Surveying engineering teaching system and method based on digital twinning and behavior fingerprint library

Abstract

The application discloses a surveying and mapping engineering teaching system and method based on digital twinning and behavior fingerprint library. The method comprises the following steps: acquiring unmanned aerial vehicle oblique image and ground laser point cloud data, fusing GIS vector data to construct a three-dimensional digital twinning scene, triggering ICP incremental update through mobile terminal to collect environmental changes; students perform surveying and mapping operations through virtual reality interactive devices, and the system collects postures, actions and instructions and feeds back force touch; multi-dimensional behavior data flow is captured through edge nodes, and after abnormal filtering, it is stored in a time sequence database to construct a teaching behavior fingerprint library and a capacity development track; a skill capacity map is constructed to map physical parameters to capacity indicators, and the evaluation threshold is dynamically adjusted to generate a class error distribution map for semantic clustering of error events; based on a weighted linear evaluation model, capacity quantitative scores are generated and personalized retraining is pushed. The application is used for immersive teaching of surveying and mapping engineering, and realizes dynamic consistency of virtual and real and intelligent evaluation.

Description

Technical Field

[0001] This invention relates to the fields of educational informatization and surveying and mapping engineering technology, specifically to a surveying and mapping engineering teaching system and method based on digital twins and behavioral fingerprint databases. Background Technology

[0002] Surveying engineering is a highly practical discipline, and its teaching relies heavily on field training. Currently, practical teaching in surveying engineering faces the following challenges:

[0003] First, practical training resources are limited by the number of venues, equipment, and teaching cycles, resulting in a relatively limited effective practical time per student. High-precision surveying instruments (such as total stations and RTK) are expensive, and field training is greatly affected by environmental factors such as weather and electromagnetic interference.

[0004] Secondly, there are currently several types of virtual simulation teaching software on the market: the first type mainly simulates instrument operation panels, and its scenes are usually simplified geometric models without integrating high-resolution real-scene geographic information; the second type is a GIS-based 3D modeling platform, whose main functions are static scene display and vector editing, and it has not yet integrated interactive simulation of surveying and mapping operation processes and process data recording functions; the third type is a general-purpose 2D and 3D visualization platform, whose design goal is scene roaming and browsing, and it does not include task-driven training process control and operation evaluation modules.

[0005] Third, the existing related patent technologies have the following limitations: (1) CN121705961A (Experimental teaching simulation method and system based on virtual reality) adopts the technical route of "continuous behavior sequence discretization → comparison of state vector with standard template → error knowledge base query". Its data collection data is the operation sequence, and does not involve the physical state parameters of the operation (such as action speed, angle, position, duration, etc.). The evaluation granularity is relatively coarse. It uses temporary comparison templates and has no historical data accumulation. The evaluation result is a qualitative anomaly mark and no quantitative scoring system is established. The feedback is a general evaluation result and no personalized retraining push. (2) CN120183265B (Simulated mine digital twin teaching system and method based on virtual reality technology) focuses on the operation evaluation of coal mining equipment in mines. The evaluation dimension is single and no multi-dimensional behavior data collection system is established. Moreover, its digital twin scene is a static mirror and does not reflect the dynamic changes of the real environment.

[0006] Fourth, at present, the evaluation of practical teaching in most universities still relies mainly on teachers' on-site observation and records, resulting in low consistency in scoring, long feedback cycles, and difficulty in systematically identifying individual skill gaps.

[0007] Therefore, there is an urgent need for a surveying and mapping engineering practice teaching system and method that can integrate high-precision digital twin scenarios, multimodal immersive interaction, automatic collection of teaching behaviors, and intelligent evaluation feedback. Summary of the Invention

[0008] This invention aims to solve the technical problems existing in the teaching of surveying and mapping engineering, such as limited training resources, disconnect between virtual simulation scenarios and teaching evaluation, and provides a surveying and mapping engineering teaching system and method based on digital twins and behavioral fingerprint databases.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a surveying engineering teaching system and method based on digital twins and behavioral fingerprint databases, comprising the following steps:

[0010] Step S1: Acquire UAV oblique images and ground laser point cloud data of the surveying and mapping training area, integrate GIS vector data to construct a three-dimensional digital twin scene, and assign semantic labels to the surveying and mapping control points and ground objects in the scene;

[0011] Step S1A: Collect real-time change information of the field environment through the mobile terminal, and automatically trigger the incremental update process of the ICP registration algorithm to keep the three-dimensional digital twin scene and the actual surveying environment dynamically consistent.

