Laboratory Anomaly Screening and Monitoring Methods and Related Equipment Based on VR Technology

Abstract

This application provides a method and related equipment for monitoring and investigating anomalies in a laboratory using VR technology, relating to the field of electronic digital data processing. It includes loading a unified virtual laboratory scene on both the teacher's and student's ends, pre-setting anomalies at test points and non-test points, along with their corresponding locations and attributes. Then, using test point anomalies as nodes, a directed graph with traversal priority is constructed based on their relationships. Combining 3D coordinates and spatial investigation order, optimal and non-optimal standard investigation sequences are generated. Students' VR gaze trajectories are collected in real time, and gaze areas exceeding a preset dwell time are filtered out and sorted to obtain the actual gaze attention sequence. The matching degree between this sequence and the standard sequence is calculated to obtain a visual inspection logic score, which is displayed in a visual form such as progress display. Implementing this method transforms the abstract investigation path into intuitive quantitative indicators, allowing teachers to obtain the student investigation process and achieve multi-dimensional evaluation of the investigation situation.

Description

Technical Field

[0001] This application relates to the field of electronic digital data processing technology, and in particular to a method and related equipment for monitoring and investigating multiple anomalies in a laboratory based on VR technology. Background Technology

[0002] Laboratory safety anomaly screening is a crucial step in ensuring the safety of faculty and students and the normal operation of experimental equipment. Traditional laboratory safety anomaly screening mainly relies on manual inspection. This method is not only inefficient and has limited coverage, but it is also prone to overlooking important anomalies due to differences in the experience or negligence of inspectors, leading to safety accidents. Especially in complex experimental environments, some potential hazards such as gas leaks, improper storage of chemicals, and electrical equipment malfunctions are often difficult to identify and handle in a timely manner.

[0003] With the development of virtual reality (VR) technology, as exemplified by patent document CN111127968A, current VR anomaly detection training systems typically employ a model where students enter a virtual laboratory. The system pre-sets multiple safety anomaly points in the virtual environment. Students are required to view and identify these pre-set anomalies by moving around and clicking on suspicious areas within a specified time. Teachers can switch to the student's first-person perspective or select a third-person perspective (i.e., a global perspective) through the VR system backend to observe the student's actions in real time. The teacher then evaluates the student's anomaly detection performance based on the number of anomalies discovered and the accuracy of their handling in the virtual laboratory.

[0004] However, traditional methods rely excessively on the quantitative statistics of the final investigation results, lacking quantitative consideration of the standardization of the operational process. In practical applications, some subjects may accidentally discover more anomalies and obtain high scores through random interaction and discrete search; while the process data value of those behaviors that strictly follow the standard investigation sequence and operate according to preset logic cannot be reflected through a single result statistical analysis. Throughout the process, teachers also randomly or based on their own judgment switch students' VR perspectives to view the data, lacking real-time behavioral data to trigger guidance, thus failing to provide targeted guidance to students in establishing correct safety anomaly investigation strategies. Summary of the Invention

[0005] This application provides a laboratory multi-person anomaly investigation and monitoring method and related equipment based on VR technology, which is used to realize the quantitative calculation of the conformity of the anomaly investigation operation sequence of the operation object and improve the effectiveness of data feedback from the monitoring end.

[0006] Firstly, this application provides a method for monitoring and investigating multiple anomalies in a laboratory based on VR technology, applied to an anomaly investigation system. This system includes at least one teacher's end and at least one student's end, connected via a network. The method includes:

[0007] A unified virtual laboratory scene is loaded on both the teacher's and student's ends. This scene includes multiple pre-defined target anomalies and their corresponding locations and attributes. These anomalies include both test-point and non-test-point anomalies. Multiple standard screening sequences are determined, each consisting of an ordered sequence of test-point locations corresponding to multiple test-point anomalies. Real-time behavioral data for each student within the virtual laboratory scene is acquired, including at least the VR gaze point movement trajectory. Multiple VR gaze areas where the VR gaze point remains for more than a pre-defined observation duration are selected from these trajectories. These VR gaze areas are then sorted chronologically to obtain the student's actual VR gaze attention sequence. For a target student, multiple matching degrees are determined based on the actual VR gaze attention sequence and each standard screening sequence. A visual inspection logic score matching the target student is then determined based on these matching degrees. This visual inspection logic score is displayed in a pre-defined visual format next to the corresponding student identification module on the teacher's interface, including a progress display matching the visual inspection logic score.

[0008] By adopting the above technical solutions, a unified virtual scene ensures synchronized information between teachers and students, a standardized screening sequence provides professional reference, real-time eye trajectory collection and dwell area screening quantify process data, and comparison between the actual attention sequence and the standard sequence generates logical scores that are then visualized. These interconnected features transform the abstract screening path into intuitive, quantifiable indicators, enabling teachers to access the student screening process, improving the objectivity and professionalism of the assessment, and achieving multi-dimensional evaluation of anomaly screening operation data.

[0009] In conjunction with some embodiments of the first aspect, in some embodiments, the step of determining multiple standard screening sequences specifically includes: setting each test point anomaly as a node, and constructing an anomaly relationship directed graph that defines the node traversal priority based on the correlation between test point anomalies, as a logical association of the test point anomalies, the association including one of causal relationship, dependency relationship or standard operating procedure relationship; obtaining the three-dimensional coordinates of each test point anomaly in the virtual laboratory scene, and using the pre-set spatial screening order as the spatial order of the test point anomalies; generating at least one optimal standard screening sequence with the logical association as the main constraint and the spatial order as the optimization objective; generating at least one non-optimal standard screening sequence that includes the test point locations corresponding to all test point anomalies, but in a different order than the optimal standard screening sequence.

[0010] By adopting the above technical solution and constructing an "anomaly relationship directed graph," the causal and dependency relationships between test points are used as the primary constraint, ensuring the correctness and rigor of the generated troubleshooting sequence in professional operating procedures. For example, the main power must be turned off before repairing individual devices. Secondly, a path planning algorithm is introduced to calculate the shortest physical path as the optimization objective, making the standard sequence not only logically correct but also the most efficient in virtual space, conforming to the efficient working habits of professionals in reality. Through the combination of primary and secondary constraints, the generated "optimal standard troubleshooting sequence" becomes a gold standard that combines logical rigor and spatial efficiency. At the same time, generating "non-optimal standard troubleshooting sequences" provides more dimensions for subsequent evaluation, enabling the system to effectively distinguish between logically consistent but spatially redundant data samples and completely discrete data samples, thereby improving the system's granularity and quantification accuracy in recognizing different characteristic operation sequences.

[0011] In some embodiments of the first aspect, the step of determining multiple matching degrees for a target student based on the actual VR gaze attention sequence and each of the standard screening sequences specifically includes: selecting all test point attention subsequences from the actual VR gaze attention sequence that correspond to the test point locations in the standard screening sequences and maintain their original temporal order; comparing the test point attention subsequences with each of the standard screening sequences in terms of order and quantity, and calculating multiple logical completeness matching scores containing both order and quantity; determining a final logical completeness matching score based on the multiple logical completeness matching scores; counting the number of test point attention belonging to test point locations and the number of non-test point attention belonging to test point locations in the actual VR gaze attention sequence; obtaining a screening focus score to characterize the student's screening efficiency by calculating the proportion of the number of test point attention in the total number of attention in the entire actual VR gaze attention sequence; and generating the final matching degree corresponding to the standard screening sequence by combining the final logical completeness matching score and the screening focus score through a preset weighted model.

[0012] By employing the aforementioned technical solution, extracting the "exam point focus subsequence" and calculating the "logical completeness matching score," the order and completeness of students' screening process in the core steps were accurately assessed, directly evaluating their professional logical abilities. Furthermore, an innovative "screening focus score" was introduced. By calculating the proportion of students' attention on valid exam points and invalid areas, the duration and efficiency of their gaze during the screening process were quantified, effectively identifying students who, although they found some exam points, had lengthy processes and high dispersion in their gaze trajectories. Finally, the final matching degree obtained through the weighted model is no longer a single-dimensional score, but a comprehensive profile of the student's overall screening performance, making the evaluation results more objective, accurate, and richer in data dimensions.

[0013] In conjunction with some embodiments of the first aspect, in some embodiments, the step of determining the final logical completeness matching score based on multiple logical completeness matching scores specifically includes: assigning a high-level coefficient to the optimal standard screening sequence and assigning a low-level coefficient to the non-optimal standard screening sequence; if it is determined that the subsequence of the test point of concern has the highest logical completeness matching score with the optimal standard sequence in terms of order and quantity, then the final logical completeness matching score is determined by combining the high-level coefficient; if it is determined that the subsequence of the test point of concern has the highest logical completeness matching score with the non-optimal standard sequence in terms of order and quantity, then the final logical completeness matching score is determined by combining the low-level coefficient.

