Intelligent operation system for pre-post training examination of power construction unit

Abstract

The invention provides an intelligent operation system for a pre-post training examination of a power construction unit, and relates to the technical field of intelligent operation, and the system comprises an examination task analysis module which is used for analyzing examination task information to obtain a plurality of construction operation tasks; the operation task mapping module is used for mapping a plurality of construction operation tasks to a plurality of operation functional blocks; the time window configuration module is used for configuring a scheduling time window for the operation functional block according to the operation duration identifier; the scheduling and sequencing module is used for scheduling and sequencing the operation functional blocks according to the examination sequence identifiers to obtain a functional block scheduling sequence; and the examination behavior monitoring module is used for performing examination behavior monitoring on the examination user and outputting fusion cheating alarm information. The technical problems that traditional examination monitoring mostly depends on manual inspection or fixed camera monitoring, the behaviors of examinees cannot be comprehensively covered, the abnormal behavior detection accuracy is low, and the accuracy and fairness of examination monitoring are insufficient are solved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent operation technology, and in particular to an intelligent operation system for pre-job training and examination of electric power construction units. Background Art

[0002] As a pillar industry of the national economy, the power industry has special safety needs. In the pre-job training examination of power construction units, it is necessary not only to verify the professional ability of the staff, but also to emphasize the importance of safe operation. Therefore, it is very necessary to use high-tech means to improve the quality of training and the fairness and authenticity of the examination.

[0003] There are many problems with the current safety training examinations for power construction teams. On the one hand, traditional examination monitoring mainly relies on manual proctoring or video surveillance, and has a weak ability to analyze abnormal behaviors of examinees. It is difficult to effectively prevent and identify cheating behaviors, such as taking exams for others, consulting materials, and accepting guidance from others. It is also difficult to deal with new cheating methods, such as obtaining external information through subtle movements or shifting of vision, making it impossible to ensure the fairness and authenticity of the examination. On the other hand, the traditional examination system lacks a task scheduling mechanism and cannot be dynamically optimized according to the time and resource requirements of different tasks, which can easily lead to resource waste or scheduling conflicts, making it difficult to ensure the efficient operation of the system. Summary of the invention

[0004] This application provides an intelligent operating system for pre-job training examinations for power construction units, aiming to solve the technical problem that traditional examination monitoring relies heavily on manual patrols or fixed camera monitoring, which cannot fully cover the behavior of candidates, resulting in low accuracy in abnormal behavior detection, and thus leads to insufficient accuracy and fairness in examination monitoring.

[0005] The present application discloses an intelligent operating system for pre-job training examinations for electric power construction units. The system includes: an examination task parsing module, which is used to obtain multiple construction operation tasks by parsing the examination task information after interactively obtaining the examination task information, wherein the multiple construction operation tasks have multiple examination order identifiers and multiple operation duration identifiers; an operation task mapping module, which is used to map the multiple construction operation tasks to multiple operation function blocks, wherein each operation function block is deployed with a head behavior monitoring task, a limb behavior monitoring task and an eye movement trajectory monitoring task; a time window configuration module, which is used to configure multiple scheduling time windows for the multiple operation function blocks according to the multiple operation duration identifiers; a scheduling sorting module, which is used to perform scheduling sorting of the multiple operation function blocks according to the multiple examination order identifiers to obtain a function block scheduling sequence; an examination behavior monitoring module, which is used to monitor the examination behavior of the examination user performing the multiple construction operation tasks with the multiple scheduling time windows as scheduling duration constraints and the function block scheduling sequence as scheduling order constraints, and output fusion cheating alarm information.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: By acquiring and parsing the test task information, the structure and requirements of the test can be quickly determined, providing efficient task analysis capabilities. Through analysis, specific construction operation tasks and their attributes, including the test order and operation duration, are obtained, laying the foundation for subsequent task allocation and monitoring; according to the analyzed task characteristics, each construction operation task is mapped to the corresponding operation function block. This mapping ensures the best match between the task and the monitoring environment, improves the pertinence and effectiveness of monitoring, and deploys dedicated monitoring tasks for each functional block, including head behavior, limb behavior, and eye movement trajectory monitoring, allowing multiple tasks to be executed in parallel, greatly improving monitoring efficiency and monitoring accuracy; according to the duration of the operation task, a scheduling time window is configured for each functional block to optimize the task. The time management of execution avoids resource conflicts and time waste by accurately controlling the start and end time of tasks, and improves the efficiency and effectiveness of the overall examination process; through systematic scheduling and sorting, it ensures that each functional block is executed in the most logical and efficient order, which not only makes the examination process smoother, but also reduces waiting time and potential monitoring blind spots through reasonable scheduling; using monitoring technology, the behavior of test users is monitored in all directions and in real time, and at the same time, multi-modal monitoring data is comprehensively analyzed through data fusion technology to improve the recognition accuracy and speed of cheating behavior. When potential cheating behavior is monitored, the fusion cheating alarm information is output in time. This information integrates data from various monitoring dimensions and provides reliable decision support for examination management. In summary, the intelligent operating system significantly improves the monitoring efficiency and cheating detection capabilities of pre-job training examinations of power construction units through refined task scheduling and efficient parallel monitoring technology, which ensures the fairness and accuracy of the examination and improves the training quality and the safety of power employees.

[0007] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A schematic diagram of the structure of an intelligent operating system for pre-job training examinations for power construction units provided in an embodiment of the present application.

