A live pig slaughtering technology virtual and simulation training system

Abstract

The application provides a kind of live pig slaughtering technology virtual and simulation training system, relating to virtual reality technical field.The live pig slaughtering technology virtual and simulation training system includes data acquisition module, based on live pig slaughtering process and veterinary hygiene inspection specification, adopts multi-source data fusion method, and the data of live pig anatomical structure, slaughtering operation and pathological characteristics are collected, labeled and standardized, to generate original training data.Through the actual business object of live pig slaughtering, anatomical structure, operation process and pathological morphology are taken as unified data core, forming a continuous processing link driven by real business rules, so that the data always maintains consistent semantics and traceable relationship from acquisition, modeling, simulation to evaluation, and the structured and standardized processing of multi-source heterogeneous data is completed in the acquisition stage, so that the subsequent three-dimensional reconstruction and physical property mapping no longer stay in appearance reproduction.

Description

Technical Field

[0001] This invention relates to the field of virtual reality technology, specifically to a virtual and simulation training system for pig slaughtering technology. Background Technology

[0002] Virtual reality (VR) technology is a system based on computer graphics, 3D modeling, human-computer interaction, sensing, and simulation technologies. It constructs immersive, interactive, and perceptible virtual environments through the collaborative development of hardware and software. This technology can digitally model and simulate real-world objects, scenes, and their operational processes, providing users with near-realistic visual, auditory, and operational feedback experiences within a virtual environment. One example is a virtual and simulation training system for pig slaughtering technology. This system, built on VR technology, is used for digital simulation and interactive training of pig slaughtering and inspection procedures, operational standards, and pathological identification processes. Its purpose is to provide slaughtering inspection personnel with an immersive and repeatable training environment, enabling training in slaughtering operations, pathological identification, operational behavior evaluation, and skill enhancement. This reduces actual training costs and improves personnel skill levels and operational standardization.

[0003] Existing virtual reality training programs often focus on general scenario demonstrations and basic interactions, emphasizing visual immersion while neglecting the internal structure and physical properties of real-world business objects. This leads to significant differences in feel, feedback, and results between virtual and real-world operations, requiring trainees to adapt to the real-world environment over a considerable period. Training content is typically based on pre-set cases or fixed pathological samples, limiting the number and types of samples and failing to cover complex and varied real-world testing scenarios. This can easily lead to path dependence on typical cases and insufficient ability to identify combined lesions or minor abnormalities. Operational assessments often rely on completion or correctness of results, lacking continuous analysis of operational sequence, process standardization, and detailed behaviors. This results in coarse assessments that fail to reflect true skill differences. Observational behaviors lack quantitative analysis during training, and no effective link is established between attention distribution and pathological identification results, easily leading to situations where critical areas are not adequately observed or detected in a timely manner. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a virtual and simulation training system for pig slaughtering technology. This system solves the problem that existing virtual reality training programs often focus on general scene displays and basic interactions, emphasizing visual immersion effects in actual operation while neglecting the internal structure and physical properties of real business objects. This results in significant differences between virtual and actual operations in terms of feel, feedback, and results, and trainees still need a long adaptation period in the real environment.

[0005] To achieve the above objectives, the present invention provides a virtual and simulation training system for pig slaughtering technology, comprising the following modules:

[0006] The data acquisition module, based on the pig slaughtering process and veterinary hygiene inspection standards, uses a multi-source data fusion method to collect, label, and standardize pig anatomical structure data, slaughtering operation data, and pathological characteristic data to generate raw training data.

[0007] The digital twin module, based on the original training data, uses a biological structure parameterized modeling method to perform three-dimensional reconstruction and physical property mapping of pig muscles, bones, internal organs and pathological tissues, generating a pig digital twin model.

[0008] The scenario construction module, based on the aforementioned pig digital twin model, adopts a process-driven scenario construction method and constructs virtual simulation inspection environments corresponding to head and hoof inspection, internal organ inspection, carcass inspection, and re-inspection according to the pig slaughter sequence and veterinary hygiene inspection post division rules, thereby generating virtual inspection scenarios;

[0009] The pathology generation module, based on the virtual inspection scenario, uses a pathology feature generation method to dynamically generate pathological features of pigs and embed them into the pig digital twin model through deep learning lesion morphology deduction and lesion severity control, thereby generating a pathology simulation sample set.

[0010] The interactive simulation module, based on the pathological simulation sample set, adopts an immersive interactive simulation method. Through motion capture, force feedback, and multi-channel interaction, it records the trainees' operation actions, examination sequence, and interaction status during the virtual examination process, generating trainee interactive behavior data.

[0011] The behavior analysis module, based on the trainee interaction behavior data, uses a time-series behavior recognition method to analyze and evaluate the trainee's operational standardization, process completeness, and verification consistency in order to generate operational ability assessment results.

[0012] The attention assessment module, based on the operational ability assessment results, uses an attention modeling method to evaluate the observation effectiveness and omission risk in the pathological identification process by matching and analyzing the trainee's eye movement trajectory with the key pathological attention areas, and generates pathological attention assessment results.

[0013] The adaptive training module, based on the pathological attention assessment results, uses an adaptive training decision-making method to dynamically adjust the difficulty level, pathological type, and training focus of the training content, generating a personalized training plan.

[0014] The feedback management module, based on the personalized training plan, summarizes and analyzes the training process and evaluation results of the trainees, and generates a visualized evaluation result and competency profile, as well as a training closed-loop report.

[0015] Preferably, the data acquisition module includes:

[0016] The structural acquisition submodule collects data on the outline of the pig's body surface, the proportion of its skeleton, and key anatomical structures to generate raw structural data of the pig.

[0017] The image analysis submodule, based on the original data of pig structure, performs tissue boundary analysis on the pig image data to generate tissue image analysis data;

[0018] The pathological annotation submodule annotates the pathological features of pigs based on the tissue image analysis data to generate pathological feature annotation data.

[0019] The standardization and organization submodule standardizes and organizes the pathological feature annotation data to generate original training data.

[0020] Preferably, the digital twin module includes:

[0021] The morphological modeling submodule constructs a three-dimensional morphological model of pigs based on the original training data to generate a morphological model of pigs.

[0022] The organization mapping submodule constructs a multi-layered organization structure based on the pig morphology model to generate an organization structure mapping model;

[0023] The parameter calibration submodule calibrates the tissue physical parameters based on the tissue structure mapping model to generate a calibrated tissue model.

[0024] The model validation submodule performs consistency validation based on the calibrated tissue model to generate a digital twin model of pigs.