[0012] Step S2: The trainee performs surveying operations in the digital twin scene through a virtual reality interactive device. The system collects the trainee's head posture, hand movements and operation commands in real time, and provides feedback on force and tactile information.

[0013] Step S3: Capture multi-dimensional behavioral data streams during student operations in real time through edge computing nodes, filter for anomalies, and store them in a time-series database to build a teaching behavior fingerprint database;

[0014] Step S3A: Based on the teaching behavior fingerprint database, establish a capability development trajectory sequence for each student and record the evolution of capability scores in each training session;

[0015] Step S4: Calculate the behavioral data stream based on the preset evaluation model to generate a quantitative score of the trainee's ability, and push personalized retraining content according to the score results.

[0016] Step S4A: Construct a surveying and mapping engineering skills and capabilities map, map the physical state parameters in the multi-dimensional behavioral data stream to each capability node of the skills and capabilities map, and generate an evaluation result that includes capability quantitative scores and capability shortcoming semantic annotations;

[0017] Step S4B: Based on the capability development trajectory sequence, dynamically adjust the threshold standards of each evaluation dimension in the weighted linear model to achieve progressive capability assessment;

[0018] Step S4C: Perform semantic clustering on the error event data in the teaching behavior fingerprint database, generate a class error distribution map, and push a report of common problems in the class to the teacher.

[0019] Further, in step S1, the ground resolution of the UAV tilted image is ≤2cm, and the density of the ground laser point cloud is ≥100 points / m²; the three-dimensional digital twin scene adopts a multi-resolution LOD mechanism to achieve line-of-sight adaptive loading; the measured point cloud data is aligned with the three-dimensional digital twin scene through the ICP registration algorithm.

[0020] The error function of the ICP registration algorithm is:

[0021] Where R is the rotation matrix, t is the translation vector, and w_i is the weight of the corresponding point. The optimal transformation is solved by singular value decomposition.

[0022] Furthermore, in step S2, the virtual reality interactive device includes a head-mounted display device, a spatial positioning system, and a force feedback glove; the hand movements are recognized by a CNN+LSTM hybrid neural network model, and the recognition categories include 12 preset mapping operation gestures.

[0023] The loss function of the CNN+LSTM model is cross-entropy loss:

[0024] Where N is the number of samples, C is the number of gesture categories (12 categories), yi,c is the ground truth label, and is the predicted probability. The model output is passed through a softmax layer to obtain the final action classification.

[0025] Furthermore, in step S3, the multi-dimensional behavioral data stream includes IMU attitude data, three-dimensional spatial coordinates, operation duration, and error event data, with a data dimension of no less than 20 dimensions; the anomaly filtering adopts the 3σ principle: for a certain index sequence x1,x2,...,xn, calculate the mean μ and standard deviation σ. If |xi−μ|>3σ, it is determined to be an outlier and removed to ensure the purity and reliability of the behavioral data.

[0026] Furthermore, in step S4, the preset evaluation model is a weighted linear model:

[0027] Score = 0.4 × Accuracy + 0.3 × Efficiency + 0.2 × Standardization + 0.1 × Reusability

[0028] The accuracy is measured by the mean Euclidean distance between the measurement point and the true value, the efficiency is the normalized ratio of the completion time to the standard time, the standardization statistics include the frequency of violations, and the reusability assessment includes the verification behavior of historical control points. After generating a quantitative capability score, a capability radar chart, an operation heat map, and a three-dimensional trajectory playback are also generated to identify individual capability weaknesses. Based on the identified capability weaknesses, customized training modules are matched and pushed.

[0029] Furthermore, in step S4A, the surveying engineering skill capability map includes multiple capability nodes, each capability node is bound to a quantitative evaluation rule, and the physical state parameters are converted into quantitative scores for each capability node through a pre-trained mapping model.

[0030] Furthermore, in step S4B, the dynamic threshold adjustment adopts a sliding window algorithm: based on the ability score sequence of the most recent N training sessions, the mean μ_n and standard deviation σ_n of each ability dimension are calculated, and the dynamic threshold is set to μ_n - k·σ_n (lower limit) and μ_n + k·σ_n (upper limit), where k is a configurable coefficient.