[0009] Figure 2 A schematic diagram of the execution flow of the examination behavior monitoring module in the intelligent operating system for pre-job training examinations for power construction units provided in an embodiment of the present application.

[0010] Explanation of the reference numerals: examination task parsing module 10 , operation task mapping module 20 , time window configuration module 30 , scheduling and sorting module 40 , examination behavior monitoring module 50 . DETAILED DESCRIPTION

[0011] The embodiment of the present application solves the technical problem that traditional examination monitoring mostly relies on manual inspections or fixed camera monitoring, which cannot fully cover the behavior of candidates, resulting in low accuracy of abnormal behavior detection, and thus leads to insufficient accuracy and fairness of examination monitoring, by providing an intelligent operating system for pre-job training examinations for power construction units.

[0012] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically introduced below in conjunction with the accompanying drawings of the specification. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0013] Embodiment 1, as Figure 1 As shown, the embodiment of the present application provides an intelligent operating system for pre-job training examinations for electric power construction units, and the system includes: The test task parsing module 10 is used to obtain multiple construction operation tasks by parsing the test task information after interactively obtaining the test task information, wherein the multiple construction operation tasks have multiple test sequence identifiers and multiple operation duration identifiers.

[0014] Interact with the examination management system or the academic affairs management system to obtain specific information about the examination task, including the type of examination, examination time, examination location, examination object, and specific construction operation requirements. Parse the examination task information, extract detailed information about the construction operation, and obtain multiple construction operation tasks. The multiple construction operation tasks have multiple examination sequence identifiers and multiple operation duration identifiers, where the examination sequence identifier is the order in which each construction operation task is executed, ensuring that the examination process is consistent with the actual operation sequence of the construction; the operation duration identifier is the length of time required for each operation task, which is crucial for allocating examination time and monitoring the use of time during the examination process.

[0015] The operation task mapping module 20 is used to map the multiple construction operation tasks to multiple operation function blocks, wherein each operation function block is deployed with a head behavior monitoring task, a limb behavior monitoring task and an eye movement trajectory monitoring task.

[0016] The operation function blocks are set up for specific construction operation tasks. Each block is a virtual space configured to complete a specific set of operations. Each block is for a specific construction operation task. These blocks are designed according to the characteristics of the construction operation, such as electrical work, welding work, etc. Each type of operation corresponds to a block. Each operation function block is deployed with head behavior monitoring tasks, limb behavior monitoring tasks and eye movement trajectory monitoring tasks. Among them, the head behavior monitoring task uses a camera facing the examinee to monitor the movement and positioning of the head, such as whether there is frequent looking around or other examination areas, which may indicate an attempt to seek external help; the limb behavior monitoring task uses motion capture technology or video analysis technology to monitor the examinee's hand and body movements, evaluate whether the movements meet the operation standards, and whether there are abnormal movements, such as unnecessary repetitive movements or non-examination operation movements; the eye movement trajectory monitoring task uses an eye tracking device to record the examinee's line of sight movement during the test, analyze their gaze points, and check whether they stay abnormally on test materials or illegal materials, which can reveal the examinee's possible cheating intentions or behaviors.

[0017] Each construction operation task is mapped to the corresponding functional block according to its type and requirements. For example, if a task involves high-precision electrical installation, it is mapped to a functional block that is specially equipped with precision operation monitoring technology. Each monitoring task needs to be configured with corresponding sensors and monitoring equipment, such as cameras, microphones, eye trackers, etc. Through such settings, every behavior of candidates in the construction operation test can be effectively monitored to ensure the fairness and accuracy of the test.

[0018] The time window configuration module 30 is used to configure multiple scheduling time windows for the multiple operation function blocks according to the multiple operation duration identifiers.

[0019] Parse the operation duration identifier of each construction operation task. These identifiers indicate the length of time required to complete each task. According to the operation duration identifier, configure the corresponding scheduling time window for each operation function block. The scheduling time window is the time range within which the specified task must start and be completed. The configuration of the time window takes into account the complexity and urgency of the task to ensure that there is enough time to complete the task without affecting the execution of subsequent tasks.

[0020] The scheduling and sorting module 40 is used to schedule and sort the multiple operation function blocks according to the multiple test order identifiers to obtain a function block scheduling sequence.

[0021] The scheduling and sorting of the operation function blocks are based on the examination order identifiers. These identifiers specify the order in which each task is executed. The examination order identifiers are converted into an operational sequence to ensure that each task can be performed in the predetermined order. According to the examination order identifiers, all operation function blocks are sorted. This sequence determines the specific time point of each block activity. The function block scheduling sequence takes into account the dependencies and time windows of each function block to ensure the continuity and logic of the entire examination process.

[0022] The examination behavior monitoring module 50 is used to monitor the examination behavior of the examination user in performing the multiple construction operation tasks with the multiple scheduling time windows as scheduling duration constraints and the functional block scheduling sequence as scheduling order constraints, and output fused cheating alarm information.

[0023] Examination behavior monitoring is strictly carried out according to the set scheduling time window. The activities of each operation function block must be started and completed within the specified time window. The time window ensures that the monitoring task can be started and ended in time to avoid affecting other tasks, while ensuring that each task has enough time for monitoring; monitoring activities are carried out according to the function block scheduling sequence. This sequence stipulates the startup order of each operation function block. Compliance with the sequence ensures the logic and coherence of the examination process, so that monitoring can be smoothly switched between various function blocks.