[0014] By adopting the above technical solution, this method introduces a more refined level distinction when evaluating students' investigation logic, making the evaluation results more instructive. By assigning different level coefficients to the "optimal standard investigation sequence" and the "non-optimal standard investigation sequence," the system can clearly distinguish the level of students' professional competence. If a student's investigation sequence highly matches the "optimal sequence," which balances logic and efficiency, they will receive a higher logic score. This not only affirms their correctness but also rewards their efficient professional ability. Conversely, if their sequence only matches the "non-optimal sequence," which is logically correct but has a circuitous path, the score will be relatively low. This tiered calculation mechanism transforms discrete operation sequence data from a single binary judgment of "match" to a continuous quantitative indicator reflecting the degree of path optimization, significantly improving the system's quantitative precision and the richness of evaluation dimensions for complex investigation behavior data.

[0015] In conjunction with some embodiments of the first aspect, in some embodiments, the step of filtering out all test point attention subsequences from the actual VR gaze attention sequence that correspond to the test point locations in the standard screening sequence and maintain the original temporal order includes: real-time acquisition of student interaction behavior data with virtual items associated with test points in the test point attention subsequence, the interaction behavior data including the student's virtual physical trajectory in the virtual laboratory scene and interaction operations with test point anomalies; when the student's actual gaze attention and the virtual physical trajectory are both in the test point anomaly area at the same time, and the student performs an abnormal interaction operation, it is determined that the student's test point anomaly is a valid exploration state; a valid screening operation sequence is formed based on multiple valid exploration states; and the final test point attention subsequence is obtained based on the valid screening operation sequence.

[0016] By adopting the above technical solution, this application addresses the problem that existing technologies relying solely on gaze-based data are prone to misjudgment (i.e., unable to distinguish between random gaze scanning and substantive interactive operations at the data level). It establishes a multi-dimensional verified 'valid exploration state' by integrating virtual physical coordinate trajectories and interactive command data. This mechanism sets strict AND-gate triggering logic: the system only generates a valid exploration signal when the proximity of spatial coordinates, the coverage of the gaze area, and the generation of interactive commands (such as trigger clicks, grabs, and other control signals) are synchronized in time. This processing method can accurately filter out discrete gaze-based noise lacking interactive command support, ensuring that each node in the final generated 'exam point focus subsequence' is a valid interactive event cross-validated by multi-source data.

[0017] In conjunction with some embodiments of the first aspect, in some embodiments, after determining the visual inspection logic score matching the target student based on the matching degree, the method further includes: processing the actual VR gaze attention sequence to divide it into non-examination point attention subsequences other than the examination point attention subsequence, wherein the examination point attention region in the examination point attention subsequence represents that the student's actual VR gaze falls within any examination point abnormal region, and the non-examination point attention subsequence represents that the student's actual VR gaze falls within any non-examination point abnormal region; determining an efficiency bonus item based on the number of examination points focused on in the examination point attention subsequence and the total dwell time at the examination points; determining an efficiency deduction item based on the number of non-examination points focused on in the non-examination point attention subsequence and the total dwell time at non-examination points; and updating the visual inspection logic score based on the efficiency bonus item and the efficiency deduction item.

[0018] By adopting the above technical solution, speed and efficiency are equally crucial in real-world security screening scenarios. This solution achieves a quantitative assessment of students' time management abilities by clearly defining the time spent on "exam-focused" and "non-exam-focused" tasks, and accordingly establishing "efficiency bonus points" and "efficiency deduction points." Students who can quickly locate and handle core anomalies receive bonus points, while those who spend a lot of time on irrelevant areas receive deduction points. By differentiating screening efficiencies through bonus and deduction rules, the assessment balances logical completeness with operational timeliness, better aligning with the application needs of actual screening scenarios.

[0019] In conjunction with some embodiments of the first aspect, in some embodiments, the method further includes: when any student's VR gaze point is within the abnormal area of ​​the test site, the abnormal test site is highlighted and flashed in a first color on the teacher's interface to generate a prompt signal indicating that the abnormal test site has been covered by the gaze area; obtaining the real-time dwell time of any student on the abnormal test site and displaying the real-time dwell time next to the corresponding student identification module; calculating the total dwell time of each student on the abnormal test site based on multiple real-time dwell times; when the total dwell time exceeds a preset abnormal dwell time threshold, the abnormal test site is marked with a second color on the teacher's interface, the preset abnormal dwell time threshold matching the standard investigation time required for the corresponding target abnormality; if the abnormal test site is found by a student, the abnormal test site is marked with a third color.

[0020] By adopting the above technical solution, a multi-dimensional real-time monitoring interface is built for teachers. Different colors are used to highlight abnormalities at test sites, intuitively presenting the level of attention, investigation time, and identification results. Teachers can monitor all investigation activities in parallel, quickly identify common characteristics or special situations during the investigation process, providing data support for timely intervention and targeted guidance, and improving the efficiency of teaching monitoring and the accuracy of guidance.

[0021] In a second aspect, this application provides an anomaly detection system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the anomaly detection system to perform the method described in the first aspect and any possible implementation thereof.

[0022] Thirdly, this application provides a computer-readable storage medium including instructions that, when executed on an anomaly detection system, cause the anomaly detection system to perform the method described in the first aspect and any possible implementation thereof.

[0023] Fourthly, this application provides a computer program product that, when run on an anomaly detection system, causes the anomaly detection system to perform the method described in the first aspect and any possible implementation thereof.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] 1. By constructing a standard fault diagnosis knowledge graph and collecting multi-dimensional interaction data of trainees in the virtual environment in real time (such as operation sequence, dwell time, and gaze focus), and by quantitatively comparing and logically verifying the collected data with the knowledge graph, the technical problems of existing assessment methods relying on manual observation or simple result verification, resulting in a single assessment dimension, untraceable process, and strong subjectivity, are effectively solved. This achieves the technical effect of objective, quantitative, and multi-dimensional accurate assessment of the trainee's fault diagnosis process.

[0026] 2. By employing a technical approach that categorizes interactive objects in the virtual scene into core test points and non-test points, and cross-analyzes students' deviations from the task from two dimensions—"interaction breadth" (the proportion of attention to non-test points) and "dwelling depth" (the duration of dwell time on non-test points)—this approach effectively solves the technical problem of existing technologies that can only judge efficiency through total time but cannot accurately attribute and differentiate inefficient behaviors. This results in a precise profile of the distribution characteristics of gaze data and interaction redundancy, providing reliable data support for personalized feedback and guidance.

[0027] 3. By adopting the technical means of dividing the test points and non-test points into sub-sequences and setting efficiency bonus and deduction items to update the final logical completeness matching score, the technical problem of existing technologies not taking into account the accuracy of investigation and the timeliness of operation is effectively solved. This achieves the technical effect of guiding students to improve operational efficiency while ensuring the rigor of investigation logic, and making the assessment results more in line with the actual work needs. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of a scenario of a laboratory multi-person anomaly detection and monitoring method based on VR technology in an embodiment of this application;

[0029] Figure 2 This is a flowchart illustrating a method for investigating and monitoring multiple abnormalities in a laboratory based on VR technology, as described in this application.

[0030] Figure 3 This is another flowchart illustrating the laboratory multi-person anomaly investigation and monitoring method based on VR technology in this application embodiment;

[0031] Figure 4 This is a schematic diagram of the physical device structure of an anomaly detection system in this application embodiment. Detailed Implementation

[0032] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to and includes any or all possible combinations of one or more of the listed items.

[0033] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0034] To facilitate understanding, the application scenario used in this application will be introduced below. Please refer to... Figure 1 This is a schematic diagram of a scenario for a laboratory multi-person anomaly detection and monitoring method based on VR technology in an embodiment of this application.

[0035] like Figure 1 Figure (a) shows the task preparation and startup interface on the teacher's side, namely the lobby of the "Laboratory Multi-person Anomaly Troubleshooting VR System". This interface is the starting point for the entire training or assessment, and its core function is to organize and manage the participating students.

[0036] Multi-user access and status display: The interface features multiple (six in this example) student seats. These students can be regular students or other users undergoing anomaly detection exams. Students log in using their VR devices and join the virtual room. The teacher's interface updates the status of each seat in real time. For example, in the image, student A at seat "01" has already donned their device and confirmed readiness, displaying "Network ready"; student B at seat "02" is loading or connecting, displaying "Student B preparing..."; the remaining seats display "Student not yet joined". This design allows teachers to clearly see the readiness status of all students.