[0025] Preferably, the scene construction module includes:

[0026] The process mapping submodule constructs the slaughtering operation process logic based on the pig digital twin model to generate slaughtering process mapping data;

[0027] The job zoning submodule constructs the spatial distribution of inspection jobs based on the slaughter process mapping data to generate an inspection job spatial model;

[0028] The environment rendering submodule generates a virtual slaughtering environment based on the inspection post space model to generate a simulation environment model;

[0029] The rhythm control submodule controls the work rhythm based on the simulation environment model to generate a virtual inspection scenario.

[0030] Preferably, the pathology generation module includes:

[0031] The lesion generation submodule generates pathological morphological data based on the virtual examination scenario;

[0032] The severity control submodule generates graded pathological data of different lesion severity based on the pathological morphology data;

[0033] The feature embedding submodule embeds pathological features into the pig digital twin model based on the graded pathological data to generate a pathological embedding model.

[0034] The sample expansion submodule generates a pathological simulation sample set based on the pathological embedding model.

[0035] Preferably, the interactive simulation module includes:

[0036] The motion acquisition submodule collects trainees' operational actions based on the pathological simulation sample set to generate operational action data;

[0037] The force feedback submodule generates interactive feedback data based on the operation action data;

[0038] The interaction recording submodule records the interaction process based on the interaction feedback data to generate an interaction process record;

[0039] The status synchronization submodule generates student interaction behavior data based on the interaction process records.

[0040] Preferably, the behavior analysis module includes:

[0041] The action recognition submodule generates action recognition results based on the student interaction behavior data;

[0042] The sequence verification submodule generates a process verification result based on the action recognition result;

[0043] The compliance analysis submodule generates compliance analysis results based on the process verification results.

[0044] The capability modeling submodule generates operational capability assessment results based on the compliance analysis results.

[0045] Preferably, the attention assessment module includes:

[0046] The line-of-sight analysis submodule generates line-of-sight trajectory data based on the operational capability assessment results;

[0047] The hotspot matching submodule generates attention matching results based on the gaze trajectory data;

[0048] The deviation calculation submodule generates attention deviation data based on the attention matching results;

[0049] The risk assessment submodule generates pathological attention assessment results based on the attention deviation data.

[0050] Preferably, the adaptive training module includes;

[0051] The difficulty adjustment submodule generates training difficulty parameters based on the pathological attention assessment results;

[0052] The key reinforcement submodule generates reinforcement training content based on the training difficulty parameters;

[0053] The content reorganization submodule generates a combination of training content based on the enhanced training content;

[0054] The path planning submodule generates personalized training schemes based on the training content.

[0055] Preferably, the feedback management module includes:

[0056] The results summary submodule generates training result summary data based on the personalized training scheme.

[0057] The competency profiling submodule generates a trainee competency profile based on the aggregated training results data.

[0058] The report generation submodule generates a training assessment report based on the trainee competency profile.

[0059] The document management submodule generates a training closed-loop report based on the training evaluation report.

[0060] This invention provides a virtual and simulation training system for pig slaughtering technology. It has the following beneficial effects:

[0061] This invention uses anatomical structures, operational processes, and pathological morphology as a unified data core, centered around the actual business operations of pig slaughtering. This forms a continuous processing chain driven by real business rules, ensuring consistent semantics and traceability throughout the data acquisition, modeling, simulation, and evaluation stages. Multi-source heterogeneous data undergoes structuring and standardization during the acquisition phase, enabling subsequent 3D reconstruction and physical attribute mapping to go beyond mere appearance reproduction. It reflects the mechanical differences and pathological changes of different tissues during the operation process, allowing virtual objects to exhibit response characteristics close to real inspection objects during interaction. The virtual environment constructed around the slaughtering sequence and job division establishes a one-to-one correspondence between the training process and the actual inspection procedures, avoiding cognitive biases caused by the disconnect between the scenario and business operations. The dynamic generation of pathological features within the virtual objects... The integration and embedding of data allow samples to move beyond static or fixed types, covering varying degrees of severity and combinations, significantly expanding training coverage. Based on records of real operational actions, interactive states, and temporal behaviors, assessment results are built upon complete behavioral trajectories rather than single-point judgments, thereby improving the accuracy of identifying operational standardization and process integrity. Combining line-of-sight trajectory with matching analysis of key pathological areas establishes a correlation between observational behavior and identification results, uncovering potential omission risks. Training content is dynamically reorganized based on assessment results, allowing training paths to adjust with changes in ability, avoiding efficiency decline caused by repetitive training. Assessment results are preserved in the form of competency profiles and process records, forming a traceable training loop, providing a basis for continuous improvement, and enhancing the overall authenticity, relevance, and sustainability of training. Attached Figure Description

[0062] Figure 1 This is a system block diagram of the present invention;

[0063] Figure 2 This is a schematic diagram of the data acquisition module of the present invention;

[0064] Figure 3 This is a schematic diagram of the digital twin module of the present invention;

[0065] Figure 4 This is a schematic diagram of the scene construction module of the present invention;

[0066] Figure 5 This is a schematic diagram of the pathology generation module of the present invention;

[0067] Figure 6 This is a schematic diagram of the interactive simulation module of the present invention;

[0068] Figure 7 This is a schematic diagram of the behavior analysis module of the present invention;

[0069] Figure 8 This is a schematic diagram of the attention assessment module of the present invention;

[0070] Figure 9 This is a schematic diagram of the adaptive training module of the present invention;

[0071] Figure 10 This is a schematic diagram of the feedback management module of the present invention. Detailed Implementation

[0072] 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.

[0073] Example:

[0074] like Figure 1-10 As shown in the figure, this embodiment of the invention provides a virtual and simulation training system for pig slaughtering technology, including the following modules:

[0075] The data acquisition module, based on the pig slaughtering process and veterinary hygiene inspection standards, uses a multi-source data fusion method to collect, label, and standardize pig anatomical structure data, slaughtering operation data, and pathological characteristic data to generate raw training data.

[0076] The described digital twin module, based on the original training data, uses a biological structure parameterized modeling method to perform three-dimensional reconstruction and physical property mapping of pig muscles, bones, internal organs and pathological tissues, generating a digital twin model of pigs.

[0077] The scenario construction module, based on the digital twin model of pigs, adopts a process-driven scenario construction method. According to the pig slaughter sequence and the division rules of veterinary hygiene inspection positions, it constructs virtual simulation inspection environments for head and hoof inspection, internal organ inspection, carcass inspection and re-inspection, and generates virtual inspection scenarios.

[0078] The pathology generation module, based on a virtual inspection scenario, adopts a pathology feature generation method. Through deep learning of lesion morphology deduction and lesion severity control, it dynamically generates pathological features of pigs and embeds them into the digital twin model of pigs to generate a pathology simulation sample set.

[0079] The interactive simulation module, based on a pathological simulation sample set, adopts an immersive interactive simulation method. Through motion capture, force feedback, and multi-channel interaction, it records the trainees' operational actions, examination sequence, and interactive status during the virtual examination process, generating trainee interactive behavior data.