[0031] Furthermore, in step S4C, the error event data is semantically clustered according to error type, and the class error distribution map is presented in the form of a heat map or pie chart, which supports teachers to filter and view by error type.

[0032] Furthermore, in step S1A, the real-time change information of the field environment includes changes in control point status, new obstacles, and terrain changes. This information is collected via a mobile app and uploaded to the cloud to trigger incremental updates for ICP registration.

[0033] A surveying and mapping engineering teaching system based on digital twins and behavioral fingerprint databases includes:

[0034] Digital twin scene construction module: used to acquire UAV oblique images and ground laser point cloud data of the surveying and mapping training area, integrate GIS vector data to construct a three-dimensional digital twin scene, and assign semantic labels to surveying and mapping control points and ground objects in the scene;

[0035] Twin dynamic update module: Used to collect real-time change information of the field environment through mobile terminal, automatically trigger the incremental update process of ICP registration algorithm, so that the 3D digital twin scene and the actual surveying and mapping environment maintain dynamic consistency.

[0036] Multimodal immersive interaction module: includes virtual reality interaction device, which is used to enable trainees to perform surveying operations in the digital twin scene, and to collect trainees’ head posture, hand movements and operation commands in real time, while providing feedback of force and tactile information;

[0037] Teaching behavior data acquisition module: includes a data acquisition agent deployed on edge nodes, used to capture multi-dimensional behavioral data streams during student operations in real time, and store them in a time-series database after anomaly filtering to build a teaching behavior fingerprint database;

[0038] Ability Development Trajectory Modeling Module: Used to establish an ability development trajectory sequence for each student based on the teaching behavior fingerprint database, and dynamically adjust the threshold standards of the evaluation dimensions;

[0039] Group Analysis Module: Used to perform semantic clustering on error event data in the teaching behavior fingerprint database, generate class error distribution maps, and push class common problem reports to teachers;

[0040] Intelligent assessment and feedback module: used to calculate the behavioral data stream based on a preset assessment model, generate a quantitative score of the trainee's ability, and push personalized retraining content based on the score results.

[0041] The specific preferred technical parameters of each module mentioned above are consistent with the limitations in the corresponding method steps, and will not be repeated here.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] 1. By fusing drone oblique imagery with ground-based laser point clouds, a semantically tagged 3D digital twin is constructed, supporting ICP registration and incremental updates, ensuring dynamic consistency between the virtual environment and the field scene.

[0044] 2. Based on the spatial positioning system and CNN+LSTM gesture recognition model, the preset surveying operation gestures are recognized; the force feedback glove provides tactile feedback during the operation.

[0045] 3. Capture no less than 20 physical state parameters (IMU attitude data, three-dimensional spatial coordinates, operation duration, error events, etc.) in real time, and store them in a time-series database after anomaly filtering to build a traceable teaching behavior fingerprint database.

[0046] 4. Construct a surveying and mapping engineering skills and capabilities map, map the physical state parameters in the multi-dimensional behavioral data stream to each capability node of the skills and capabilities map, and generate an evaluation result that includes capability quantitative scores and semantic annotations of capability shortcomings.

[0047] 5. Based on historical data from the teaching behavior fingerprint database, establish a capability development trajectory sequence for each student and dynamically adjust the threshold standards for each assessment dimension in the assessment model.

[0048] 6. Perform semantic clustering on error event data to generate a class error distribution map and push a report of common problems in the class to the teacher's end.

[0049] 7. A weighted linear model is used to generate a quantitative score of ability, and an ability radar chart, operation heat map and 3D trajectory playback are generated to identify individual ability shortcomings. Based on the identified ability shortcomings, customized training modules are matched and pushed.

[0050] 8. Collect information on changes in the field environment via mobile devices to trigger incremental updates of ICP registration, ensuring that the virtual scene remains dynamically consistent with the actual environment. Attached Figure Description

[0051] Figure 1 This is an overall flowchart of the method of the present invention;

[0052] Figure 2 This is an interactive flowchart of the system of the present invention. Detailed Implementation

[0053] 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.