[0037] Students can enter the preparation state at any time without time restrictions, while the final start of the assessment rests entirely with the teacher. Once the teacher observes that all scheduled students are ready, they can click the "Start" button at the bottom of the interface. This action triggers a command, causing all students in the "Ready" state to start simultaneously and enter the same virtual chemistry lab scenario, ensuring the fairness and synchronicity of the assessment. It should be noted that the "same virtual chemistry lab" here could mean each student's device has an independent virtual chemistry lab that other students cannot enter, facilitating the examination of individual student anomalies. In some embodiments, it could also mean multiple students entering the same shared virtual chemistry lab space, facilitating the examination of competition among multiple students. This application primarily focuses on the first scenario in its embodiments.

[0038] Once the teacher clicks "Start," all ready students will simultaneously enter the virtual laboratory scenario to troubleshoot anomalies. At this time, the teacher's interface will automatically switch to... Figure 1 The real-time monitoring view shown in (b) provides teachers with a global, top-down monitoring platform. The core interface for the teacher is an interactive 3D model of the laboratory. Unlike the traditional first-person perspective, teachers can observe the layout of the entire laboratory and the positions of all students from a top-down view. As shown, teachers can clearly see the real-time activity trajectories of students (represented by small icons or models) within the laboratory. This 3D perspective is not fixed; teachers can freely switch viewing angles and zoom in or out using mouse dragging, scroll wheel zooming, and other operations. This allows teachers to monitor any corner of the laboratory without blind spots, whether it's an area where students gather or a secluded corner, their dynamics are fully under control.

[0039] On the left side of the interface, there is an information panel that clearly lists all participating students (e.g., "Student A", "Student B") and displays the number of anomalies they found in real time (e.g., "0 / 10"). This allows teachers to quickly understand the overall progress of each student. In some implementations, the number of anomalies here can be an anomaly investigation progress bar to realistically reflect the students' investigation status, rather than simply relying on the number of anomalies found. This is not a limitation here. The "Countdown: 07:58" in the lower left corner indicates that this assessment has a time limit, allowing teachers to monitor the remaining time and control the overall pace. Each student entry has a "View Progress" button, suggesting that teachers can not only perform macro-level monitoring but also click to view detailed investigation data or a first-person perspective for a specific student (this function is not fully shown in the image, but the presence of the button indicates this capability).

[0040] The following describes the process of the method provided in this implementation, using the scenario above as an example. Please refer to [link / reference]. Figure 2 This is a flowchart illustrating a method for monitoring and investigating multiple anomalies in a laboratory based on VR technology, as described in this application.

[0041] S201. Load a unified virtual laboratory scene on the teacher's end and the student's end. The virtual laboratory scene has multiple target anomalies and their corresponding location points and attributes preset. The target anomalies include test point anomalies and non-test point anomalies.

[0042] The teacher's end refers to the terminal device used by teachers to operate, monitor, and evaluate students' anomaly screening. This includes computers and VR hosts with corresponding VR adaptation software installed, used for issuing training tasks, viewing student data, and providing feedback and guidance. The student's end refers to the terminal device used by students to perform virtual anomaly screening operations, primarily VR headsets, paired with peripherals such as controllers, for students to immerse themselves in virtual laboratory screening training. The target anomaly refers to a pre-set set of risk data within the virtual laboratory scene that needs to be logically identified by the system; it is the core interactive object of the screening task. The specific types of target anomalies include, but are not limited to, pre-set anomalies within the scene. Static safety hazards (such as exposed wiring), dynamic safety anomalies triggered by interactions (such as chemical leaks), potential operational hazards (such as high-temperature and high-pressure equipment), and improper pre-set operational logic that characterizes non-standard interactive processes (such as incorrect reagent addition order or failure to turn on the fume hood), etc.; Exam-point anomalies refer to target anomalies included in the scope of the investigation and assessment. Their discovery and handling are directly related to students' assessment scores and are components of the standard investigation sequence. Non-exam-point anomalies refer to target anomalies not included in the core assessment scope. They can be used to interfere with or test students' judgment abilities, avoiding students from only conducting targeted investigations of exam-point anomalies, and are more in line with real and complex scenarios.

[0043] This step is the initial preparation stage for the anomaly investigation system to execute the entire monitoring method. It is executed before students start anomaly investigation training and after teachers complete the training task configuration. The scenario needs to be loaded synchronously on all student terminals and teacher monitoring terminals participating in the training to ensure the consistency of the scenario on all terminals when training starts.

[0044] First, the anomaly detection system must pre-store at least one set of virtual laboratory scene templates. These templates must be digitally recreated based on the layout and equipment configuration of a real laboratory, including but not limited to 3D modeling of core elements such as lab benches, fume hoods, chemical cabinets, electrical equipment, and piping. They must also recreate environmental details such as the laboratory's spatial dimensions, aisle distribution, and lighting conditions, and even simulate changes in scene lighting under different weather conditions or at different times to enhance immersion. The scene templates must support personalized configuration by the teacher. For example, teachers can select different types of laboratory scenes (such as chemistry labs, physics labs, biology labs, etc.), or adjust the placement of equipment within the scene, adding or removing some non-core equipment to adapt to different training needs.

[0045] Secondly, when loading scenes, the system must ensure scene synchronization between the teacher's end and all student ends. Specifically, the system server first sends the selected scene file to each terminal. After receiving the file, the terminal starts the VR rendering engine to load the scene. During the loading process, the progress needs to be reported to the server in real time. The server monitors the loading status of all terminals. When the terminal completes loading and reports a "ready" status, the system sends a "training can start" prompt to the teacher's end to remind students that they can enter the experiment for testing.

[0046] Once the scene is loaded, the system will automatically insert multiple target anomalies at preset locations and associate them with corresponding location points and attribute information. The setting of target anomalies must adhere to the principle of "realistic relevance and coverage of key points." For example, in a chemistry lab scenario, anomalies such as mixed chemical storage, broken and leaking reagent bottles, and fume hoods not being turned on can be set; in a physics lab scenario, anomalies such as short circuits, poor instrument grounding, and unstable hanging of objects at height can be set. The location point of each target anomaly is three-dimensional coordinate data (X, Y, and Z axis values), precisely corresponding to its specific location in the virtual scene, ensuring that students can accurately focus on the anomaly area when moving within the scene. Attribute information is stored in the system backend and can be triggered and displayed through student interaction. For example, when a student clicks on an anomaly with a virtual controller, a pop-up will appear displaying the anomaly type, risk level, and basic handling prompts, but it will not directly indicate whether it is an exam-related anomaly to avoid affecting the investigation and judgment.

[0047] Finally, the target anomalies are divided into test-point anomalies and non-test-point anomalies to simulate the scenario where "valid anomalies" and "distractors" coexist in actual troubleshooting. The number and distribution of test-point anomalies need to be reasonably planned, usually covering the core knowledge points of laboratory safety, and their locations should facilitate the construction of a logically coherent troubleshooting sequence. Non-test-point anomalies can be randomly distributed in the scenario, and their number can be adjusted according to the training difficulty. For example, non-test-point anomalies account for 30% in basic difficulty and 50% in advanced difficulty. Their attribute settings can be similar to those of test-point anomalies to ensure the authenticity and effectiveness of the training. After all target anomalies are set up, the virtual laboratory scenario is ready, waiting for the teacher to start the troubleshooting training instruction.

[0048] S202. Determine multiple standard screening sequences, each of which is an ordered sequence composed of multiple test point locations corresponding to test point anomalies;

[0049] The standard screening sequence refers to the ordered path preset by the system that conforms to the professional logic of laboratory safety anomaly screening, and is used to perform sequence comparison and differential analysis on the real-time operation data collected subsequently.

[0050] This step is executed after the virtual laboratory scene has been loaded and before the students begin their screening operations. It is a key step for the system to build a "reference standard" for subsequent competency assessment. The core purpose is to establish a scientific and standardized screening sequence to provide a basis for comparing the students' actual screening logic and to solve the problem of the lack of a unified logical standard in traditional assessments.