[0024] During the exam, monitoring equipment deployed in each functional block, such as cameras and sensors, is used to capture the behavioral data of candidates in real time, including head, limb and eye movement trajectories, analyze the collected multimodal data, identify behavioral patterns that do not meet the regulations or abnormal behaviors that are significantly different from normal behaviors, and generate fusion cheating alarm information once cheating or other abnormalities are detected. These alarm information are based on the results of comprehensive analysis, including behavioral data, time stamps and specific abnormal details. Through such monitoring and cheating detection mechanisms, the intelligent operating system can ensure the fairness and effectiveness of pre-job training examinations for power construction units, while improving the utilization rate of monitoring data and the accuracy of cheating detection.

[0025] Furthermore, if Figure 2 As shown, the multiple scheduling time windows are used as scheduling duration constraints, and the functional block scheduling sequence is used as scheduling order constraints, the test user's test behavior of performing the multiple construction operation tasks is monitored, and fusion cheating alarm information is output, including: When the test user executes the first construction operation task in sequence according to the multiple test sequence identifiers, the first operation function block is scheduled from the function block scheduling sequence; the first operation function block receives the monitoring video stream returned by the multimodal sensor array to execute parallel monitoring tasks until the duration meets the first scheduling time window, and outputs the first cheating alarm information; after the computing power resource configuration of the second operation function block is updated based on the first cheating alarm information, the second operation function block is scheduled from the function block scheduling sequence; the second operation function block receives the monitoring video stream returned by the multimodal sensor array to execute parallel monitoring tasks until the duration meets the second scheduling time window, and outputs the second cheating alarm information; and so on, the computing power resource configuration of the next operation function block is updated and the test behavior monitoring is performed according to the cheating alarm output of the previous operation function block, until the test behavior monitoring of the multiple construction operation tasks is completed, and the fused cheating alarm information is output.

[0026] When the test user starts to perform the first construction operation task according to the set test order, the first operation function block is activated according to the function block scheduling sequence, and the test user performs a specific construction operation task in the first operation function block, which has been pre-set to meet the requirements of these operations.

[0027] The first operation function block monitors the behavior of test users in real time through a multimodal sensor array, including video cameras, infrared sensors, etc. These monitoring tasks include head behavior, limb behavior and eye movement trajectory monitoring, which can be executed in parallel. During the monitoring process, if any abnormal or possible cheating behavior is monitored, and once the monitoring time reaches the requirements of the first scheduling time window, the first cheating alarm information is generated and output. This information includes detailed information such as student ID, type of cheating behavior detected, time of occurrence and related operation steps.

[0028] According to the first cheating alarm information, the computing power resource configuration of the second operating function block is updated, including increasing computing power to improve the accuracy of monitoring or adjusting resources to respond to specific detected behavior patterns. After updating the resource configuration, the second operating function block is scheduled to continue monitoring the examination process, which ensures the optimal use of resources and adjusts the monitoring strategy based on real-time feedback.

[0029] The second operation function block continues to receive the monitoring video stream sent back by the multimodal sensor array and executes the multimodal monitoring task, including monitoring of the head, limb behavior and eye movement trajectory. The monitoring continues until the predetermined scheduling time window of the second operation function block, i.e., the second scheduling time window, is met. Once the monitoring data shows abnormal behavior, a second cheating alarm message is generated, which includes detailed information such as the student ID, the type of cheating behavior detected, the time of occurrence and the relevant operation steps.

[0030] Similar to the previous steps, the computing resource configuration of the subsequent functional blocks is adjusted according to the cheating alarm information output from each functional block. This dynamic adjustment is based on the cheating alarm result of the previous operation functional block, aiming to optimize resource allocation and improve monitoring effect. All construction operation tasks are continuously monitored. After each task is completed, the monitoring resources of the subsequent tasks will be updated according to the cheating alarm information and monitoring data. After completing the monitoring of all construction operation tasks, a fusion cheating alarm information is output. This information summarizes all relevant cheating data and analysis results collected during the entire examination process, providing a comprehensive overview of cheating behavior.

[0031] Furthermore, when the test user performs the first construction operation task according to the plurality of test sequence identifiers, the first operation function block is scheduled from the function block scheduling sequence, before that, the method includes: After a random number array is generated through a random number pool, the random number array is sent to the test user; the real-time voice data and real-time facial video stream of the test user are collected in real time through the multimodal sensor array; the target sample voice is retrieved and called from the voice sample library according to the test number of the test user; after the real-time voice data is falsified by using the target sample voice, the synchronization of the real-time voice data and the real-time facial video stream is checked; if the result of the synchronization check is passed, a voice instruction is generated to prompt the test user to perform the first construction operation task.

[0032] A random number pool is used to generate a random array, which is used as a one-time verification code or task sequence to verify the test user's response or operation, and the generated random array is sent to the test user.

[0033] A multimodal sensor array is configured, including but not limited to cameras and microphones, to collect the voice data and facial video streams of the test users in real time. Through these devices, the real-time voice and facial expressions of the test users are collected in real time, and these data can be used for subsequent identity authentication and behavioral analysis.

[0034] Each test user has a unique test number, which is used to retrieve the corresponding target sample voice from the pre-prepared voice sample library. The retrieved target sample voice is used to match the real-time collected voice data to check the consistency of the voice, thereby verifying the identity of the test user or detecting possible proxy test behavior.