[0054] Example 1

[0055] I. Digital Twin Scenario Construction

[0056] A DJI Phantom 4 RTK drone was used to perform oblique photography of the training area at a flight altitude of 50m and a ground resolution of 2cm, acquiring 1200 images. Simultaneously, a FARO Focus S350 ground laser scanner was used to collect point cloud data at a density of 120 points / m². The oblique images and point cloud data were imported into ContextCapture software and integrated with a campus GIS vector map to construct a 3D digital twin scene. The scene includes 18 preset surveying control points, each bound with semantic tags such as real coordinates and grade identifiers. A multi-resolution LOD mechanism was employed, loading the highest precision texture when the line-of-sight is <20m, and switching to a simplified model when the line-of-sight is >100m.

[0057] When the on-site measured data (such as RTK real-time measurement) needs to update the virtual scene, the ICP registration algorithm is called: the source point cloud P is the measured point cloud, the target point cloud Q is the virtual scene point cloud, and the rotation matrix R and translation vector t are solved iteratively to make the registration error E(R,t) ≤ 0.5cm.

[0058] Dynamic Update Example: One day, a construction fence was added to the east side of the training area, obscuring control point CP-05. The instructor on-site photographed the fence location using a mobile app and uploaded the image. The system automatically recognized the environmental change, triggered an incremental update of the ICP, added a fence model to the corresponding location in the virtual scene, and updated the visibility attribute of control point CP-05 to "Occluded." When students subsequently entered the system, the virtual scene was consistent with the actual environment.

[0059] II. Multimodal Immersive Interaction

[0060] Trainees wore HTC VIVE Pro 2 headsets and Manus Quantum force feedback gloves. Four Lighthouse 2.0 positioning base stations were deployed around them for spatial tracking. The system pre-set 12 types of surveying operation gestures (such as "total station coarse aiming," "fine-tuning the auger," "RTK centering rod alignment," and "data recording"). Hand motion data was input into a pre-trained CNN+LSTM model: the CNN part used 3 convolutional layers (3×3 kernels, stride 1, ReLU activation) to extract spatial features such as hand joint angles and velocities; the LSTM part used 2 hidden layers (128 units per layer) to capture temporal dependencies. Based on the laboratory test dataset, the model output, after passing through a softmax layer, yielded the probability distribution of the 12 gesture types, achieving an accuracy of over 95%.

[0061] The force feedback glove applies different damping depending on the operation type: when simulating the leveling screw of a total station, the feedback torque is 0.5~1.0 N·m (linearly adjusted with screw resistance); when simulating the centering rod of an RTK, the offset force is 1.5~2.5 N (varying with tilt angle). Trainees complete the entire process of "control point setup - instrument setup - observation - recording" in a virtual environment.

[0062] III. Data Collection on Teaching Behaviors

[0063] Agents are deployed on edge computing nodes to capture 20-dimensional data streams in real time: head IMU (3-axis angular velocity + 3-axis acceleration), hand joints (6 degrees of freedom), 3D spatial coordinates (x, y, z), operation time (time per step), and error event types (such as "unbraked horizontal spiral" or "incorrect measurement sequence"). After data alignment with NTP timestamps, outliers are removed using the 3σ principle (e.g., if a student's operation time significantly exceeds three times the standard deviation of the mean, it is considered an anomaly and discarded). Valid data is written to the InfluxDB time-series database in JSON format, with each record containing student ID, timestamp, operation ID, and behavioral feature vector. A total of 120 hours of data are collected to construct a behavioral fingerprint database.

[0064] Example of constructing a capability development trajectory: In five training sessions, trainee Li's accuracy scores were 62, 71, 78, 83, and 85, respectively; his efficiency scores were 55, 63, 72, 78, and 83, respectively. The system records these capability development trajectory sequences and dynamically adjusts the evaluation thresholds based on a sliding window algorithm.

[0065] IV. Skill Competency Mapping and Semantic Mapping

[0066] This invention constructs a surveying and mapping engineering skills competency map, which includes the following competency nodes and their mapping rules: Centering accuracy Inclination angle of the center rod and horizontal offset distance A tilt angle < 0.5° and a deviation < 2mm earn 100 points; deduct 5 points for every 10% exceeding the limit. Instrument stability Amplitude of hand tremor (variance of acceleration) A tremor amplitude of less than 0.1g earns 100 points; deduct 5 points for every 20% exceeding the limit. Operational Standards Error event types and frequency 10 points will be deducted for each violation. Work efficiency Completion time of each step Normalized score based on ratio to standard time Environmental adaptability Operational adjustments in response to environmental changes A perfect score is awarded for a correct response; otherwise, points will be deducted.