[0080] The behavior analysis module, based on trainee interaction behavior data, uses time-series behavior recognition methods to analyze and evaluate the trainees' operational standardization, process completeness, and inspection consistency in order to generate operational ability assessment results.

[0081] The attention assessment module, based on the operational ability assessment results, uses an attention modeling method to evaluate the observation effectiveness and omission risk in the pathological identification process by matching and analyzing the trainees' eye movement trajectories with key pathological attention areas, and generates pathological attention assessment results.

[0082] The adaptive training module, based on the pathological attention assessment results, uses an adaptive training decision-making method to dynamically adjust the difficulty level, pathological type, and training focus of the training content, generating a personalized training plan.

[0083] The feedback management module, based on personalized training plans, summarizes and analyzes the training process and evaluation results of trainees, generates visualized evaluation results and competency profiles, and produces a training closed-loop report.

[0084] The data acquisition module includes:

[0085] The structural acquisition submodule collects data on the outline of the pig's body surface, the proportion of its skeleton, and key anatomical structures to generate raw structural data of the pig.

[0086] During data collection, the first step is to collect data on the pig's body contour, skeletal proportions, and key anatomical structures. This data is crucial for understanding the pig's overall structure. Using laser scanning or 3D imaging technology, the pig's body contour is accurately acquired and quantified, generating corresponding raw structural data. This data can be collected using precision instruments and high-resolution cameras, capturing the pig's skeletal proportions and other anatomical landmarks point-by-point, such as vertebral position, rib number and distribution, ensuring that each data point reflects the pig's actual body shape and structure. For example, in practical operation, laser scanning technology is used to accurately measure the ratio between the pig's head and chest cavity, determining the skeletal proportions of a specific individual in this area, and storing the data in a database for subsequent analysis.

[0087] The image analysis submodule, based on the original data of pig structure, performs tissue boundary analysis on the pig image data to generate tissue image analysis data;

[0088] When performing image analysis on the raw anatomical data of collected pigs, high-resolution imaging technology is first used to photograph each anatomical part of the pig, and boundary recognition is performed to obtain the precise location and morphological information of each tissue. Software tools are then used to automatically identify boundaries in the image data; for example, image processing algorithms are used to extract and classify the edges of different tissues, such as segmentation algorithms to distinguish the boundaries of muscles, bones, and skin, and to generate image analysis data. For instance, if a lesion is present in the abdominal area of ​​a sample, the analysis process can accurately mark the lesion area, providing precise data for subsequent analysis. This data provides the foundation for subsequent pathological annotation and analysis.

[0089] The pathology annotation submodule annotates pathological features of pigs based on tissue image analysis data to generate pathological feature annotation data.

[0090] In the pathological annotation stage, deep learning algorithms are used to annotate pathological features in the image data. Existing pathological datasets and expert annotation information are used to train a pathological recognition model, which automatically annotates lesion areas in pig images, generating pathological feature annotation data. This annotation data includes the precise location of tumors, inflammation, or other lesions. For example, based on an existing pathological image database, the algorithm identifies inflammatory responses in the intestinal region and specifically marks this lesion area during annotation to ensure accuracy. The annotated pathological feature data provides a standardized basis for subsequent training data processing, ensuring the consistency and high quality of the system's training data.

[0091] The standardization and organization submodule standardizes and organizes pathological feature-annotated data to generate original training data.

[0092] Data standardization involves structuring and standardizing the aforementioned collected and labeled data. Using data cleaning and formatting tools, pathological feature data is formatted in a unified way and combined with anatomical data of live pigs to generate raw training data that meets training requirements. This process includes removing duplicate data, standardizing units, and standardizing table formats to ensure data accuracy and consistency. For example, when processing data on different body sizes of live pigs, it may be necessary to standardize units from different data sources to ensure that all data formats are consistent with the annotations, ultimately outputting a standard dataset that can be input into the simulation training system.

[0093] The digital twin module includes;

[0094] The morphological modeling submodule constructs a three-dimensional morphological model of pigs based on the original training data to generate a morphological model of pigs;

[0095] Based on the original training data, advanced 3D modeling technology is used to construct a 3D morphological model of a pig. First, anatomical data of the pig, such as skeletal proportions, body contours, and key anatomical points, is collected and modeled using computer graphics software (such as Blender or 3ds Max). By defining specific dimensional parameters of the pig's body shape, a realistic 3D model is constructed, ensuring the accurate reproduction of each anatomical feature. In practical applications, assuming that a 150 kg pig needs to be modeled during training, its length and body contours are measured to generate a 3D model consistent with the actual pig, thus providing an interactive object for the virtual training system. The accuracy of the model will affect the effectiveness of subsequent simulation training, ensuring that trainees can perform operational training in a highly realistic environment.

[0096] The organization mapping submodule constructs a multi-layered organizational structure based on the morphological model of pigs to generate an organizational structure mapping model;

[0097] After generating a 3D morphological model of the pig, a hierarchical mapping of its tissue structure is performed. A hierarchical modeling method is used to separate and map the various tissue layers of the pig, including muscle, bone, fat, and skin. This process uses digital technology to bind the physical properties (such as elasticity and density) of each tissue to its tissue type. For example, by using the physical characteristics of each layer in the aforementioned model, modeling and mapping are performed. For instance, a higher density value is set for the pig's skeleton, while a lower density value is set for the muscle, generating a multi-layered tissue structure mapping model to ensure that the model accurately reflects the structure and physical characteristics of the pig's internal tissues. The accuracy at this stage directly affects the subsequent training and simulation results, ensuring that different tissues can respond to external interactive feedback.

[0098] The parameter calibration submodule calibrates the tissue physical parameters based on the tissue structure mapping model to generate a calibrated tissue model.

[0099] After mapping the pig's tissue structure model, parameter calibration is performed based on the physical properties of each layer of pig tissue, such as elasticity, density, and temperature. This calibration process adjusts the simulation parameters according to the physical data of actual anatomical samples. For example, the simulation parameters are corrected one by one based on the actual responsiveness of pig muscles and the elasticity of skin. Assuming that muscle elasticity needs to be calibrated by adjusting simulation parameters to match actual anatomical data during simulation, this process involves dynamically adjusting the physical parameters of each tissue layer to ensure consistency with reality. Through this calibration process, a calibrated tissue model that accurately reflects the real pig body is generated, making the trainee's interactive experience more realistic in subsequent simulation operations, and enabling the model to accurately respond to training movements.

[0100] The model validation submodule performs consistency verification based on the calibrated tissue model to generate a digital twin model of pigs.