[0035] Use the target sample voice retrieved from the voice sample library to verify the real-time voice data collected in real time to ensure the authenticity and consistency of the voice data, so as to prove the identity of the test user or discover any proxy test behavior. After passing the falsification check, further synchronization verification of the real-time voice data and the real-time facial video stream is performed. This verification confirms the temporal consistency of the voice and facial expression data to ensure that the recorded data has not been tampered with or delayed, and to ensure the correspondence of the audio and video data.

[0036] If the synchronization check result is passed, it indicates that the real-time voice and facial video streams are matched in time and the data is authentic and reliable. After the check is passed, a voice command is generated to prompt the test user to start the first construction operation task. This command is conveyed to the test user through the system's audio output device to guide the user to perform the next specific operation.

[0037] Furthermore, the first operation function block receives the monitoring video stream returned by the multimodal sensor array, executes the parallel monitoring task until the duration meets the first scheduling time window, and outputs the first cheating alarm information, including: 1 / N of the first scheduling time window is used as a multimodal monitoring scale; when the image acquisition sensor in the multimodal sensor array returns a first periodic video stream to the first operation function block with the multimodal monitoring scale as a constraint, the first operation function block activates the behavior stream division module to divide the first periodic video stream into a head behavior video stream, a limb behavior video stream and an eye movement trajectory video stream; through the head behavior monitoring task, the limb behavior monitoring task and the eye movement trajectory monitoring task of the first operation function block, the abnormal behavior recognition of the head behavior video stream, the limb behavior video stream and the eye movement trajectory video stream is mapped and executed in parallel to obtain a first head abnormal sequence, a first limb abnormal sequence and a first eye movement abnormal sequence; by analogy, the first operation function block analyzes the N periodic video streams returned by the image acquisition sensor, and outputs N head abnormal sequences, N limb abnormal sequences and N eye movement abnormal sequences; by performing time series fusion analysis on the N head abnormal sequences, N limb abnormal sequences and N eye movement abnormal sequences, the first cheating alarm information is output.

[0038] 1 / N of the first scheduling time window is defined as the time scale of multimodal monitoring. N is set according to specific needs and computing power. This means that within the first scheduling time window, data is periodically acquired from the multimodal sensor array according to this time scale. By setting such a time scale, data can be collected and processed regularly to ensure continuous updating of monitoring data and avoid information overload.

[0039] The image acquisition sensor in the multimodal sensor array transmits the video stream back to the first operating function block according to the set multimodal monitoring scale. The video stream is transmitted back according to the predetermined monitoring scale to ensure that there is new video data for analysis in each cycle, so as to capture any possible abnormal behavior or cheating behavior in time.

[0040] When the first cycle video stream is received, the first operation function block activates the behavior stream segmentation module, which uses the region of interest (ROI) to subdivide the video stream into three main parts, including head behavior video stream, limb behavior video stream and eye movement track video stream. Each video only targets the behavior of one human body area. Among them, the head behavior video stream focuses on capturing the candidate's head movements, such as turning the head and nodding; the limb behavior video stream monitors the movements of the candidate's hands and other limbs to check whether there are gestures or movements that do not comply with the regulations; eye movement track video stream: tracks the candidate's eye movements, analyzes his or her attention focus and checks whether the behavior is normal.

[0041] In the first operation function block, the head behavior monitoring task, the limb behavior monitoring task and the eye movement trajectory monitoring task are executed in parallel. Each monitoring task focuses on its corresponding video stream, namely the head behavior video stream, the limb behavior video stream and the eye movement trajectory video stream. These video streams are analyzed through image processing and machine learning techniques to identify abnormal behaviors. These technologies may include behavior pattern recognition, action recognition algorithms and anomaly detection mechanisms for detecting situations where behaviors deviate from normal or expected patterns. After abnormal behavior identification, the first head abnormal sequence, the first limb abnormal sequence and the first eye movement abnormal sequence are obtained. Through these detailed abnormal sequences, any possible cheating behavior can be responded to in a timely manner, thereby ensuring the effectiveness and fairness of the examination.

[0042] Similarly, the first operating function block receives N periodic video streams sent back by the image acquisition sensor, and each periodic video stream is captured regularly according to the previously set multimodal monitoring scale to ensure the continuity and coverage of the behavior monitoring data.

[0043] The head behavior video stream is analyzed cycle by cycle to generate N abnormal head sequences, which describe the abnormal head behavior characteristics within a specific time period, such as the number of head turns, abnormal head movements, etc.; the limb behavior video stream is analyzed cycle by cycle to generate N abnormal limb sequences, which describe the abnormal limb behavior characteristics within a specific time period, such as abnormal gestures, suspicious operation behaviors, etc.; the eye movement behavior video stream is analyzed cycle by cycle to generate N abnormal eye movement sequences, which describe the abnormal eye movement behavior characteristics within a specific time period, such as frequent shifting of attention, focusing on non-examination areas for too long, etc.

[0044] A comprehensive time series analysis is performed on N abnormal head, limb and eye movement sequences. The time series fusion technology is used to integrate these sequence data to obtain global behavioral characteristics. Time series analysis methods such as dynamic time warping or long short-term memory network are used. The goal of the fusion is to detect the correlation between behaviors, such as whether the head abnormality is synchronized with the limb abnormality, and its time evolution law. Time series fusion can discover complex cheating patterns that cannot be reflected in a single sequence. For example, head, limb and eye movement behaviors may be abnormal at the same time at certain time points, indicating a higher possibility of cheating. Combined with the time series analysis results, the first cheating alarm information is generated, including the student ID, the type of cheating behavior detected, the time of occurrence, and the relevant operation steps.