[0067] Semantic mapping example: During the centering and leveling process of a total station, the system collected a variance of 0.28g in the amplitude of hand tremor acceleration (standard ≤ 0.1g). The mapping model output a "instrument stability" score of 72 points and generated a semantic annotation: "Your hand stability during the centering and leveling process is insufficient (tremor amplitude 0.28g, standard ≤ 0.1g). It is recommended to strengthen the training of instrument holding stability."

[0068] V. Intelligent Assessment and Feedback

[0069] The system uses a weighted linear model to score the student's performance in this exercise.

[0070] Accuracy: Calculate the average Euclidean distance between each measurement point and the true value. In this example, the average error of the 5 control points is 0.023m, and the accuracy score is 85 points (out of 100).

[0071] Efficiency: Total completion time was 12 minutes and 30 seconds, standard time was 15 minutes, normalized efficiency = 12.5 / 15 = 0.833, efficiency score was 83.3 points.

[0072] Standardization: The number of violations is counted. In this case, there was one instance of "failure to perform rough leveling first", so 10 points are deducted for standardization, and the score is 90 points.

[0073] Reusability: The trainee checked the same control point twice in the first three training sessions, and the reusability score was 80 points.

[0074] Total score = 0.4×85 + 0.3×83.3 + 0.2×90 + 0.1×80 = 85.0 points.

[0075] Example of dynamic threshold adjustment: In the first training session, student Li's "centering accuracy" passing threshold was set at 5cm (a lenient standard). Li's actual error was 4.2cm, and he scored 75 points. In the third training session, based on the previous two ability development trajectories (62→71), the system dynamically adjusted the passing threshold to 3cm. Li's actual error was 2.8cm, and he scored 85 points. In the fifth training session, the threshold was further adjusted to 2cm. Li's actual error was 1.5cm, and he scored 90 points. This achieved a "progressive" ability assessment, avoiding premature high standards that could lead to frustration.

[0076] The system automatically generates a capability radar chart (five dimensions: accuracy, efficiency, standardization, reusability, and comprehensiveness), an operation heatmap (overlaying gaze time and hand activity density in a virtual scene), and a 3D trajectory playback (marking the location of errors). It identifies the trainee's weaknesses as "low efficiency" and "omission of coarse leveling steps." The system matches a "quick instrument setup training package" and "leveling-specific exercises" from the retraining resource library and pushes them to the trainee's terminal. Simultaneously, the teacher receives class error distribution statistics: out of 32 students, 12 had "omission of coarse leveling steps," accounting for 37.5%, allowing the teacher to focus on explaining this common problem.

[0077] VI. Comparison of Results

[0078] Compared with traditional field training, in this embodiment, the average effective operation time per student has increased from 2 hours / semester to 15 hours / semester; the maintenance cost of training equipment has been significantly reduced; there are no safety risk incidents; the evaluation and feedback cycle has been shortened from 3 days to real time; and in this embodiment, the retake pass rate has increased from 68% to 92%.

[0079] Example 2: Individual Ability Deficiency Identification and Precise Retraining Based on Teaching Behavior Fingerprint Database

[0080] This embodiment is basically the same as embodiment 1. This embodiment uses the same digital twin scenario as embodiment 1. The student is Zhang, a third-year student majoring in surveying and mapping engineering, who has never used this system before.

[0081] (1) Initial Operation and Data Acquisition

[0082] Zhang was performing a "control point measurement" task. The system captured his 20-dimensional behavioral data stream through edge nodes. Key events are recorded as follows:

[0083] t=120s: During the instrument setup process, Zhang forcibly rotated the instrument without releasing the horizontal brake screw, and the system recorded an error event ERROR_001 (non-standard operation).

[0084] t=245s: When aiming at the prism, the line of sight stays for more than 8 seconds (the standard is 3 seconds), and the system records WARN_002 (inefficiency).

[0085] (2) Skills and Abilities Mapping and Scoring

[0086] The system maps Zhang's 20-dimensional physical state parameters to a skill ability graph:

[0087] Centering accuracy: Centering rod tilt angle 0.8° (standard ≤0.5°), score 78 points.

[0088] Instrument stability: Hand tremor amplitude 0.22g (standard ≤0.1g), score 68 points.

[0089] Operational compliance: 2 violations, 75 points.

[0090] Work efficiency: Total time spent: 18 minutes (standard 12 minutes), score: 67 points

[0091] Total score = 0.3×78 + 0.2×68 + 0.2×75 + 0.3×67 = 72.1 points (weights are adjusted according to the importance of the ability nodes).

