrated with the audio material library and motion picture case library, enabling it to recommend relevant music clips, rhythm templates, or classic performance videos.

[0050] Through the above expansion, this embodiment extends the application boundaries of digital twins from static visual arts to dynamic, multi-sensory art creation, enabling the system to fully support the growing demand for cross-media creative ability cultivation in contemporary art education, and further enhancing the system's universality and teaching value.

Claims

1. An information management system for cultivating art and design skills based on digital twins, characterized in that: include: The physical perception and interaction layer is used to collect multimodal data during the physical creation process and feed back virtual commands to the physical creation environment; The digital twin construction and simulation layer is used to receive and fuse multimodal data streams from the physical perception and interaction layer, construct and drive a digital twin synchronized with the physical creation process, and provide a virtual simulation environment; The capability modeling and analysis layer is used to conduct in-depth analysis of the full-process data recorded by the digital twin, build a multi-dimensional capability quantification model of the learner, and conduct developmental assessment and bottleneck diagnosis. The instructional decision-making and intervention layer is used to generate personalized instructional decisions and trigger corresponding intervention actions based on the output of the competency modeling and analysis layer.

2. The information management system for cultivating art and design abilities based on digital twins according to claim 1, characterized in that, The physical perception and interaction layer includes a high-precision image acquisition device, a 3D scanning device, a pressure and posture sensing device, an environmental parameter sensing device, and a physical feedback execution unit. The high-precision image acquisition device is used to capture a two-dimensional image of the creation surface. The 3D scanning device is used to acquire point cloud data of the three-dimensional model or installation artwork. The pressure and posture sensing device is used to acquire data on pen pressure, pen speed, acceleration, and the creator's hand and body posture. The environmental parameter sensing device is used to monitor the light intensity, color temperature, and ambient sound decibels of the creation space. The physical feedback execution unit is used to receive instructions and dynamically adjust the lighting conditions, reference image projection, and surface angle of the physical creation environment.

3. The information management system for cultivating art and design abilities based on digital twins according to claim 2, characterized in that, The digital twin construction and simulation layer includes a twin data fusion engine, a real-time rendering engine, and a parametric simulation engine. The twin data fusion engine is used to spatiotemporally align and correlate two-dimensional image sequences, three-dimensional point cloud sequences, brush stroke sensor data, and environmental parameters based on a unified time reference, generating a structured twin temporal state dataset. The real-time rendering engine is used to render the brush stroke layer evolution animation of a two-dimensional work or the geometric shape growth animation of a three-dimensional work in real time in virtual space based on the twin temporal state dataset. The parametric simulation engine is used to allow non-destructive parameter modification and effect preview of the work in a virtual environment based on the current state of the twin.

4. The information management system for cultivating art and design abilities based on digital twins according to claim 3, characterized in that, The capability modeling and analysis layer includes a feature extraction module, a multi-dimensional capability quantification module, and a development trajectory prediction module; the feature extraction module is used to extract predefined creative process feature vectors from the twin's temporal state data. The multidimensional ability quantification module is used to input the extracted feature vectors into a set of pre-trained ability assessment models and output the learner's standardized scores on multiple core dimensions; the development trajectory prediction module is used to predict the learner's development trend curves on each ability dimension in the future teaching cycle based on the learner's historical multidimensional ability score sequence and to identify bottleneck dimensions.

5. The information management system for cultivating art and design skills based on digital twins according to claim 4, characterized in that, The teaching decision-making and intervention layer includes a strategy matching engine, a resource recommendation engine, and an intervention instruction generator. The strategy matching engine is used to match learners' current ability gaps and predicted development bottlenecks with the intervention strategy map in the teaching knowledge base. The resource recommendation engine is used to dynamically select and assemble personalized learning resource packages from the digital resource library based on the matched teaching strategies. The intervention instruction generator is used to transform decisions into executable instructions, which include feedback instructions, environmental adjustment instructions, and task push instructions.

6. The information management system for cultivating art and design abilities based on digital twins according to claim 5, characterized in that, The parametric simulation engine operates as follows: upon receiving a simulation request, a copy instance of the digital twin is created in memory; a set of parametric controls is provided, allowing users to adjust specific parameters in the copy instance, including the overall replacement of the color system, the displacement and scaling of compositional elements, the structural topology optimization of the 3D model, and the replacement of material textures; a real-time rendering engine performs visual rendering of the copy instance after parameter modification; all simulation operations and results are recorded and are not written back or affected in the data stream of the main twin corresponding to the ongoing entity creation process.

7. The information management system for cultivating art and design abilities based on digital twins according to claim 6, characterized in that, The feature extraction module calculates the line smoothness index as follows: extracts continuous coordinate points from the pen stroke sensing data sequence and calculates the rate of change of the direction angle between adjacent points; uses a sliding time window to smooth the rate of change of the direction angle sequence to filter out high-frequency jitter noise; the line smoothness index is defined as the reciprocal of the standard deviation of the smoothed sequence, which is calculated in segments according to the time window and finally incorporated into the feature vector of the creation process.

8. The information management system for cultivating art and design abilities based on digital twins according to claim 7, characterized in that, The matching logic of the strategy matching engine is based on a weighted rule network. This weighted rule network uses the difference between the learner's ability dimension score and the target score as input weights, with the larger the difference, the higher the weight of the dimension. Each intervention strategy in the strategy graph is labeled with its primary ability dimension label and secondary ability dimension label. The matching process calculates the matching degree between the label of each strategy and the learner's high-weight ability dimensions, and prioritizes recommending intervention strategies with a matching degree exceeding a preset threshold of 85%.

9. The information management system for cultivating art and design abilities based on digital twins according to claim 8, characterized in that, If multiple strategies have similar matching degrees, then further ranking and selection are performed by combining the learner's historical strategy execution performance data.

10. An information management method for cultivating art and design skills based on digital twins, characterized in that: The information management system for cultivating art and design abilities based on digital twins, as described in any one of claims 1-9, is used to manage information on cultivating art and design abilities.

Images