[0051] First, the system needs to complete the dual preparation of "logical association basic data" and "spatial inspection sequence rules." On the one hand, it continues the basic work of logical association construction: extracting the unique identifier, attributes (anomaly type, risk level, associated equipment), three-dimensional coordinates, and preset logical relationship labels (causality, dependency, standard process) of all test point anomalies from the backend database to ensure that the logical attributes of each test point anomaly are accurately labeled. For example, there is a dependency relationship between "fume hood malfunction" and "chemical volatilization on the lab bench" (the fume hood needs to be checked first), and there is a causal relationship between "short circuit in the distribution box" and "storage of flammable materials in the surrounding area" (short circuit is the cause). On the other hand, the system loads the pre-set spatial inspection sequence rules. These rules need to conform to the spatial operation specifications of real laboratory inspections and can be selected or customized by the teacher based on the layout type of the virtual laboratory scene (such as linear layout, island layout, layered layout). For example, a linear layout chemistry lab can be pre-planned with the spatial sequence of "left wall lab bench → middle workbench → right wall equipment area → corner storage cabinet"; a layered layout physics lab can be pre-planned with the spatial sequence of "first floor instrument area → second floor reagent area → third floor lab area". The rules should specify the three-dimensional coordinate range of each spatial area (e.g., the left wall lab bench area is X: 0-5m, Y: 0-15m, Z: 0-3m) to facilitate the subsequent association with abnormal test points.

[0052] Secondly, a directed graph of anomaly relationships is constructed to establish logical connections. This part is consistent with the previous core logic, but the clarity of "logical priority" needs to be strengthened. After mapping each test point anomaly to a node, the system draws directed edges based on logical relationship labels and assigns priority weights: causal relationships have the highest weight (1.0), meaning that cause anomalies must take precedence over result anomalies; dependency relationships have the next highest weight (0.9), meaning that prerequisite anomalies take precedence over subsequent anomalies; standard process relationships have a weight of 0.8, meaning that anomalies earlier in the process take precedence over anomalies later in the process. After drawing, a topological sort is performed to ensure that the directed graph has no circular dependencies (if a cycle of "device A depends on device B, device B depends on device A" is found, the system will automatically pop up a window to prompt the teacher to correct the logical relationship), ultimately forming a directed graph of anomaly relationships with clear node traversal priorities. For example, in a biology laboratory, there are three abnormalities: A (UV lamp on the clean bench is not turned on), B (petition dish contaminated), and C (sterilized gown). A and B have a dependency relationship (A→B, weight 0.9), and C and A have a standard procedure relationship (C→A, weight 0.8). In this case, the priority of traversing the directed graph is C > A > B.

[0053] Furthermore, the spatial order is determined by combining the three-dimensional coordinates with the preset spatial inspection order. The system first matches the three-dimensional coordinates of each test point anomaly with the "spatial region" in the preset spatial inspection order rules to determine the spatial region to which each test point anomaly belongs (e.g., the three-dimensional coordinates of test point anomaly A are X: 2m, Y: 3m, Z: 1m, belonging to the "left wall experimental table region"). Then, according to the region order of the preset spatial inspection order, test point anomalies in the same region are sorted according to the "proximity principle" (the closest three-dimensional coordinate distance), finally forming the overall spatial order. For example, the preset spatial order is "left wall experimental table → middle operating table → right wall equipment area". The left wall experimental table region contains test points A and D, the middle operating table contains test points B and E, and the right wall equipment area contains test points C and F. The three-dimensional coordinate distance between A and D in the left wall region is the closest, so the order is A→D. The middle region is B→E, and the right wall region is C→F. Therefore, the overall spatial order is A→D→B→E→C→F. During this process, the system will verify the rationality of the spatial order. If the three-dimensional coordinates of the test points in a certain spatial area are abnormally scattered, the system will prompt the teacher to adjust the area division to ensure that the spatial order conforms to the actual inspection habit of "dividing areas and not jumping".

[0054] Then, the optimal standard screening sequence is generated. It strictly adheres to the principle of "logical association as the primary constraint and spatial order as the optimization objective": first, the "hard order" of the test points is determined based on the directed graph of abnormal relationships (e.g., C must precede A, and A must precede B); then, without violating this constraint, the positions of the test points are adjusted to fit the spatial order. For example, if the logical requirement is C→A→B, and the spatial order is A→D→B→C→E, the optimal sequence after fusion is C→A→D→B→E, which satisfies all logical constraints and conforms to the spatial habits of regional screening.

[0055] Finally, non-optimal standard screening sequences are generated. The core method is "random concatenation of all test point locations + exclusion of optimal sequences": First, the system extracts the unique identifiers of all test point locations and uses a random sorting algorithm to randomly concatenate them, generating a large number of random sequences containing all test point locations. For example, 10 test points can generate dozens to hundreds of random sequences, ensuring that the randomness of the sequences covers the possibility of different screening orders. Second, sequences that are completely identical to the optimal standard screening sequence are removed from all random sequences, and the remaining sequences are non-optimal standard screening sequences. Third, the system filters non-optimal sequences according to a preset number (usually 2-5). During the filtering, only cases that are completely identical to the optimal sequence are excluded, without setting other additional constraints, preserving the pure randomness of the sequences. For example, if the optimal sequence is C→A→D→B→E, the filtered non-optimal sequences could be A→C→B→D→E, D→B→A→C→E, etc., all of which have different orders from the optimal sequence and no fixed logic. These non-optimal standard screening sequences are configured as reference templates for identifying discrete or unordered operation instructions. The system compares the features of the real-time collected sequence to be detected with the non-optimal standard screening sequence. By calculating the difference in matching degree between the sequence and the optimal and non-optimal sequences, it realizes automatic classification and quantification of ordered signal sets that follow preset topological constraints and discrete signal sets that exhibit random distribution characteristics.

[0056] In some embodiments, the optimal standard screening sequence can be set by the teacher on the teacher's end based on their own experience standards. There can be multiple sequences or one sequence, and no limitation is made here.

[0057] S203. Acquire real-time behavioral data of each student in the virtual laboratory scene, including at least the VR gaze point movement trajectory;

[0058] This step is performed after the student starts the virtual lab investigation training and continues until the student completes the training and submits the results or the training time ends.

[0059] First, the system needs to complete the pre-configuration for data collection, including determining the collection frequency, defining the data collection scope, and establishing a data transmission channel. The collection frequency needs to be set comprehensively based on data accuracy requirements and device performance. Too high a frequency (e.g., 50 times / second) may cause device lag, while too low a frequency (e.g., 1 time / second) may miss key behaviors. Typically, a setting of 5-20 times / second is used to ensure that detailed behaviors such as eye movement and operation triggers are captured without putting excessive performance pressure on the student's device. In addition to the core VR eye movement trajectory, the data collection scope can also include virtual operation behavior data (such as operation records of clicking, dragging, and triggering abnormal interactions with virtual controllers), virtual physical movement trajectory (changes in the student's position coordinates in the virtual scene), and device status data (wearing status of the VR headset, operation status of the controllers), etc. Teachers can choose whether to collect additional data according to assessment needs. By default, only VR eye movement trajectory and core operation data are collected.

[0060] The eye-tracking module built into the VR headset is the hardware foundation for collecting gaze data. This module captures information such as the student's eye rotation angle and pupil position, and combines this with the VR headset's spatial positioning data to calculate the three-dimensional coordinates (X, Y, and Z axis values) of the gaze focus in the virtual scene. The system continuously collects this coordinate data at a preset frequency, obtaining a gaze point with each collection. Consecutive gaze points are sorted by collection timestamps to form the VR gaze point movement trajectory. In addition to the three-dimensional coordinates, each gaze point is associated with corresponding metadata such as a timestamp (accurate to milliseconds), student identification, and scene identification, facilitating accurate positioning and filtering during subsequent data processing. For example, the student's gaze focus coordinates at 0.1 seconds are (10.2, 5.3, 3.1), and at 0.2 seconds, they are (10.5, 5.4, 3.2). These consecutive coordinate points together constitute the student's gaze movement trajectory within that time period, reflecting the process of shifting observation from one scene area to another.

[0061] Simultaneously, the system will collect other auxiliary behavioral data (if configured). Virtual operation behavior data is generated by capturing information such as button triggers and posture changes of the virtual controller. For example, when a student clicks on an abnormal area with the controller, the system will record the click time, the corresponding abnormality indicator, and the number of clicks. Virtual physical movement trajectory is collected through the spatial positioning module of the VR headset, recording the changes in the student's position coordinates in the virtual scene and reflecting their movement path in the scene. Device status data is used to determine the student's training status, such as whether the VR headset has fallen off or whether the controller has been disconnected. If a device abnormality occurs, the system will pause data collection and send a prompt to the teacher to ensure the effectiveness of data collection.