[0101] After calibrating the physical parameters of pig tissues, the model undergoes consistency verification to ensure its reliability and accuracy as a digital twin. This process typically includes two parts: First, verifying the geometric consistency of the model by comparing it with the anatomical features of actual pigs to confirm the correct proportions and positions of each part. Second, performing physical consistency verification using standardized verification tests, such as applying certain mechanical pressure or simulating operations, to examine the model's feedback in actual operation. For example, by applying certain pressure or simulating collisions, the elastic response of the muscle layer is examined to see if it matches reality. If deviations are found, the corresponding model parameters are adjusted for correction. This verification process ensures that the generated digital twin model of pigs has high accuracy and consistency in different scenarios, guaranteeing that the final digital twin model can run reliably in actual training.

[0102] The scene construction module includes:

[0103] The process mapping submodule constructs the slaughter operation process logic based on the pig digital twin model to generate slaughter process mapping data;

[0104] Based on the established digital twin model of pigs, the various operational nodes involved in the slaughtering process are broken down. A single operation process is subdivided into multiple consecutive steps such as entry, positioning, processing, and transfer. The sequence, duration, and interrelationship of each step are analyzed. During the actual execution, by reading the pig's body size, posture, and location information recorded in the digital twin model, the space occupation and time consumption required for each operation node are calculated item by item. For example, using the average processing time of 60 seconds per pig as a benchmark, the actual time consumption of different nodes is compared. When the time consumption of a node exceeds the benchmark by 20%, it is judged as a slow-paced interval; if it is less than the benchmark by 10%, it is classified as a fast-paced interval. Based on this, the process sequence is rearranged or merged. During the simulation, by repeatedly calculating the overall time consumption data under different sequences, process combinations with high continuity and few interruptions are selected. The node sequence, time parameters, and state switching conditions corresponding to the combination are recorded as process mapping results, forming data content that can be directly called.

[0105] The job zoning submodule constructs the spatial distribution of inspection jobs based on slaughter process mapping data to generate an inspection job spatial model;

[0106] Using the above process mapping results as input, the inspection-related positions are spatially decomposed. The manual operation points and equipment operation points involved in the process are identified as independent position units. The workshop size parameters in the digital twin model are read, and the floor area of ​​each position unit is calculated. For example, three square meters is used as the reference benchmark value for a single inspection position. When the actual occupied area is less than 2.5 square meters, it is classified as a small interval, and when it is more than 3.5 square meters, it is classified as a large interval. Based on this, the relative positions between positions are compared. By calculating the straight-line distance between adjacent positions, when the distance is less than one meter, it is marked as a close interval, and when it is greater than two meters, it is marked as a dispersed interval. The position positions are adjusted multiple times in combination with the process sequence to keep the positions of the preceding and following processes within a reasonable distance range. Then, the coordinate position, occupied area, and mutual spacing parameters of each position are uniformly organized to form a stable and reusable position spatial distribution result, which serves as the output of the spatial model.

[0107] The environment rendering submodule generates a virtual slaughtering environment based on the inspection post space model, in order to generate a simulation environment model;

[0108] After the job space model is determined, the spatial coordinates, equipment dimensions, and aisle width parameters of each job position are read. The outlines of the ground, walls, and equipment in the virtual environment are constructed item by item. During the execution process, by comparing with commonly used size data in real workshops, the aisle width is divided into a narrow range of less than 1.2 meters, a normal range of 1.2 to 1.8 meters, and a wide range of more than 1.8 meters. The aisle shape is adjusted according to the distribution of jobs. At the same time, parameters such as light intensity and ground material in the environment are numerically set. For example, based on the light value of the normal work area, the darker or brighter areas are compared and corrected. The image value state under different parameter combinations is recorded in multiple rendering tests. The combination that meets the spatial parameter requirements is retained. Finally, the environmental structure parameters, rendering parameters, and spatial relationships are encapsulated to generate a complete virtual environment data model.

[0109] The rhythm control submodule controls the work rhythm based on the simulation environment model to generate a virtual inspection scenario.

[0110] Based on a virtual environment model, parameters related to the work rhythm are set. The processing time of each node recorded in the process mapping is used as the initial data. The overall work cycle is broken down, and a threshold for the number of times the next node can be entered per unit time is set. For example, one pig per minute is used as the benchmark value for the rhythm. When the actual operation exceeds this value three times in a row, it is classified as too fast. If it is less than 20% of the benchmark value, it is classified as too slow. During the simulation operation, the time consumption value of each node is continuously recorded and compared with the benchmark value item by item. By repeatedly adjusting the node waiting time and entry order, the overall rhythm parameters are kept within the preset range. Finally, the rhythm parameters, node switching conditions and time recording data are summarized to form virtual inspection scenario data that can be called.

[0111] The pathology generation module includes:

[0112] The lesion generation submodule generates pathological morphology data based on a virtual examination scenario;

[0113] Based on the virtual inspection scene data generated previously, the appearance, tissue structure display, and inspection perspective parameters of individual pigs in the scene are broken down item by item to extract basic information that can be used to construct pathological morphology. During execution, the surface areas of the pigs in the virtual scene are first divided and marked, with the head, trunk, limbs, etc., numbered separately, and the morphological parameter values ​​of the corresponding areas under normal conditions are read. Then, the abnormal morphologies that can be observed during the inspection are compared and analyzed. For example, the change in surface texture is recorded in the trunk area. When the texture change value falls within 10% of the normal reference range, it is classified as a slight change range, and when it exceeds 30%, it is classified as a significant change range. On this basis, by statistically analyzing the data of the same area in multiple scene runs, the combinations of recurring abnormal areas are selected and used as input conditions for constructing pathological morphology. At the same time, a corresponding set of morphological parameters is assigned to each type of abnormal morphology, and finally, a set of pathological morphological data that can be called is formed.

[0114] The severity control submodule generates graded pathological data of different lesion severity based on pathological morphology data;

[0115] Using the aforementioned pathological morphology data as the basic input, the numerical characteristics corresponding to each type of lesion morphology are graded. In specific execution, parameters such as area ratio, color deviation degree, and morphological distortion amplitude involved in the pathological morphology are first read, and reference benchmark values ​​are set for each. For example, the area ratio of the normal morphology is used as the benchmark value. When the lesion area ratio is lower than the benchmark value plus 15%, it is classified as a low severity interval; when it is higher than 40%, it is classified as a high severity interval; and the middle range is classified as a medium severity interval. During this process, different parameters are compared horizontally. When multiple parameters fall into the same interval, their severity levels are determined to be consistent. If there are distribution differences, values ​​are taken according to the pre-set weight order. For example, the area ratio weight is set to a higher level, and the color deviation weight is set to a medium level. The stability of the grading results is verified by substituting multiple rounds of sample data. Finally, the grade label and parameter combination corresponding to each lesion morphology are recorded together to form the graded pathological data.