[0045] Further, including: Interactively obtain multiple head violation image sets of multiple sample head violation types, and use the multiple head violation image sets to construct a head violation recognition model; preset a frame extraction interval scale to perform frame extraction processing on the head behavior video stream to obtain multiple head behavior frames, wherein the multiple head behavior frames have multiple acquisition time identifiers; the head behavior monitoring task drives the head violation recognition model to perform one-by-one identification of the multiple head behavior frames to obtain multiple violation identification results, wherein the multiple violation identification results have multiple head violation type identifiers; perform inter-frame behavior consistency detection on the multiple violation identification results, and calculate the duration of the violation behavior based on the detection results, and output the first head abnormality sequence.

[0046] By interacting with the examination system or using data collection methods, we can obtain a head violation image set of various sample head violation types. These image sets include head movements that are not allowed during the examination, such as wandering eyes, frequent sideways or upward glances, etc. The image sets cover a variety of violation behavior types to ensure that the recognition model can cover all possible violations. These head violation image sets are used to train the head violation recognition model, which uses machine learning or deep learning technology, such as convolutional neural networks, to perform image analysis and pattern recognition. The trained model can analyze the examination video stream in real time to identify whether there is wandering eyes or other head movements that do not meet the examination specifications.

[0047] When processing the head behavior video stream, key frames are extracted according to the preset frame extraction interval scale. This interval scale is set according to the behavior dynamics and monitoring requirements, aiming to optimize the balance between data processing efficiency and monitoring accuracy. For example, one frame can be extracted every second, or the frame extraction rate can be dynamically adjusted according to the density of action changes. The extracted frames are marked with acquisition time identifiers, which are associated with specific time points of the original video stream to facilitate subsequent behavior analysis and time alignment.

[0048] The head behavior monitoring task drives the head violation recognition model, which is trained in the previous steps and is specifically used to analyze head behavior. The model analyzes each extracted head behavior frame one by one to identify whether there is a violation. The output violation recognition results include various types of head violations, such as wandering eyes, abnormal head rotation, etc. Each recognition result is accompanied by a head violation type identifier, which records the specific category of the violation in detail, providing a clear classification basis for subsequent processing.

[0049] The behavior in multiple consecutive video frames is checked for consistency. Only when consecutive frames are identified as violations and the types of violations are consistent, are they confirmed as valid violations. This method helps to eliminate accidental or misjudged violations and increase the accuracy of recognition. By identifying violations in consecutive frames, the duration of the violations is calculated. This duration is used to judge the severity of the violations and determine the corresponding penalties. Based on the above analysis, the first head abnormal sequence is output, which includes the type identification and duration of each violation, providing detailed data for subsequent monitoring reports and analysis.

[0050] Similarly, the eye movement trajectory monitoring task drives the eye movement violation recognition model to identify and analyze the examinee's eye movement behavior, identify anomalies and calculate the duration of the anomalies, and generate a first eye movement anomaly sequence. The method for handling abnormal eye movement behavior is the same as the method for handling abnormal head behavior, which will not be repeated here.

[0051] Further, including: Interactively obtain multiple limb violation image sets of multiple sample limb violation types, and use the multiple limb violation image sets to construct a limb violation recognition model; the limb behavior monitoring task drives the limb violation recognition model to perform abnormality recognition of the limb behavior video stream to obtain the first limb abnormality sequence.

[0052] By interacting with the examination system or automating data collection, image sets reflecting various types of physical violations are collected. These image sets include various physical movements that do not comply with examination regulations, such as impermissible gestures, movements of carrying or using prohibited items, etc. The collected image data covers a wide range to ensure that the recognition model can adapt to different violation situations. The physical violation recognition model is trained using the collected physical violation image set. The model can be based on deep learning technology, such as convolutional neural networks, which are specially designed to identify and classify various physical violations. After the model training is completed, it can accurately identify physical violations in real-time video streams.

[0053] The physical behavior monitoring task drives the trained physical violation recognition model to analyze the physical behavior video stream in the exam. The abnormality recognition process of the physical violation recognition model is similar to that of the head violation recognition model. For the sake of brevity, it is not repeated here. After abnormality recognition, the first physical violation sequence is output. The physical violation sequence records in detail the type, occurrence time, duration and other information of the violation, providing data support for further analysis and monitoring reports.

[0054] Furthermore, by performing time series fusion analysis on the N head abnormality sequences, the N limb abnormality sequences, and the N eye movement abnormality sequences, the first cheating alarm information is output, including: According to the data acquisition timing, the N head abnormality sequences, N limb abnormality sequences and N eye movement abnormality sequences are spliced ​​to obtain the test head abnormality sequence, the test limb abnormality sequence and the test eye movement abnormality sequence; interactively obtain an associated abnormality scoring table, wherein the associated abnormality scoring table records multiple cheating quantitative scores for multiple associated abnormal behaviors; use the multiple associated abnormal behaviors to traverse the test head abnormality sequence, the test limb abnormality sequence and the test eye movement abnormality sequence, and after locating M associated abnormal behaviors, remove the M associated abnormal behaviors from the test head abnormality sequence, the test limb abnormality sequence and the test eye movement abnormality sequence to obtain a single-dimensional head abnormality sequence, a single-dimensional limb abnormality sequence and a single-dimensional eye movement abnormality sequence. ; According to the M associated abnormal behaviors, the associated abnormal scoring table is traversed to obtain the associated abnormal cheating coefficient by calculation; the head abnormal scoring table, the limb abnormal scoring table and the eye movement abnormal scoring table are obtained interactively; the head abnormal scoring table, the limb abnormal scoring table and the eye movement abnormal scoring table are traversed by adopting the one-dimensional head abnormal sequence, the one-dimensional limb abnormal sequence and the one-dimensional eye movement abnormal sequence mapping, and the one-dimensional abnormal cheating coefficient is obtained based on the sum of the traversed scoring results; the associated abnormal cheating coefficient and the one-dimensional abnormal cheating coefficient are added to obtain the first comprehensive cheating coefficient; the first comprehensive cheating coefficient, the examination head abnormal sequence, the examination limb abnormal sequence and the examination eye movement abnormal sequence are stored in association as the first cheating alarm information.