[0092] The system-generated capability radar chart shows that Zhang's "instrument stability" and "operational efficiency" dimensions are significantly deficient. The operation heatmap shows that his heat density is highest (representing the longest time) in the "instrument leveling" and "target aiming" stages. The semantic annotation output is: "Your hand stability is insufficient in the leveling stage (tremor amplitude 0.22g). It is recommended to strengthen the training of instrument holding stability."

[0093] (3) Personalized refresher training recommendations

[0094] The system automatically matches and pushes resources from the resource library:

[0095] "Quick Leveling Techniques" (3 minutes)

[0096] The "Timed Leveling Challenge" training module requires trainees to complete the entire process from unpacking to leveling within 90 seconds. The system times the time in real time and provides feedback on the time taken for each step.

[0097] The virtual "coach" voice prompt will automatically pop up at the corresponding location when Zhang enters the scene again: "Please check if the horizontal brake screw is loose."

[0098] After receiving two rounds of specialized refresher training, Zhang performed the same task again, and his instrument stability score improved to 82 points, his work efficiency score improved to 78 points, and his total score reached 85.3 points, effectively making up for his shortcomings.

[0099] Example 3 (Extended Application Scenarios)

[0100] This embodiment is essentially the same as Embodiment 1, except that the digital twin scenario is replaced with an open-pit mine area for pre-job training of new employees. The resolution of the UAV oblique photography is increased to 1.5cm, the point cloud density is 150 points / m², and semantic annotations for dangerous areas such as mine slopes and transport roads are added. The gesture recognition model adds four new safety operation gestures, including "obstacle avoidance warning" and "emergency braking," for a total of 16 categories. A "safety standard" weight factor (0.15) is added to the evaluation model, and other weights are adjusted accordingly. Training results show that the violation rate of trainees during actual operation decreased by 65%.

[0101] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0102] 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 teaching method for surveying and mapping engineering based on digital twins and behavioral fingerprint databases, comprising the following steps: Step S1: Acquire UAV oblique images and ground laser point cloud data of the surveying and mapping training area, integrate GIS vector data to construct a three-dimensional digital twin scene, and assign semantic labels to the surveying and mapping control points and ground objects in the scene; Step S2: The trainee performs surveying operations in the digital twin scene through a virtual reality interactive device. The system collects the trainee's head posture, hand movements and operation commands in real time and provides feedback on force and tactile information. Step S3: Capture multi-dimensional behavioral data streams during student operations in real time through edge computing nodes, filter for anomalies, and store them in a time-series database to build a teaching behavior fingerprint database; Step S4: Calculate the behavioral data stream based on the preset evaluation model to generate a quantitative score of the trainee's ability, and push personalized retraining content according to the score results.

2. The surveying engineering teaching method based on digital twins and behavioral fingerprint database according to claim 1, characterized in that: Step S1 further includes: collecting real-time change information of the field environment through a mobile terminal, automatically triggering the incremental update process of the ICP registration algorithm, so that the three-dimensional digital twin scene and the actual surveying and mapping environment maintain dynamic consistency; Step S3 further includes: based on the teaching behavior fingerprint database, establishing a capability development trajectory sequence for each trainee, and recording the capability score evolution process of each training session.

3. The surveying engineering teaching method based on digital twins and behavioral fingerprint database according to claim 1, characterized in that: Step S4 further includes: Construct a surveying and mapping engineering skills and capabilities map, map the physical state parameters in the multi-dimensional behavioral data stream to each capability node of the skills and capabilities map, and generate an evaluation result that includes capability quantitative scores and semantic annotations of capability shortcomings; Based on the aforementioned capability development trajectory sequence, the threshold standards for each assessment dimension in the assessment model are dynamically adjusted to achieve progressive capability assessment. Semantic clustering is performed on the error event data in the teaching behavior fingerprint database to generate a class error distribution map, and a report of common class problems is pushed to the teacher's end.

4. The surveying engineering teaching method based on digital twins and behavioral fingerprint database according to claim 1, characterized in that: In step S1: the ground resolution of the UAV tilt image is ≤2cm, and the density of the ground laser point cloud is ≥100 points / m²; the three-dimensional digital twin scene adopts a multi-resolution LOD mechanism to achieve line-of-sight adaptive loading; the measured point cloud data is aligned with the three-dimensional digital twin scene through the ICP registration algorithm.