[0116] The feature embedding submodule embeds pathological features into the pig digital twin model based on graded pathological data to generate a pathological embedding model;

[0117] After obtaining graded pathological data, it is matched item by item with existing pig digital twin models. During the process, the tissue hierarchy in the digital twin model is first located to determine the specific tissue layer and spatial location corresponding to the pathological features. For example, the pathological features of the skin layer are limited to the range of surface tissue numbers. Then, the corresponding morphological parameter values ​​in the graded pathological data are read and embedded into the corresponding parameter slots of the digital twin model. During the embedding process, the original parameters are compared with the pathological parameters. When the pathological parameters deviate from the original parameters by more than a preset range, a replacement operation is performed. If they do not deviate, an overlay adjustment is performed. At the same time, the parameter changes involved in each embedding operation are recorded. The consistency of parameters is verified by repeatedly embedding multiple sample models, a stable embedding method is selected, and the embedded model state is saved to form digital twin model data containing pathological features.

[0118] The sample expansion submodule generates a pathological simulation sample set based on the pathological embedding model.

[0119] Based on the digital twin model with completed pathological feature embedding, the sample generation process is extended. During execution, a baseline value for the number of expanded samples is first set, for example, ten samples corresponding to a single pathological grade as the initial setting. Then, multiple approximate but not completely identical model instances are generated by adjusting the subtle range of pathological parameters. The parameter combinations used in each generation process are recorded. When the parameter variation is below the set threshold, they are classified as samples of the same type; when it is above the threshold, they are classified as different sample types. By statistically analyzing the parameter distribution of the generated samples, it is determined whether the range of values ​​for each severity interval is covered. If the number of samples in a certain interval is too small, the parameter values ​​are adjusted again to supplement them. Finally, all generated model instances are organized according to pathological grade and feature type to form a pathological simulation sample data set with a clear structure and traceable parameters.

[0120] The interactive simulation module includes:

[0121] The motion acquisition submodule collects trainees' operational actions based on a pathological simulation sample set to generate operational action data;

[0122] Based on the established pathological simulation sample set, the operational behaviors of trainees in the simulation environment were decomposed and collected. Specifically, the trainees' operations on the interactive device were first broken down into multiple action parameters such as position movement, angle change, and contact duration, and recorded in chronological order. During the collection process, the displacement values ​​between consecutive sampling points were compared. When the displacement value per unit time was less than a preset benchmark interval, it was determined to be a slow operation interval; when it was greater than the upper limit of the interval, it was classified as a fast operation interval. For example, using a movement of five millimeters per second as a reference benchmark, when three consecutive sampling results exceeded eight millimeters, it was recorded as a relatively fast state. At the same time, the operation duration was statistically analyzed. When the duration of a single operation was less than two seconds, it was classified as a short-duration operation interval; when it was more than five seconds, it was classified as a long-duration operation interval. The above parameters were filtered from multiple operation samples, and obviously abnormal jittery data was removed, retaining only stable and continuous data segments. The action parameters were then organized and stored in a unified format to finally form structured operation action data.

[0123] The force feedback submodule generates interactive feedback data based on the operation action data;

[0124] Using the compiled operation data as input, the interactive feedback generated during the trainee's operation is generated item by item. During execution, the contact position, the trend of applied force change, and the duration in the action data are analyzed first. The contact position is mapped to the corresponding pathological simulation sample area. Then, based on the force change record, it is determined whether it is in the low intensity range, medium intensity range, or high intensity range. For example, using the conventional operation force calibrated by the equipment as the benchmark, when the actual collected value is lower than 20% of the benchmark, it is classified as a low intensity range, and when it is higher than 30%, it is classified as a high intensity range. At the same time, the stability of the force change is compared. When the difference between adjacent sampling points is small, it is judged as a stable input, and when the difference fluctuates greatly, it is marked as an unstable input. Based on this, the position, force range, and change state are combined and recorded, and corresponding feedback parameter items are generated for each interaction. After multiple operation data verifications, a complete interactive feedback data set is formed.

[0125] The interaction recording submodule records the interaction process based on interaction feedback data to generate an interaction process record;

[0126] Based on the generated interactive feedback data, the interaction of trainees throughout the simulation process is continuously recorded. During execution, each piece of feedback data is numbered according to the timeline and associated with the corresponding operation stage. For example, feedback data generated continuously under the same pathological sample is grouped into the same interaction segment. The number of feedbacks, duration, and distribution of feedback types within each segment are statistically analyzed. When a certain type of feedback accounts for more than 50% of the total, it is recorded as the main interaction type, and when it is less than 20%, it is recorded as the secondary type. At the same time, the interaction interruption is judged. When the time interval between two adjacent feedback data exceeds the preset range, it is marked as an interruption point. After all the recording is completed, the time stamps, feedback parameters, and interruption markers are organized together to form a complete interactive process record data.

[0127] The status synchronization submodule generates student interaction behavior data based on the interaction process records.

[0128] After obtaining the complete interaction process record, the trainees' interaction behavior is synchronously organized. During the execution process, the interaction records are first classified according to the trainee's identifier, and the interaction records of the same trainee under different pathological samples are summarized separately. Then, the operation frequency, feedback intensity distribution, and number of interruptions in each group of records are compared and analyzed. For example, when a trainee's operation frequency in the same type of sample is significantly higher than the average range of other trainees, their operation rhythm is classified into the faster range, and if it is lower than the average range, it is classified into the slower range. At the same time, the stability of the interaction behavior over time is judged. When the fluctuation amplitude of multiple recorded parameters is close, it is considered a stable state. When the difference is large, it is recorded as a fluctuating state. Finally, all kinds of behavioral parameters are synchronously output according to a unified field to form trainee interaction behavior data that can be directly called.

[0129] The behavior analysis module includes;

[0130] The action recognition submodule generates action recognition results based on student interaction behavior data;

[0131] Based on the previously generated student interaction data, the recorded operation trajectory, operation rhythm, and interaction state changes are broken down and processed. During execution, each interaction is first segmented according to time sequence. Data that is continuous and the interval is less than a preset time base is divided into the same action segment. Then, the displacement change amplitude, directional consistency, and duration of each action segment are compared item by item. For example, when the displacement change in a segment is concentrated in a similar numerical range and the directional deviation is less than a preset range, it is marked as the same type of action. When the directional change is frequent and the amplitude fluctuates greatly, it is classified as another type of action. On this basis, an action reference library is introduced, and the parameter combination of the current segment is compared item by item with the existing action parameters in the reference library. When multiple key parameters fall into the same judgment range, the corresponding action category is confirmed. At the same time, the matching degree is numerically recorded, for example, the similarity is recorded in a percentage system. When the similarity is less than 60, it is not classified and enters the unconfirmed set. Through repeated comparison and screening of multiple data samples, a clear action recognition result is finally generated for each student operation.