[0062] S204. Select multiple VR viewing areas from the VR viewing point movement trajectory where the VR viewing point stays for more than a preset observation duration;

[0063] The anomaly detection system collects real-time coordinate data of students' gaze points using VR devices on the student's end. This coordinate data is arranged chronologically to form VR gaze point movement trajectories. The system then timestamps each gaze point in the trajectory and determines whether the dwell time of each gaze point and its adjacent consecutive gaze points within the same spatial range exceeds a preset observation dwell time. If the dwell time of consecutive gaze points within a certain spatial range exceeds this threshold, that spatial range is defined as a VR gaze area. This step effectively filters out instantaneous noise coordinates (i.e., interference areas) caused by rapid displacement from the continuous gaze trajectory data stream, thereby extracting stable dwell area data that meets the preset time span requirements. This step eliminates data redundancy caused by discrete random trajectories during data preprocessing, preventing invalid 3D spatial coordinates from being incorrectly mapped to sequence nodes, thus significantly improving the data signal-to-noise ratio and the confidence of the calculation results in subsequent sequence matching operations.

[0064] S205. Sort the multiple VR gaze areas in chronological order to obtain the student's actual VR gaze attention sequence;

[0065] The actual VR gaze attention sequence refers to an ordered set formed by arranging multiple VR gaze areas in the order in which students focused on them, reflecting the students' gaze attention path during anomaly investigation. For example, if a student first focuses on the lab bench area (timestamp 10:00:02), then the chemical storage area (timestamp 10:00:05), and finally the electrical equipment area (timestamp 10:00:08), these three VR gaze areas, ordered chronologically, constitute the actual VR gaze attention sequence.

[0066] The system retrieves relevant data for each VR gaze region obtained in step S204 from the database, including the start timestamp of each VR gaze region's attention. Then, based on these timestamps, the system sorts all VR gaze regions in ascending order, arranging them according to the order in which the student focuses on a particular region first. After sorting, these ordered VR gaze regions are combined to form the student's actual VR gaze attention sequence and stored in the corresponding data module of the system. This step transforms the student's scattered attention areas into an ordered gaze attention path, fully presenting the student's gaze movement logic during anomaly detection. This provides an intuitive basis for subsequent comparison with standard detection sequences and solves the problem that traditional methods cannot capture the student's detection process.

[0067] S206. For the target student, determine multiple matching degrees based on the actual VR gaze attention sequence and each of the standard screening sequences, and determine the visual inspection logic score that matches the target student based on the matching degree.

[0068] This step is executed after obtaining the student's actual VR gaze sequence, when the teacher needs to evaluate a student's screening logic. The scenario is the stage where the teacher monitors and evaluates the student screening process on the teacher's end.

[0069] After completing the detailed matching degree calculations (see S301-S306) to obtain multiple matching degrees between students and various standard screening sequences, the system needs to convert these technical indicators into scoring results that are easy for teachers to understand. The system first selects the highest value from the multiple matching degrees as the student's best matching degree, indicating which standard screening sequence the student's screening behavior is closest to. Then, the system converts the matching degree into a visual inspection logic score using a preset score mapping function, typically on a percentage or ten-point scale. For example, the mapping function can be set as: Visual Inspection Logic Score = Best Matching Degree × 100 (percentage) or Best Matching Degree × 10 (ten-point scale). For a student with a matching degree of 0.85, their visual inspection logic score is 85 points (percentage) or 8.5 points (ten-point scale).

[0070] The system may also set more complex mapping rules according to teaching needs, such as introducing score range divisions: a matching degree of 0.9 or above corresponds to excellent (90-100 points), a matching degree of 0.8-0.9 corresponds to good (80-89 points), and so on. In addition, the system may combine other student performance indicators, such as the time to complete the investigation and the number of anomalies found, to fine-tune the visual inspection logic score in order to comprehensively reflect the student's investigation performance.

[0071] In some embodiments, after obtaining the student's actual VR gaze attention sequence, the efficiency of the student's attention during the screening process can be further evaluated, and the efficiency factor can be incorporated into the final score. The system first loads a preset map of abnormal test site areas, containing the location information of all abnormal test sites and their surrounding effective identification range. Then, the system iterates through each attention point in the student's actual VR gaze attention sequence and its corresponding dwell time, determining whether the point falls within any abnormal test site area by comparing spatial coordinates. If the attention point's coordinates are within the boundary range of the abnormal test site area, it is classified as a test site attention point and added to the test site attention subsequence; otherwise, it is classified as a non-test site attention point and added to the non-test site attention subsequence. The system accurately records the location coordinates, timestamp, and dwell time information of each attention point in both types of subsequences.

[0072] After completing the sequence segmentation, the system separately analyzes the feature data of the test point focus subsequence and the non-test point focus subsequence. For the test point focus subsequence, the system counts the number of different test points (excluding those with duplicate focus) and the total time students spend at all test points. The system calculates the test point coverage rate (number of focus points / total number of test points) and the average dwell time (total dwell time / number of focus points), comparing them with the preset standard observation time. Based on this data, the system calculates the efficiency bonus using a preset formula: Efficiency Bonus = α × Test Point Coverage Rate + β × min (average dwell time ratio, maximum effective dwell time ratio), where α and β are weighting coefficients, and the maximum effective dwell time ratio is capped to prevent excessively long dwell times from resulting in unreasonably high scores.

[0073] Simultaneously, the system counts the number of different non-exam points in the non-exam point attention subsequence and the total time students spend at these non-exam points. The system may classify non-exam points according to their nature, such as "completely irrelevant areas" and "low-relevance areas," assigning different weights. The system calculates the efficiency deduction item using a preset formula: Efficiency Deduction Item = γ × (Number of Non-Exam Points / Total Number of Attention Points) + δ × (Total Time Spent at Non-Exam Points / Total Time Spent at All Points), where γ and δ are weighting coefficients that can be adjusted according to actual conditions, requirements, and task difficulty.

[0074] Finally, the system adjusts the visual inspection logic score previously calculated through step S206 based on efficiency bonuses and deductions. A typical update formula is: Updated visual inspection logic score = Original visual inspection logic score × (1 + Efficiency bonuses - Efficiency deductions), and may set an adjustment cap (e.g., ±20%) to ensure that efficiency factors do not excessively affect the basic score. The updated score will replace the original score and be displayed on the teacher's interface, providing teachers with more comprehensive information on student screening.

[0075] This step allows for a comprehensive evaluation of the logic and efficiency of the screening process, effectively distinguishing between ordered professional screening and random clicking or unordered searching. It also addresses the problem that traditional evaluation methods cannot reflect the efficiency of students' gaze duration allocation. By quantitatively analyzing the distribution of students' gaze at examination sites and non-examination sites, the system can identify operational behaviors that conform to standard screening logic and have high timeliness, providing a scientific basis for teachers to provide targeted guidance, thereby improving the effectiveness of laboratory safety education and students' safety awareness.

[0076] S207. Display the visual inspection logic score in a preset visual format next to the corresponding student identification module in the teacher's interface. The preset visual format includes a progress display that matches the visual inspection logic score.

[0077] The student identification module refers to the area on the teacher's interface used to identify each student, typically including the student's name, student ID, and photo, making it easier for teachers to distinguish between different students. Progress display is a preset visual format that uses the fill level of a progress bar to correspond to the visual inspection logic score. For example, if the maximum score is 100 and the score is 80, the progress bar will be 80% full.

[0078] The system determines the location of the target student's corresponding student identification module on the teacher's interface. Then, according to preset visualization rules, it converts the visual inspection logic score into a corresponding progress display format; for example, when the score is 60, a progress bar filled to 60% is generated. Finally, the system displays this progress display along with the corresponding visual inspection logic score next to the target student's identification module, ensuring that teachers can quickly and intuitively see each student's visual inspection logic score. This step allows teachers to grasp the investigation logic status of each student in real time and clearly, without manually analyzing complex data, greatly improving the efficiency of teacher monitoring and evaluation, and also facilitating targeted guidance to students.

[0079] In the above embodiment, by using a unified virtual scene to preset anomalies and multiple sets of standard investigation sequences, the system tracks students' VR gaze trajectory in real time and extracts an orderly actual attention sequence. Then, through sequence matching degree calculation, a visual inspection logic score reflecting the investigation logic is obtained and displayed in a visual form on the teacher's end. Therefore, it is possible to identify the degree of conformity between students' investigation thinking logic and the standard investigation sequence, which solves the problem of traditional methods that only evaluate based on results and are difficult to quantify the investigation process. This enables a quantitative evaluation of students' abnormal investigation behavior process and provides data support for teaching guidance.

[0080] Following the above embodiments, the method provided in this embodiment will now be described in more detail. Please refer to [link / reference]. Figure 3 This is another flowchart illustrating the laboratory multi-person anomaly screening and monitoring method based on VR technology in this application embodiment.