[0055] The data collection sequence is the time sequence of video stream collection. The N head abnormality sequences, N limb abnormality sequences and N eye movement abnormality sequences collected during the entire test are spliced ​​according to the data collection sequence. Such splicing ensures the temporal continuity and integrity of the sequences, so that each type of abnormal behavior can be fully recorded during the entire test. After splicing, three complete test abnormality sequences are formed, namely, the test head abnormality sequence, the test limb abnormality sequence and the test eye movement abnormality sequence. These sequences record in detail the occurrence time, duration and specific type of each abnormal behavior, providing a comprehensive data basis for subsequent analysis and evaluation.

[0056] The associated abnormal scoring table is obtained through automated database query and other means. This table records the quantitative cheating scores of various associated abnormal behaviors. Each cheating behavior is assigned different scores according to its severity and probability of cheating. These scores reflect the potential threat of various abnormal behaviors to the fairness of the examination. Through quantitative scoring, the abnormal behaviors of different examinations and candidates can be evaluated and compared more objectively and uniformly.

[0057] Using multiple associated abnormal behaviors, we traverse the integrated test head abnormality sequence, test limb abnormality sequence, and test eye movement abnormality sequence to find possible associated abnormal behaviors. Associated abnormal behaviors refer to abnormal behaviors that occur simultaneously or continuously in a short period of time in different dimensions (head, limb, eye movement). By analyzing the timestamps and abnormality types in each abnormal sequence, we identify M associated abnormal behaviors, which may manifest as coordinated or continuous abnormal patterns during the test.

[0058] When the associated abnormal behaviors are identified, these behaviors are removed from the original abnormal sequence. After removal, the remaining abnormal behaviors form new unidimensional head abnormal sequences, unidimensional limb abnormal sequences, and unidimensional eye movement abnormal sequences. This step is used to clearly distinguish isolated abnormal behaviors from compound abnormal behaviors caused by collaborative cheating.

[0059] The M identified associated abnormal behaviors are matched with the records in the associated abnormal scoring table. The scoring table provides quantitative scores for associated abnormal behaviors of different types and patterns. These scores reflect the potential cheating severity of various behaviors. For each associated abnormal behavior, calculations are performed based on the scores in the scoring table, and the cheating scores of all associated abnormal behaviors are summarized to form a comprehensive associated abnormal cheating coefficient. This coefficient provides a quantitative and comprehensive cheating risk assessment for test monitoring personnel, which helps to accurately determine the severity of cheating behavior and the suspicion of cheating in the test.

[0060] Three types of abnormal scoring tables are obtained through the automatic data interface, namely the head abnormality scoring table, the limb abnormality scoring table and the eye movement abnormality scoring table. Each scoring table records the different levels of the corresponding type of abnormal behavior and its corresponding scores. These scores are quantified based on the potential impact of abnormal behavior on the fairness of the examination.

[0061] The one-dimensional head abnormality sequence, one-dimensional limb abnormality sequence and one-dimensional eye movement abnormality sequence after removing associated abnormalities are mapped to the corresponding scoring table respectively. In the scoring table, each abnormal behavior has a predetermined score. These scores are found and applied according to the abnormality type in the sequence. For each abnormality type, the corresponding scoring result is obtained. Finally, the scores of the scoring results of each abnormality are added up to calculate the one-dimensional abnormal cheating coefficient. This coefficient provides a detailed assessment of the cheating behavior of candidates in each monitoring dimension.

[0062] The calculated associated abnormal cheating coefficient and single-dimensional abnormal cheating coefficient are added together. The associated abnormal cheating coefficient reflects the potential coordinated cheating behavior among multiple monitoring dimensions, while the single-dimensional abnormal cheating coefficient reflects the abnormal behavior in each independent dimension. The total score after the addition is defined as the first comprehensive cheating coefficient. This coefficient provides a quantitative and comprehensive assessment, indicating the overall potential severity of cheating by candidates throughout the entire examination process.

[0063] The first comprehensive cheating coefficient is associated with the related abnormal sequences, namely the abnormal head sequence, the abnormal limb sequence and the abnormal eye movement sequence, and stored. These data together form detailed first cheating alarm information, which not only includes the quantitative score of cheating, but also records the specific type and time information of cheating behavior in detail. The first cheating alarm information can be used by test monitoring personnel and managers to help them conduct detailed cheating behavior analysis, evaluate the severity of cheating behavior, and formulate corresponding disposal measures based on this information, ensuring the integrity and fairness of the test.

[0064] Furthermore, after updating the computing resource configuration of the second operating function block based on the first cheating alarm information, scheduling the second operating function block from the function block scheduling sequence includes: The first comprehensive cheating coefficient is called from the first cheating alarm information; the first comprehensive cheating coefficient is used to traverse the computing power allocation adjustment table to obtain the first computing power update information; after the computing power resource configuration of the second operating function block is updated by using the first computing power update information, the second operating function block is scheduled from the function block scheduling sequence.