[0132] The sequence verification submodule generates process verification results based on the action recognition results;

[0133] After obtaining clear action recognition results, the sequence of actions performed by the trainee in the complete operation process is verified. During execution, a standard action sequence list is first extracted according to the established operation process document, and a sequence number is assigned to each standard action. Then, the trainee's action recognition results are arranged in chronological order and compared item by item with the standard sequence. For example, when the trainee's first action number matches the first item in the standard list, it is recorded as a sequence match. If there is a jump or repetition of the number, it is marked as a sequence deviation. At the same time, the number and position of the sequence deviation are counted. When the number of deviations is less than 10% of the total number of actions, it is classified as a slight deviation range, and when it is more than 30%, it is classified as a significant deviation range. During the verification process, missed actions and multiple actions are recorded separately. When a certain standard action does not appear in the complete sequence, it is counted as a missing item. When the same action appears consecutively more than a preset number of times, it is counted as a redundant item. Finally, the sequence matching results, deviation ranges, and abnormal item list are integrated to form the process verification results.

[0134] The compliance analysis submodule generates compliance analysis results based on process verification results;

[0135] Based on the generated process verification results, the compliance status of trainees' operations is analyzed item by item. During execution, the sequence deviation markers, omission records, and redundant records in the verification results are first read and classified according to pre-set compliance judgment rules. For example, records with no omissions and sequence deviations in the mild range are classified as high compliance range; records with a single omission or multiple sequence deviations are classified as medium compliance range; and records with multiple omissions and obvious sequence deviations are classified as low compliance range. Different types of problems are assigned different weights during the analysis process. For example, omission problems have a higher weight than redundant problems. All types of problems occurring within the same operation process are weighted and summarized. By comparing multiple trainee samples, the stable range of compliance range division is confirmed. Finally, the compliance level and problem distribution corresponding to each complete operation are output together to form the compliance analysis results.

[0136] The capability modeling submodule generates operational capability assessment results based on compliance analysis results.

[0137] After obtaining the compliance analysis results, the overall operational capabilities of trainees are modeled. During the process, the compliance analysis results of the same trainee in different training scenarios are first summarized, and the number of occurrences in the high compliance, medium compliance, and low compliance intervals is counted. The proportion of each interval is calculated. For example, when the high compliance interval accounts for more than 70% of all records, the trainee's stable performance parameters are recorded as a higher interval. When the proportion of the low compliance interval exceeds 30%, it is recorded as a lower interval. At the same time, the action stability data recorded in the preceding action recognition is combined to judge the consistency of the trainee's performance in different operation stages. When the parameter fluctuations are small in multiple scenarios, it is recorded as a stable state. When the fluctuation range is large, it is recorded as an unstable state. Finally, the compliance interval distribution, problem type weight, and stability label are comprehensively sorted to generate a structured operational capability assessment result.

[0138] The attention assessment module includes:

[0139] The line-of-sight analysis submodule is used to generate line-of-sight trajectory data based on the operational capability assessment results;

[0140] Based on the established operational ability assessment results, the attention allocation of trainees during simulation operations is broken down and processed. During execution, the assessment results are first read to determine the operational phase divisions, stable action intervals, and abnormal frequency markers, thus identifying the operational time periods requiring focused attention. Then, the trainee's gaze tracking records within these time periods are extracted synchronously. The continuously sampled gaze points are arranged chronologically, and the corresponding screen coordinates and dwell time for each gaze point are recorded. For example, if a coordinate point is sampled more than three times consecutively with a single dwell time exceeding 0.3 seconds, that point is marked as a valid gaze point. Points with dwell times less than 0.1 seconds are classified as rapid scanning intervals and discarded. Based on this, the valid gaze points are linked together to form continuous gaze movement paths. Simultaneously, the number of turns and dwell density within the paths are statistically analyzed. When the number of turns per unit time is in a low range, it is recorded as a stable trajectory; when it is in a high range, it is recorded as a frequent jump trajectory. Through these steps, structured gaze trajectory data is finally generated.

[0141] The hotspot matching submodule is used to generate attention matching results based on gaze trajectory data;

[0142] Based on the generated gaze trajectory data, the distribution of trainees' attention areas is matched. During execution, the key attention areas pre-marked in the pathological simulation scene are first read, and the area is divided into multiple hot zone units according to spatial range. Each hot zone unit is assigned a corresponding reference number. Then, the effective gaze points in the gaze trajectory are mapped to the corresponding hot zone units one by one. The cumulative number of gazes and the total dwell time in each hot zone are counted. For example, when the cumulative dwell time of a hot zone exceeds 40% of the total gaze time during the complete operation, the hot zone is recorded as a high attention interval. When it is between 20% and 40%, it is recorded as a medium attention interval. When it is below 20%, it is recorded as a low attention interval. At the same time, hot zones with multiple gaze back and forth are additionally marked to indicate that there is repeated attention behavior in the area. After completing the statistics of all hot zones, the actual attention distribution of trainees is compared with the expected attention distribution item by item to form the corresponding attention matching results.

[0143] The bias calculation submodule is used to generate attention bias data based on the attention matching results;

[0144] After obtaining the attention matching results, the deviation of the trainees' attention distribution is quantified. During the execution process, the attention interval marker corresponding to each hot zone is first read and compared with the preset standard attention interval. The part that matches is recorded as no deviation state, and the part that does not match is recorded as deviation state. Then, the deviation state is distinguished according to the direction. For example, when the actual attention interval is lower than the standard attention interval, it is recorded as insufficient attention, and when it is higher than the standard attention interval, it is recorded as excessive attention. At the same time, the number of deviations and their distribution ratio in the entire operation process are counted. When the deviation ratio is less than 15%, it is recorded as a slight interval, between 15% and 35%, it is recorded as a moderate interval, and above 35%, it is recorded as a significant interval. In this process, the degree of deviation of different hot zones is summarized to form attention deviation data that includes deviation type, deviation interval, and distribution location.

[0145] The risk assessment submodule is used to generate pathological attention assessment results based on attention bias data.

[0146] Based on the collected attentional deviation data, the attentional risk status of trainees in pathological simulation operations is assessed and processed. During the execution process, the deviation interval records generated by trainees in different operation stages are first summarized according to the trainee's dimension, and the occurrence frequency of each type of deviation interval is counted. Then, deviations that recur in consecutive stages are marked. For example, when the same key hot area is recorded as a significant deviation interval in more than three consecutive stages, the situation is listed as a high-risk record. When deviations only appear sporadically in non-consecutive stages, they are recorded as low-risk records. At the same time, combined with the distribution of deviation types, the numbers of insufficient attention and excessive attention are counted separately. When the proportion of one type is significantly higher than that of another type, it is marked separately. Finally, the deviation interval distribution, continuity characteristics, and type statistics results are integrated to form the corresponding pathological attention assessment results.

[0147] The adaptive training module includes:

[0148] The difficulty adjustment submodule generates training difficulty parameters based on pathological attention assessment results.