[0081] S301. From the actual VR gaze attention sequence, select all the test point attention subsequences that correspond to the test point locations in the standard screening sequence and maintain the original time sequence.

[0082] The test point focus subsequence refers to the subsequence selected from the actual VR visual focus sequence that corresponds to the test point location in the standard screening sequence and maintains the original chronological order. For example, if the actual VR visual focus sequence is [A (non-test point), B (test point 1), C (non-test point), D (test point 2), E (test point 3)], and the test point locations in the standard screening sequence are [test point 1, test point 2, test point 3], then the test point focus subsequence is [B, D, E].

[0083] When students enter the virtual laboratory scene to investigate anomalies, the system acquires their VR gaze movement trajectory in real time and filters out VR gaze areas that linger for longer than a preset observation time, thus forming an actual VR gaze attention sequence. To accurately determine the match between the student's investigation logic and the standard investigation logic, a subsequence of test point attention related to the standard investigation sequence needs to be extracted from this sequence. The system first iterates through each VR gaze area in the actual VR gaze attention sequence, determining whether the area corresponds to a test point location in the standard investigation sequence; if a correspondence is found, it is retained. Simultaneously, the chronological order of these retained VR gaze areas in the actual VR gaze attention sequence must be strictly maintained, without reordering, ultimately forming the test point attention subsequence. The effect of this step is to filter out the core parts of the actual VR gaze attention sequence that are related to the standard investigation logic, eliminating irrelevant non-test point attention areas as interference, laying the foundation for subsequent calculation of the complete logical matching score, and avoiding the impact of too many non-test point attention areas on the accurate judgment of the student's investigation logic.

[0084] In some embodiments, the effectiveness of the initial test point focus subsequence can be further verified by considering whether students have performed screening operations.

[0085] Specifically, during the student screening process, the system first acquires real-time data on the interaction behavior of students with virtual items associated with test points in the test point focus subsequence. This data mainly includes two aspects: first, the virtual physical trajectory formed by the student controlling the virtual character to move in the virtual laboratory scene, with the system recording the coordinate changes of the virtual character in real time using positioning technology; second, the interactive operations performed by the student on test point anomalies, such as clicking, dragging, and using virtual tools. Next, the system analyzes and judges the acquired data. When it detects that the student's actual VR gaze focus area and virtual physical trajectory are both stationary in a certain test point anomaly area at the same time, and the student performs a preset abnormal interactive operation in that area, it can be determined that the student is in a valid exploration state for that test point anomaly. This step is to exclude the situation where the student's gaze accidentally sweeps across the test point area without actually exploring it. Then, the multiple valid exploration states formed by the student throughout the screening process are arranged in chronological order to form a valid screening operation sequence. Finally, based on the valid screening operation sequence, the test point focus area corresponding to each valid exploration state is extracted and organized in chronological order to obtain the final test point focus subsequence. The effect of this step is to verify the authenticity and effectiveness of the test point focus subsequence through multi-dimensional behavioral data, avoid misjudging a student's occasional gaze as a valid screening, solve the problem of misjudgment that may occur when screening test point focus subsequences based solely on gaze trajectory, and provide a more accurate and reliable data foundation for subsequent calculation of logical matching scores, making the evaluation of student screening results more objective and accurate.

[0086] S302. Compare the subsequence of the test point with each standard screening sequence in terms of order and quantity, and calculate multiple logical complete matching scores that include both order and quantity.

[0087] Among them, the sequential comparison refers to comparing the consistency between the order of the test points in the test point focus subsequence and the order of the test points in the standard screening sequence; the quantitative comparison refers to comparing the difference between the number of test point positions contained in the test point focus subsequence and the number of test point positions in the standard screening sequence.

[0088] This step is performed after obtaining the student's test point focus subsequence. Its purpose is to evaluate the degree of consistency between the student's screening logic and the standard path. The system comprehensively compares the test point focus subsequence with each standard screening sequence, evaluating it from two dimensions: order and quantity. For order comparison, the system uses the Longest Common Subsequence (LCS) algorithm to calculate the longest common part between the test point focus subsequence and the standard screening sequence, obtaining the order matching degree. For example, if the standard screening sequence is [A, B, C, D, E], and the student's test point focus subsequence is [A, C, D, F], then the longest common subsequence is [A, C, D], and the order matching rate is 3 / 5 = 0.6 or 60%. For quantity comparison, the system calculates the percentage of test points discovered by the student relative to the total number of test points in the standard screening sequence, obtaining the quantity matching degree. For the above example, the quantity matching rate is 4 / 5 = 0.8 or 80% (considering F is an additional discovered test point). The system then calculates a comprehensive logical completeness matching score using a pre-set weighted formula (e.g., Logical Completeness Matching Score = 0.7 × Sequential Matching Rate + 0.3 × Quantitative Matching Rate). Since the system may have pre-set multiple standard screening sequences (e.g., based on different strategies such as logical association priority or spatial order priority), multiple logical completeness matching scores will be obtained. This step objectively quantifies the degree to which students' screening behavior conforms to professional standards, solving the problem of traditional methods focusing only on results and neglecting the process, and providing a scientific basis for teachers to evaluate students' professional screening performance.

[0089] S303. Determine the final logical completeness matching score based on multiple logical completeness matching scores;

[0090] This step is performed after calculating the logical completeness matching score between the student's focus subsequence and each standard screening sequence, aiming to determine a final score that comprehensively reflects the student's screening logic process. The system first compares multiple logical completeness matching scores, identifying the highest matching score and its corresponding standard screening sequence type. If the student's focus subsequence has the highest matching score with the optimal standard screening sequence (such as a logical association priority sequence based on anomaly relation directed graphs), the system applies a preset high-level coefficient (such as 1.2) to multiply the matching score; if the student's screening logic has the highest matching degree with a non-optimal standard sequence (such as a shortest path sequence based on spatial order), a lower-level coefficient (such as 0.9) is applied to multiply the matching score.

[0091] For example, if a student scores 0.75 for matching the optimal standard screening sequence and 0.85 for matching a non-optimal sequence, the system will calculate 0.85 × 0.9 = 0.765 < 0.75 × 1.2 = 0.9, therefore the final logical completeness matching score is 0.9. In some cases, the system may also consider the comprehensive matching of the student's screening logic with multiple standard sequences, determining the final score through a weighted average to comprehensively evaluate the student's screening logic. This step allows for comparison of the student's screening behavior with multiple professional standards and assigns differentiated evaluations based on the importance of different standards, solving the problem that traditional evaluation methods cannot quantify and distinguish different screening logics, and providing a data foundation for evaluating the student's screening behavior.

[0092] S304. Count the number of test point attention items belonging to the test point location and the number of non-test point attention items not belonging to the test point location in the actual VR gaze attention sequence;

[0093] Among them, "Number of Examination Points Focused On" represents the number of times a student's gaze lingers on an examination point location in the actual VR gaze sequence, characterizing the distribution density of VR gaze data at key anomaly coordinate points. "Non-Examination Point Locations" is used to represent areas in the virtual laboratory that do not contain examination point anomalies, which may be normal equipment or insignificant areas. "Number of Non-Examination Points Focused On" refers to the number of times a student's gaze lingers on a non-examination point location in the actual VR gaze sequence, characterizing the discrete distribution of VR gaze data in non-critical areas.

[0094] This step is performed after obtaining the students' actual VR gaze sequence, aiming to quantify the students' attention allocation during the screening process. The system first divides all areas in the virtual laboratory scene into two categories: test site locations and non-test site locations. Test site locations refer to areas containing pre-set test site anomalies, while non-test site locations refer to other areas that do not contain test site anomalies.

[0095] The system iterates through each attention point in the student's actual VR gaze sequence, determining whether the point falls on a test site location or a non-test site location, and counts accordingly. For gaze points falling on test site locations, the system increments the test site attention count by 1; for gaze points falling on non-test site locations, the system increments the non-test site attention count by 1. For example, if a student's actual VR gaze sequence contains 20 attention points, with 12 at test site locations and 8 at non-test site locations, the test site attention count is 12, and the non-test site attention count is 8. The system may also record the number of times each test site is attended, as well as the specific duration the student spends at each location, providing more granular data for subsequent analysis. This step reflects the distribution of gaze dwell time during anomaly detection, solving the problem that traditional methods cannot quantify the distribution of gaze dwell time during detection, and providing data support for evaluating students' detection behavior.

[0096] S305. By calculating the proportion of the number of attention points for this test point in the total number of attention points in the entire actual VR gaze attention sequence, a screening focus score is obtained to characterize the student screening efficiency.