5. The surveying engineering teaching method based on digital twins and behavioral fingerprint database according to claim 1, characterized in that: In step S2: the virtual reality interaction device includes a head-mounted display device, a spatial positioning system, and a force feedback glove; the hand movements are recognized using a CNN+LSTM hybrid neural network model, and the recognition categories include 12 preset mapping operation gestures; in step S3: the multi-dimensional behavioral data stream includes IMU posture data, three-dimensional spatial coordinates, operation duration, and error event data; the anomaly filtering adopts the 3σ principle; and the time-series database is stored in a standardized JSON format.

6. The surveying engineering teaching method based on digital twins and behavioral fingerprint database according to claim 1, characterized in that: In step S4: the preset evaluation model is a weighted linear model. Score = 0.4 × Accuracy + 0.3 × Efficiency + 0.2 × Standardization + 0.1 × Reusability The accuracy is measured by the mean Euclidean distance between the measurement point and the true value, the efficiency is the normalized ratio of the completion time to the standard time, the normative statistics are the frequency of violations, and the reusability assessment is the historical control point verification behavior. After generating a quantitative ability score, it also generates an ability radar chart, operation heat map, and 3D trajectory playback to identify individual ability weaknesses and match and push customized training modules based on the identified ability weaknesses.

7. A surveying and mapping engineering teaching system based on digital twins and behavioral fingerprint databases, characterized in that: Digital twin scene construction module: used to acquire UAV oblique images and ground laser point cloud data of the surveying and mapping training area, integrate GIS vector data to construct a three-dimensional digital twin scene, and assign semantic labels to surveying and mapping control points and ground objects in the scene; Multimodal immersive interaction module: includes virtual reality interaction device, which is used to enable trainees to perform surveying operations in the digital twin scene, and to collect trainees’ head posture, hand movements and operation commands in real time, while providing feedback of force and tactile information; Teaching behavior data acquisition module: includes a data acquisition agent deployed on edge nodes, used to capture multi-dimensional behavioral data streams during student operations in real time, and store them in a time-series database after anomaly filtering to build a teaching behavior fingerprint database; Intelligent assessment and feedback module: used to calculate the behavioral data stream based on a preset assessment model, generate a quantitative score of the trainee's ability, and push personalized retraining content based on the score results.

8. The surveying and mapping engineering teaching system based on digital twin and behavioral fingerprint database according to claim 7, characterized in that: Also includes: Twin dynamic update module: Used to collect real-time change information of the field environment through mobile terminal, automatically trigger the incremental update process of ICP registration algorithm, so that the 3D digital twin scene and the actual surveying and mapping environment maintain dynamic consistency. Ability Development Trajectory Modeling Module: Used to establish an ability development trajectory sequence for each student based on the teaching behavior fingerprint database, and dynamically adjust the threshold standards of the evaluation dimensions; Group Analysis Module: Used to perform semantic clustering on error event data in the teaching behavior fingerprint database, generate class error distribution maps, and push reports of common problems in the class to the teacher's end.

9. The surveying engineering teaching system based on digital twin and behavioral fingerprint database according to claim 7, characterized in that, In the digital twin scene construction module: the UAV oblique image and the ground laser point cloud have high resolution and high density; the three-dimensional digital twin scene adopts a multi-resolution LOD mechanism to achieve line-of-sight adaptive loading, and supports real-time alignment with measured point cloud data through ICP registration algorithm; in the multimodal immersive interaction module: the virtual reality interaction device includes a head-mounted display device, a spatial positioning system, and a force feedback glove; the hand movements are recognized by a CNN+LSTM hybrid neural network model, and the recognition categories include multiple preset surveying operation gestures.

10. The surveying engineering teaching system based on digital twin and behavioral fingerprint database according to claim 7, characterized in that, In the teaching behavior data acquisition module: the multi-dimensional behavior data stream includes IMU attitude data, three-dimensional spatial coordinates, operation duration, and error event data; the anomaly filtering adopts the 3σ principle; the time-series database is stored in a standardized JSON format; in the intelligent evaluation and feedback module: the preset evaluation model is a weighted linear model, which is used to generate capability radar charts, operation heatmaps, and three-dimensional trajectory playback, identify individual capability shortcomings, and match and push customized training modules.

Images