[0065] The first comprehensive cheating coefficient is called from the stored first cheating alarm information. This coefficient combines abnormal behaviors in multiple monitoring dimensions such as head, limbs, and eye movements, and reflects the overall potential cheating level of the examinee.

[0066] The first comprehensive cheating coefficient is used to traverse the computing power allocation adjustment table. This table presets different computing power adjustment strategies according to different values ​​of the cheating coefficient, including increasing or adjusting the computing power of specific monitoring tasks according to the cheating coefficient, such as enhancing the processing capability of the deep learning algorithm for image analysis, or adjusting the priority of the data stream, etc.

[0067] According to the first computing power update information, the computing power resources of the second operating function block are updated accordingly, including increasing processing power, optimizing data processing flow, etc., in order to strengthen the monitoring capability of potential high-risk behaviors and ensure that the next monitoring tasks can be handled more effectively. After updating the computing power configuration, the second operating function block is started according to the functional block scheduling sequence, which ensures that the monitoring task can continue after resource optimization, so as to monitor the next examination link more accurately.

[0068] Furthermore, the multiple scheduling time windows are used as scheduling duration constraints, and the functional block scheduling sequence is used as a scheduling order constraint, the test user's test behavior of performing the multiple construction operation tasks is monitored, and fusion cheating alarm information is output, including: Taking the multiple scheduling time windows as scheduling duration constraints, and taking the functional block scheduling sequence as scheduling order constraints, the test user's test behavior in executing the multiple construction operation tasks is monitored in real time to obtain multiple comprehensive cheating coefficients, multiple test head abnormality sequences, multiple test limb abnormality sequences, and multiple test eye movement abnormality sequences; weight configuration is performed according to the execution difficulty of the multiple construction operation tasks to obtain multiple construction difficulty weights; the multiple construction difficulty weights are used to perform weighted calculation of the multiple comprehensive cheating coefficients to obtain a fused cheating coefficient; the multiple test head abnormality sequences, multiple test limb abnormality sequences, and multiple test eye movement abnormality sequences are spliced ​​according to the multiple test order identifier mappings to obtain a fused head abnormality sequence, a fused limb abnormality sequence, and a fused eye movement abnormality sequence; the fused cheating coefficient, the fused head abnormality sequence, the fused limb abnormality sequence, and the fused eye movement abnormality sequence are associated and stored to obtain the fused cheating alarm information.

[0069] Multiple scheduling time windows are used as scheduling time constraints to ensure that each operating functional block completes the monitoring task within the specified time, and the functional block scheduling sequence is used as the scheduling order constraint to follow the predetermined scheduling order. Such time and order constraints help the system effectively manage and optimize resources and ensure that all monitoring activities are started and ended at the appropriate time.

[0070] In each functional block, multiple construction operation tasks performed by the test users are monitored in real time. The monitoring content includes head behavior, limb behavior, eye movement trajectory, etc., to ensure that the test behavior is fully captured. Through real-time monitoring and the aforementioned analysis process, multiple comprehensive cheating coefficients and corresponding abnormal behavior sequences are generated, including multiple test head abnormal sequences, multiple test limb abnormal sequences, and multiple test eye movement abnormal sequences.

[0071] The difficulty of executing multiple construction operation tasks is evaluated. Different tasks have different difficulty levels due to their complexity, required skills and other factors. Based on the evaluation of task difficulty, a corresponding weight value is configured for each construction operation task. These weights reflect the difficulty of the task and are used to calculate the weighted cheating coefficient. The weight configuration helps to make the calculation of the cheating coefficient more accurate and ensure that abnormal behaviors in high-difficulty tasks receive more attention, as these tasks may provide greater motivation and opportunities for cheating.

[0072] Multiple set construction difficulty weights are used to perform weighted calculation on the comprehensive cheating coefficient of each construction operation task. This weighted process combines the difficulty of different tasks and reflects the potential severity of cheating behavior in more complex or technical tasks. After weighted calculation, an integrated cheating coefficient is obtained. This coefficient integrates the cheating risk assessment in all construction operation tasks and provides a quantitative measure for the cheating risk of the entire exam.

[0073] Multiple exam head abnormality sequences, multiple exam limb abnormality sequences, and multiple exam eye movement abnormality sequences are spliced ​​together according to the exam order identifier. This step ensures the continuity and integrity of the abnormal data in the time series. The spliced ​​data form fused head abnormality sequences, fused limb abnormality sequences, and fused eye movement abnormality sequences. These sequences contain all relevant abnormal information from the beginning to the end of the exam, and record in detail the time and type of abnormal behavior.

[0074] The fused cheating coefficient and the fused abnormal sequence, including the fused head abnormal sequence, the fused limb abnormal sequence and the fused eye movement abnormal sequence, are associated and stored to form comprehensive fused cheating alarm information. This alarm information includes cheating behavior assessments and detailed data in all dimensions, providing monitoring personnel with a comprehensive view of cheating behavior.

[0075] Through such steps, the intelligent operating system can achieve efficient and accurate cheating monitoring and analysis during the examination process, which not only improves the utilization efficiency of monitoring data, but also enhances the comprehensiveness and reliability of cheating detection. This comprehensive analysis helps to ensure the fairness and transparency of the examination and maintain the integrity of the examination.