[0097] This step is performed after the number of test-point and non-test-point focuses are tallied, aiming to assess the efficiency and focus of students during the screening process. The system first calculates the total number of times students actually focus on a particular VR viewing sequence, which is the sum of the number of test-point and non-test-point focuses. Then, the system divides the number of test-point focuses by the total number of focuses to obtain the test-point focus percentage, i.e., the screening focus score. The formula for calculating the screening focus score is: Screening Focus Score = Number of Test-point Focuses ÷ Total Number of Focuses × 100%. For example, if a student focuses on 15 test-points and 25 non-test-points, the total number of focuses is 40, and the screening focus score is 15 ÷ 40 × 100% = 37.5%. In practical applications, the system may adjust the formula, such as setting a base score and adding or subtracting points based on efficiency, or considering the importance of each test-point and setting weights. Furthermore, the system may also combine the proportion of test-point focus time to total observation time to more comprehensively evaluate screening efficiency. The implementation of this step can quantitatively assess the percentage of students' effective line of sight during screening, effectively solving the problem that traditional assessment methods cannot distinguish between effective screening and random clicks, and providing teachers with an objective basis for understanding students' screening strategies.

[0098] The focus score in this step is used to characterize the density of effective gaze data in the actual VR gaze attention sequence. This score does not involve the time dimension, but rather measures the accuracy of spatial overlap between the gaze point and the preset target anomaly area. A higher focus score means that most visual fixations fall within the preset "exam point anomaly" range in spatial coordinates, indicating that the check operation has high target orientation and low data redundancy; a lower focus score means that the gaze trajectory is discretely distributed in a large number of "non-exam point" coordinate areas, indicating that the operation contains more invalid interaction data or background noise, and has a low correlation with the preset target.

[0099] S306. Combining the final logical complete matching score and the investigation focus score, the final matching degree corresponding to the standard investigation sequence is generated by calculating using a preset weighted model.

[0100] The system pre-sets a weighted model that determines the weights of the final logical integrity matching score and the investigation focus score based on the key needs of laboratory anomaly investigation. Generally, the weight of the final logical integrity matching score is higher than that of the investigation focus score, for example, weights are set to 0.7 and 0.3 respectively. Then, the final logical integrity matching score is multiplied by its corresponding weight, and the investigation focus score is multiplied by its corresponding weight. The two products are then added together, and the resulting value is the final matching degree with the standard investigation sequence. If multiple standard investigation sequences exist, the corresponding matching degree is calculated separately for each. This step integrates the completeness of the student's investigation logic and the focus of their attention during the investigation process, generating a comprehensive matching degree index that provides more assessment information compared to single-dimensional evaluation. Through this matching degree, teachers can understand the student's anomaly investigation performance, including both the degree of conformity between their investigation logic and the standard sequence, and the distribution of their focus points during the investigation process, thus providing data support for teaching guidance.

[0101] In this embodiment, by selecting the test point focus subsequence corresponding to the standard screening sequence and eliminating non-test point interference, the matching degree between the screening logic and the standard sequence is accurately quantified and the weight influence of the optimal screening sequence is increased, and a comprehensive matching degree is generated. This solves the problem that traditional methods are difficult to quantify and reflect the standardization of students' screening process, thereby achieving an objective and accurate quantitative evaluation of students' abnormal screening situation and providing teachers with a practical and targeted guidance basis.

[0102] In some embodiments, in traditional monitoring modes, if a teacher wants to understand the specific investigation status of a student, they often need to manually click on the student's VR viewpoint to obtain information. However, this method has the following limitations: after clicking, only the student's current status can be viewed. To grasp more dynamics, the teacher must remain in that viewpoint continuously, and cannot simultaneously view other students. In addition, even if the teacher continues to view, they can only observe the student's current anomaly investigation action (such as whether it is operating in a certain area), and it is difficult to trace back or synchronously understand the student's investigation process for other anomalies (such as whether they have paid attention to a certain anomaly before, and for how long). Traditional methods usually only evaluate based on "the number of anomalies ultimately investigated by the student," and cannot know the student's interaction process for each anomaly (such as the duration of gaze, the order of operations, etc.), nor can they quantify the degree to which the student's investigation logic conforms to the standard sequence, resulting in a single evaluation dimension.

[0103] To address this issue and enable teachers to gain a comprehensive and accurate understanding of each student's entire screening process, the following steps can be implemented. Teachers no longer need to remain focused on a single student's perspective. Through color-coded markers and time data on the teacher's interface, they can clearly grasp the student's complete interaction process with each anomaly. This allows for a deeper understanding of the student's screening logic for each anomaly, moving away from the traditional, one-sided evaluation model that relies solely on the number of results. This enables a refined and comprehensive control over the student screening process.

[0104] Specifically, throughout the entire process of students entering the virtual laboratory scene to conduct anomaly screening, the system needs to complete multi-dimensional real-time monitoring and feedback: First, it tracks the VR gaze coordinates of all students in real time. When it detects that any student's VR gaze falls into an abnormal area of ​​a test point, it immediately triggers the visual prompt mechanism on the teacher's interface—highlighting and flashing the abnormal test point with the first color (such as yellow), allowing the teacher to instantly know that the test point has been monitored. At the same time, the system starts the timing module to record the student's real-time dwell time at the test point, and displays a dynamically updated time value (such as "Current Dwell Time: 3 seconds") next to the corresponding student's identity identifier module on the teacher's end. When the student's gaze leaves the test point area, the timing pauses; if the student re-enters, the timing restarts and accumulates, finally summing up the student's total dwell time at the test point. Subsequently, the system compares the total dwell time with a preset abnormal dwell time threshold (set according to the test point standard screening time): if the total dwell time exceeds the threshold, it indicates that the student may have difficulty screening at the test point, and the teacher's interface immediately switches the test point's marker color to the second color (such as orange). When the system detects that a student has completed a valid confirmation operation for the test site abnormality (such as submitting the correct abnormality identification result), it determines that the abnormality has been found and then switches the test site marker color to the third color (such as green).

[0105] The technical effect of the above embodiments is to realize the visualization and dynamic monitoring of the student screening process. Through color marking and time data, teachers can quickly grasp the attention status of each test point, the time spent on student screening and the results of abnormal identification. This solves the problem of "difficulty in comprehensive tracking and inability to intervene in a timely manner" in traditional manual monitoring, and provides teachers with accurate guidance for intervention.

[0106] In some embodiments, when teachers need to understand the overall interaction of a student group (such as a class or a group) with anomalies at various test points, traditional methods often only involve counting the number of anomalies that each student ultimately identifies, and then relying on the teacher's experience to judge the group's response to each anomaly. This approach not only relies on the teacher's subjective experience but also only reflects the result of "whether anomalies have been found," failing to reflect the statistical distribution of the group's interaction characteristics with anomalies, making it difficult for teachers to accurately pinpoint common difficulties in the class.

[0107] To address this need, an objective and visual assessment of the students' completion status can be achieved by performing the following steps: Based on multiple real-time dwell times, calculate the total dwell time for each student regarding the test site anomaly. Then, determine the overall completion status of all students for the test site anomaly based on the total dwell time and the standard total time required for the corresponding test site anomaly. Set the test site anomaly as an initial marker on the teacher's interface. Differentiate the test site anomaly according to the students' overall completion status. Specific differentiation methods include: when the test site anomaly is marked with the initial marker, it indicates that students have not yet effectively interacted with the test site anomaly; when the test site anomaly is marked with a second color, it indicates that the students' investigation and interaction process for the test site anomaly is lagging; when the test site anomaly is marked with a third color, it indicates that students have completed the standard identification operation for the test site anomaly; display the differentiation markers for each test site anomaly next to the corresponding test site anomaly marker on the teacher's interface.

[0108] More specifically, the system first collects the total time all students spend at a particular test site anomaly. Then, it compares this total time with the standard total time corresponding to that test site anomaly, calculating relevant statistical indicators (such as the ratio of the average total time spent to the standard total time, the percentage of students whose total time exceeds the standard total time, etc.). Based on these statistical indicators, the system determines the overall completion status of all students at that test site anomaly. Next, the system sets an initial flag (e.g., gray default state) for each test site anomaly in the teacher's interface. Next, test site anomalies are distinguished and marked according to the determined group completion status: if the statistical results show that the total time most students spend at the test site far exceeds the standard total time, or only a very small number of students can complete the identification within the standard total time, then the group completion status is judged to be low, and the test site anomaly is marked with a second color (such as orange), indicating that the interaction time of the test site anomaly exceeds the standard threshold; if the statistical results show that most students can complete the identification within the standard total time, and the average total time spent is close to the standard total time, then the group interaction efficiency is judged to be up to standard, and the test site anomaly is marked with a third color (such as green), indicating that the test site anomaly has passed the standard logic verification; if there is not enough student data or the student participation is extremely low, the test site anomaly retains its initial mark (such as gray), indicating that the corresponding test site anomaly has not yet been triggered with effective interaction. Finally, the distinguishing marks of each test site anomaly are displayed next to the corresponding test site anomaly icon (such as test site name, icon) on the teacher's interface. The technical effect of this step is to achieve macro-level cluster analysis and visualization of abnormal interaction data of all students' test points, enabling teachers to quickly and intuitively understand which test points are common nodes with high interaction latency and which test points have smooth interaction. This solves the problem that teachers have difficulty fully grasping the characteristics of group operations in traditional methods, and provides data support for teachers to carry out targeted strategy adjustments.