[0149] Based on the previously established conclusions regarding pathological attention risk, the deviation intervals, continuity markers, and type distributions of individuals across multiple training records were analyzed. During execution, the data from the five most recent training sessions were first laid out side-by-side, and a numerical label was assigned to the deviation level in each record (e.g., slight as 1 point, moderate as 2 points, significant as 3 points). Then, the scores for the same critical area across multiple records were accumulated. For example, if a trainee scored 2, 3, and 3 points for the same area in three records, the cumulative score would be 8 points. The difficulty range was then divided based on the cumulative score. Scores between three and five points are categorized into the basic range, scores between six and eight points into the intermediate range, and scores above nine points into the advanced range. A secondary check is performed using continuous marking; for example, if the same area is marked as a distinct range three times consecutively, it is directly upgraded to the advanced range. In a specific example, if a trainee has three consecutive deviations in chest image interpretation, the difficulty of their corresponding training item is recorded as a higher range. After completing the above grading and check, the range labels and values ​​corresponding to each training unit are organized and archived, outputting structured training difficulty parameters.

[0150] The key sub-module is enhanced by generating reinforcement training content based on training difficulty parameters;

[0151] Based on the established training difficulty parameters, the items in the training material library are screened and reorganized item by item. During the process, the tag information corresponding to each training material in the material library is first read, including the knowledge point type, the number of operation steps, and the historical average pass rate. Then, these tags are compared with the interval requirements in the difficulty parameters one by one. For example, when a student is marked as high-level, only materials with more than seven operation steps and a historical pass rate of less than 50% are selected. When marked as basic, materials with three to five steps and a historical pass rate of more than 60% are selected first. At the same time, additional screening conditions are set for knowledge points that are continuously marked as biased. For example, the relevant materials for this knowledge point must appear no less than ten times in the material library. In a specific example, if a student's cumulative score in the vision positioning problem reaches the high-level interval, the system will select twelve cases related to vision positioning with high step complexity from the material library, remove content with more than three repeated structures, and retain eight materials with obvious structural differences. Then, the selected material numbers, tags, and usage order are listed and organized to form the corresponding reinforcement training content.

[0152] The content reorganization submodule generates training content combinations based on reinforcement training content;

[0153] After obtaining the list of reinforcement items, the overall structure of the training content is re-examined. During the execution process, the duration of each piece of material is first counted, for example, marked as three minutes, five minutes, or eight minutes. Then, multiple pieces of material under the same knowledge point are compared side by side. Items with similar duration and highly similar step structures are merged, while items with large differences in duration and obvious structural differences are grouped and arranged. For example, among the eight pieces of material selected by a trainee, three are five minutes long and have a high degree of structural repetition, so two of them are retained, and the order of the remaining items is adjusted. At the same time, materials from different knowledge points are interspersed. For example, three consecutive image interpretation materials are split into three categories: image interpretation, operation positioning, and comprehensive case studies, and arranged alternately. In practice, a training combination can be seen to ultimately form ten pieces of content, with the duration of each piece controlled between three and eight minutes, and the total duration controlled within forty-five minutes. Each piece of content is given a sequence number and label description. After the arrangement is completed, a clear combination of training content is generated.

[0154] The path planning submodule generates personalized training plans based on combinations of training content.

[0155] Based on the compiled training content, the overall training path for a single student is planned and processed. During execution, the completion time records of the student's last ten training sessions are first read. For example, the longest completion time is 60 minutes and the shortest is 35 minutes. Then, combined with the distribution of the three most recent interruption locations, the content nodes prone to interruption are marked. For example, if multiple interruptions are found to be concentrated around the seventh content, high-complexity materials are prioritized and placed in the first three and last two positions, with medium-complexity materials inserted in the middle. At the same time, a range is set according to the student's acceptable total training time per session. For example, if the student's usual completion time is concentrated between 40 and 50 minutes, the overall combination time is controlled within this range. In a specific example, a student is assigned a training sequence of ten content. The first three are materials related to high-concern issues, the middle four are medium-complexity cases, and the last three are comprehensive cases. Each content is assigned independent start and end node numbers. Finally, the content order, single-item duration, total duration, and node numbers are integrated and output to form a personalized training plan that can be executed directly.

[0156] The feedback management module includes:

[0157] The Results Summary submodule generates summary data of training results based on the personalized training plan.

[0158] The training results data generated based on the personalized training plan first reads the completion data corresponding to each training stage and evaluates the performance during the training process. During the execution process, it is necessary to summarize the key indicators of each trainee's performance in the training, including but not limited to operation time, accuracy, efficiency, number of errors, etc. The specific execution steps are as follows: First, retrieve the operation data of each training stage in sequence. During the evaluation process, for example, compare the trainee's operation time with the standard time. If the operation time is longer than the standard time, record the deviation, record the deviation value, and accumulate the total deviation. Then, sum up the performance of each trainee, further organize and statistically analyze the comprehensive score of each training stage, and finally sort according to the score to form a complete summary of training results data. This summary data can reflect the trainee's overall performance in the entire training process.

[0159] The competency profiling submodule generates trainee competency profiles based on the aggregated training results data.

[0160] Based on the generated training results summary data, a competency profile is further constructed based on the performance data of each trainee. During the execution process, multiple scoring data of each trainee are analyzed, including indicators such as accuracy, speed, and deviation rate in the training results. Each item is quantitatively analyzed. For example, when analyzing a trainee's accuracy in the operation, if the trainee's accuracy is higher than 90% in five operations, then that indicator is displayed as high level in their competency profile. At the same time, the speed indicator is processed in the same way. If the trainee's training completion time is more than three times the standard time, then the speed is evaluated as low. In this way, each data item is analyzed and scored one by one, and finally a detailed trainee competency profile is generated, which reflects the trainee's performance level and development direction in multiple dimensions.

[0161] The report generation submodule generates training evaluation reports based on trainee competency profiles.

[0162] Training evaluation reports are generated based on trainee competency profiles. During the process, the trainee's strengths and weaknesses are analyzed based on the data in the profile. For example, if a trainee scores highly in precision operation but is slow in processing speed, this will be noted in the report. Customized evaluation content will be formed based on the trainee's performance. The training process also requires reviewing the trainee's learning progress and stage achievements. Combining the improvement of competency at each stage, the trainee's performance at each learning stage will be listed and evaluated in detail. Finally, conclusions will be drawn based on the overall performance to form the trainee training evaluation report.

[0163] The document management submodule generates a training closed-loop report based on the training evaluation report.

[0164] A training closed-loop report is generated based on the training evaluation report. During the execution process, the various evaluation data in the evaluation report are first summarized and compared to check the changes in the trainees' performance throughout the training cycle. For example, whether the trainees' training results have improved significantly. If the improvement is large, it is marked as "completed" in the closed-loop report. If the improvement is not obvious, it is marked as "not completed". Based on the completion status of each trainee in the training phase, a summary closed-loop report is generated. During the execution, if a trainee fails to complete the training multiple times or fails to complete the training as required, the closed-loop report will explain the reasons for non-completion and put forward suggestions for subsequent improvement. Finally, a training closed-loop report is formed that can serve as a basis for summary and improvement.

[0165] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A virtual and simulation training system for pig slaughtering technology, characterized in that, Includes the following modules: The data acquisition module, based on the pig slaughtering process and veterinary hygiene inspection standards, uses a multi-source data fusion method to collect, label, and standardize pig anatomical structure data, slaughtering operation data, and pathological characteristic data to generate raw training data. The digital twin module, based on the original training data, uses a biological structure parameterized modeling method to perform three-dimensional reconstruction and physical property mapping of pig muscles, bones, internal organs and pathological tissues, generating a pig digital twin model. The scenario construction module, based on the aforementioned pig digital twin model, adopts a process-driven scenario construction method and constructs virtual simulation inspection environments corresponding to head and hoof inspection, internal organ inspection, carcass inspection, and re-inspection according to the pig slaughter sequence and veterinary hygiene inspection post division rules, thereby generating virtual inspection scenarios; The pathology generation module, based on the virtual inspection scenario, uses a pathology feature generation method to dynamically generate pathological features of pigs and embed them into the pig digital twin model through deep learning lesion morphology deduction and lesion severity control, thereby generating a pathology simulation sample set. The interactive simulation module, based on the pathological simulation sample set, adopts an immersive interactive simulation method. Through motion capture, force feedback, and multi-channel interaction, it records the trainees' operation actions, examination sequence, and interaction status during the virtual examination process, generating trainee interactive behavior data. The behavior analysis module, based on the trainee interaction behavior data, uses a time-series behavior recognition method to analyze and evaluate the trainee's operational standardization, process completeness, and verification consistency in order to generate operational ability assessment results. The attention assessment module, based on the operational ability assessment results, uses an attention modeling method to evaluate the observation effectiveness and omission risk in the pathological identification process by matching and analyzing the trainee's eye movement trajectory with the key pathological attention areas, and generates pathological attention assessment results. The adaptive training module, based on the pathological attention assessment results, uses an adaptive training decision-making method to dynamically adjust the difficulty level, pathological type, and training focus of the training content, generating a personalized training plan. The feedback management module, based on the personalized training plan, summarizes and analyzes the training process and evaluation results of the trainees, and generates a visualized evaluation result and competency profile, as well as a training closed-loop report.

2. The virtual and simulation training system for pig slaughtering technology according to claim 1, characterized in that: The data acquisition module includes: The structural acquisition submodule collects data on the outline of the pig's body surface, the proportion of its skeleton, and key anatomical structures to generate raw structural data of the pig. The image analysis submodule, based on the original data of pig structure, performs tissue boundary analysis on the pig image data to generate tissue image analysis data; The pathological annotation submodule annotates the pathological features of pigs based on the tissue image analysis data to generate pathological feature annotation data. The standardization and organization submodule standardizes and organizes the pathological feature annotation data to generate original training data.

3. The virtual and simulation training system for pig slaughtering technology according to claim 1, characterized in that: The digital twin module includes: The morphological modeling submodule constructs a three-dimensional morphological model of pigs based on the original training data to generate a morphological model of pigs. The organization mapping submodule constructs a multi-layered organization structure based on the pig morphology model to generate an organization structure mapping model; The parameter calibration submodule calibrates the tissue physical parameters based on the tissue structure mapping model to generate a calibrated tissue model. The model validation submodule performs consistency validation based on the calibrated tissue model to generate a digital twin model of pigs.

4. The virtual and simulation training system for pig slaughtering technology according to claim 1, characterized in that: The scene construction module includes: The process mapping submodule constructs the slaughtering operation process logic based on the pig digital twin model to generate slaughtering process mapping data; The job zoning submodule constructs the spatial distribution of inspection jobs based on the slaughter process mapping data to generate an inspection job spatial model; The environment rendering submodule generates a virtual slaughtering environment based on the inspection post space model to generate a simulation environment model; The rhythm control submodule controls the work rhythm based on the simulation environment model to generate a virtual inspection scenario.

5. The virtual and simulation training system for pig slaughtering technology according to claim 1, characterized in that: The pathology generation module includes: The lesion generation submodule generates pathological morphological data based on the virtual examination scenario; The severity control submodule generates graded pathological data of different lesion severity based on the pathological morphology data; The feature embedding submodule embeds pathological features into the pig digital twin model based on the graded pathological data to generate a pathological embedding model. The sample expansion submodule generates a pathological simulation sample set based on the pathological embedding model.

6. The virtual and simulation training system for pig slaughtering technology according to claim 1, characterized in that: The interactive simulation module includes: The motion acquisition submodule collects trainees' operational actions based on the pathological simulation sample set to generate operational action data; The force feedback submodule generates interactive feedback data based on the operation action data; The interaction recording submodule records the interaction process based on the interaction feedback data to generate an interaction process record; The status synchronization submodule generates student interaction behavior data based on the interaction process records.

7. The virtual and simulation training system for pig slaughtering technology according to claim 1, characterized in that: The behavior analysis module includes: The action recognition submodule generates action recognition results based on the student interaction behavior data; The sequence verification submodule generates a process verification result based on the action recognition result; The compliance analysis submodule generates compliance analysis results based on the process verification results. The capability modeling submodule generates operational capability assessment results based on the compliance analysis results.

8. The virtual and simulation training system for pig slaughtering technology according to claim 1, characterized in that: The attention assessment module includes: The line-of-sight analysis submodule is used to generate line-of-sight trajectory data based on the operational capability assessment results. The hotspot matching submodule is used to generate attention matching results based on the gaze trajectory data; The deviation calculation submodule is used to generate attention deviation data based on the attention matching results; The risk assessment submodule is used to generate pathological attention assessment results based on the attention deviation data.

9. The virtual and simulation training system for pig slaughtering technology according to claim 1, characterized in that: The adaptive training module includes: The difficulty adjustment submodule generates training difficulty parameters based on the pathological attention assessment results; The key reinforcement submodule generates reinforcement training content based on the training difficulty parameters; The content reorganization submodule generates a combination of training content based on the enhanced training content; The path planning submodule generates personalized training schemes based on the training content.

10. The virtual and simulation training system for pig slaughtering technology according to claim 1, characterized in that: The feedback management module includes: The results summary submodule generates training result summary data based on the personalized training scheme. The competency profiling submodule generates a trainee competency profile based on the aggregated training results data. The report generation submodule generates a training assessment report based on the trainee competency profile. The document management submodule generates a training closed-loop report based on the training evaluation report.

Images