[0109] The anomaly detection system in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference]. Figure 4 This is a schematic diagram of the physical device structure of an anomaly detection system in this application embodiment.

[0110] It should be noted that, Figure 4 The structure of the anomaly detection system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0111] like Figure 4As shown, the anomaly troubleshooting system includes a Central Processing Unit (CPU) 401, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 402 or programs loaded from storage section 408 into Random Access Memory (RAM) 403, such as performing the methods described in the above embodiments. The RAM 403 also stores various programs and data required for system operation. The CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An Input / Output (I / O) interface 405 is also connected to the bus 404.

[0112] The following components are connected to I / O interface 405: input section 406 including audio input devices, push-button switches, etc.; output section 407 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 408 including a hard disk, etc.; and communication section 409 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 409 performs communication processing via a network such as the Internet. Drive 410 is also connected to I / O interface 405 as needed. Removable media 411, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 410 as needed so that computer programs read from them can be installed into storage section 408 as needed.

[0113] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by central processing unit (CPU) 401, it performs the various functions defined in the present invention.

[0114] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0115] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0116] Specifically, the anomaly detection system in this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the laboratory multi-person anomaly detection and monitoring method based on VR technology provided in the above embodiment.

[0117] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the anomaly detection system described in the above embodiments; or it may exist independently and not assembled into the anomaly detection system. The storage medium carries one or more computer programs, which, when executed by a processor of the anomaly detection system, enable the anomaly detection system to implement the VR-based laboratory multi-person anomaly detection and monitoring method provided in the above embodiments.

[0118] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0119] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0120] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for monitoring and investigating multiple anomalies in a laboratory based on VR technology, applied to an anomaly investigation system, wherein the anomaly investigation system includes at least one teacher terminal and at least one student terminal, the teacher terminal and the student terminal being connected via a network, characterized in that, The method includes: loading a unified virtual laboratory scene on the teacher's end and the student's end, wherein the virtual laboratory scene presets multiple target anomalies and their corresponding location points and attributes, and the target anomalies include test point anomalies and non-test point anomalies; Multiple standard screening sequences are determined, and each standard screening sequence is an ordered sequence composed of the locations of multiple test point anomalies; Real-time acquisition of each student's behavioral data in the virtual laboratory scene, the behavioral data including at least the VR gaze point movement trajectory; Filter out multiple VR gaze areas from the VR gaze point movement trajectory where the VR gaze point stays for more than a preset observation duration; The multiple VR gaze regions are sorted in chronological order to obtain the student's actual VR gaze attention sequence; For the target student, multiple matching degrees are determined based on the actual VR gaze attention sequence and each of the standard screening sequences, and a visual inspection logic score matching the target student is determined according to the matching degree; the visual inspection logic score is displayed in a preset visualization form next to the corresponding student identity identification module in the teacher's interface, and the preset visualization form includes a progress display matching the visual inspection logic score; The step of determining multiple standard screening sequences specifically includes: Each of the aforementioned test point anomalies is set as a node, and a directed graph of anomaly relationships is constructed based on the relationship between the test point anomalies, defining the node traversal priority, as the logical relationship between the test point anomalies. The relationship includes one of causal relationship, dependency relationship or standard operating procedure relationship. Obtain the three-dimensional coordinates of each test point anomaly in the virtual laboratory scene, and use the pre-set spatial inspection order as the spatial order of the test point anomalies; Using the logical association as the main constraint and the spatial order as the optimization objective, at least one optimal standard screening sequence is generated; Generate at least one non-optimal standard investigation sequence that contains the locations of all test point anomalies, but in a different order than the optimal standard investigation sequence.

2. The method according to claim 1, characterized in that, The step of determining multiple matching degrees for the target student based on the actual VR gaze attention sequence and each of the standard screening sequences specifically includes: From the actual VR gaze attention sequence, select all the test point attention subsequences that correspond to the test point locations in the standard screening sequence and maintain the original time sequence. The test point focus subsequence is compared with each of the standard screening sequences in terms of order and quantity, and multiple logical completeness matching scores containing both order and quantity are calculated. The final logical completeness matching score is determined based on multiple logical completeness matching scores. The actual VR gaze attention sequence is counted, including the number of attention points that belong to the test point location and the number of attention points that do not belong to the test point location. By calculating the proportion of the number of points of attention to the test points in the total number of points of attention in the entire actual VR gaze attention sequence, a screening focus score is obtained to characterize the student screening efficiency. Combining the final logical complete matching score and the investigation focus score, a preset weighted model is used to calculate and generate the final matching degree corresponding to the standard investigation sequence.

3. The method according to claim 2, characterized in that, The step of determining the final logical completeness matching score based on multiple logical completeness matching scores specifically includes: Assign a high-level coefficient to the optimal standard screening sequence and a low-level coefficient to the non-optimal standard screening sequence; If it is determined that the subsequence of the test point has the highest logical integrity match score with the optimal standard sequence in terms of both order and quantity, then the final logical integrity match score is determined by combining the high-level coefficient. If it is determined that the subsequence of the test point has the highest logical completeness match score with the non-optimal standard sequence in terms of both order and quantity, then the final logical completeness match score is determined by combining the low-level coefficient.

4. The method according to claim 2, characterized in that, The step of selecting all test point attention subsequences from the actual VR gaze attention sequence that correspond to the test point locations in the standard screening sequence and maintain their original chronological order includes: Real-time acquisition of interaction behavior data between students and virtual items associated with test centers in the test center attention subsequence, the interaction behavior data including students' virtual physical trajectory in the virtual laboratory scene and their interaction operations on test center anomalies; When a student's actual gaze is focused on the abnormal area of ​​the test site at the same time as the virtual physical trajectory, and the student performs an abnormal interactive operation, it is determined that the student has a valid exploration state of the abnormal test site. A valid investigation operation sequence is formed based on multiple valid investigation statuses; The final test point focus subsequence is obtained based on the effective screening operation sequence.

5. The method according to claim 1, characterized in that, After the step of determining the visual inspection logic score matching the target student based on the matching degree, the method further includes: The actual VR gaze attention sequence is processed to divide it into non-examination point attention subsequences other than the examination point attention subsequence. The examination point attention region in the examination point attention subsequence represents that the student's actual VR gaze falls within any examination point abnormal region, and the non-examination point attention subsequence represents that the student's actual VR gaze falls within any non-examination point abnormal region. Based on the number of test points focused on in the test point focus subsequence and the total time spent at the test points, efficiency bonus items are determined; Based on the number of non-exam points focused on in the non-exam point focus subsequence and the total time spent on non-exam points, efficiency deduction items are determined. The visual inspection logic score is calculated and updated based on the efficiency bonus and efficiency deduction items.

6. The method according to claim 1, characterized in that, The method further includes: When any student's VR gaze point is within the abnormal area of ​​the test point, the abnormal test point will be marked and displayed on the teacher's interface with a first color flashing, so as to generate a prompt signal to indicate that the abnormal test point has been covered by the gaze area. Obtain the real-time dwell time of any student at the test site and display the real-time dwell time next to the corresponding student identification module; Based on the multiple real-time dwell times, calculate the total dwell time of each student at the test site anomaly; When the total dwell time exceeds the preset abnormal dwell time threshold, the test site abnormality is marked with the second color on the teacher's interface. The preset abnormal dwell time threshold is matched with the standard investigation time required for the corresponding target abnormality. If the student identifies the anomaly in the test point, then the anomaly is marked with the third color.

7. An anomaly detection system, characterized in that, The anomaly detection system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the anomaly detection system to perform the method as described in any one of claims 1-6.

8. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is run on the anomaly detection system, it causes the anomaly detection system to perform the method as described in any one of claims 1-6.

9. A computer program product, characterized in that, When the computer program product is run on the anomaly detection system, the anomaly detection system performs the method as described in any one of claims 1-6.

Images