[0076] In summary, the intelligent operation method for pre-job training examination of electric power construction units provided in the embodiment of the present application has the following technical effects: By acquiring and parsing the test task information, the structure and requirements of the test can be quickly determined, providing efficient task analysis capabilities. Through analysis, specific construction operation tasks and their attributes, including the test order and operation duration, are obtained, laying the foundation for subsequent task allocation and monitoring; according to the analyzed task characteristics, each construction operation task is mapped to the corresponding operation function block. This mapping ensures the best match between the task and the monitoring environment, improves the pertinence and effectiveness of monitoring, and deploys dedicated monitoring tasks for each functional block, including head behavior, limb behavior, and eye movement trajectory monitoring, allowing multiple tasks to be executed in parallel, greatly improving monitoring efficiency and monitoring accuracy; according to the duration of the operation task, a scheduling time window is configured for each functional block to optimize the task. The time management of execution avoids resource conflicts and time waste by accurately controlling the start and end time of tasks, and improves the efficiency and effectiveness of the overall examination process; through systematic scheduling and sorting, it ensures that each functional block is executed in the most logical and efficient order, which not only makes the examination process smoother, but also reduces waiting time and potential monitoring blind spots through reasonable scheduling; using monitoring technology, the behavior of test users is monitored in all directions and in real time, and at the same time, multi-modal monitoring data is comprehensively analyzed through data fusion technology to improve the recognition accuracy and speed of cheating behavior. When potential cheating behavior is monitored, the fusion cheating alarm information is output in time. This information integrates data from various monitoring dimensions and provides reliable decision support for examination management. In summary, the intelligent operating system significantly improves the monitoring efficiency and cheating detection capabilities of pre-job training examinations of power construction units through refined task scheduling and efficient parallel monitoring technology, which ensures the fairness and accuracy of the examination and improves the training quality and the safety of power employees.

[0077] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent operating system for pre-job training and examination of power construction units, characterized by: The system comprises: An examination task parsing module, wherein the examination task parsing module is used to obtain a plurality of construction operation tasks by parsing the examination task information after interactively obtaining the examination task information, wherein the plurality of construction operation tasks have a plurality of examination sequence identifiers and a plurality of operation duration identifiers; An operation task mapping module, wherein the operation task mapping module is used to map the multiple construction operation tasks to multiple operation function blocks, wherein each operation function block is deployed with a head behavior monitoring task, a limb behavior monitoring task and an eye movement trajectory monitoring task; A time window configuration module, the time window configuration module is used to configure multiple scheduling time windows for the multiple operation function blocks according to the multiple operation duration identifiers; A scheduling and sorting module, the scheduling and sorting module is used to schedule and sort the multiple operation function blocks according to the multiple test order identifiers to obtain a function block scheduling sequence; An examination behavior monitoring module is used to monitor the examination behavior of the examination user in performing the multiple construction operation tasks using the multiple scheduling time windows as scheduling duration constraints and the functional block scheduling sequence as scheduling order constraints, and output fused cheating alarm information.

2. The intelligent operating system for pre-job training examination for electric power construction units as claimed in claim 1 is characterized in that: Taking the multiple scheduling time windows as scheduling duration constraints and the functional block scheduling sequence as scheduling order constraints, monitoring the test user's test behavior of performing the multiple construction operation tasks, and outputting fusion cheating alarm information, including: When the test user performs the first construction operation task in sequence according to the plurality of test order identifiers, scheduling the first operation function block from the function block scheduling sequence; The first operation function block receives the monitoring video stream sent back by the multi-modal sensor array, executes the parallel monitoring task until the duration meets the first scheduling time window, and outputs the first cheating alarm information; After updating the computing resource configuration of the second operating function block based on the first cheating alarm information, scheduling the second operating function block from the function block scheduling sequence; The second operation function block receives the monitoring video stream returned by the multimodal sensor array, executes the parallel monitoring task until the duration meets the second scheduling time window, and outputs the second cheating alarm information; By analogy, the computing resource configuration of the next operation function block is updated and the examination behavior is monitored according to the cheating alarm output of the previous operation function block, until the examination behavior monitoring of the multiple construction operation tasks is completed and the fused cheating alarm information is output.

3. The intelligent operating system for pre-job training examination for electric power construction units as claimed in claim 2 is characterized in that: When the test user performs the first construction operation task according to the plurality of test sequence identifiers, the first operation function block is scheduled from the function block scheduling sequence, before that, the method includes: After a random number array is generated through a random number pool, the random number array is sent to the test user; The multimodal sensor array collects the real-time voice data and real-time facial video stream of the test user in real time; Retrieving and calling a target sample voice from a voice sample library according to the test number of the test user; After the real-time voice data is verified by using the target sample voice, synchronization verification of the real-time voice data and the real-time facial video stream is performed; If the synchronization check result is passed, a voice instruction is generated to prompt the test user to perform the first construction operation task.

4. The intelligent operating system for pre-job training examination for electric power construction units as claimed in claim 2 is characterized in that: The first operation function block receives the monitoring video stream returned by the multi-modal sensor array, executes the parallel monitoring task until the duration meets the first scheduling time window, and outputs the first cheating alarm information, including: Using 1 / N of the first scheduling time window as a multimodal monitoring scale; When the image acquisition sensor in the multimodal sensing array transmits the first periodic video stream back to the first operation function block based on the multimodal monitoring scale, the first operation function block activates the behavior stream division module to divide the first periodic video stream into a head behavior video stream, a limb behavior video stream and an eye movement trajectory video stream; Through the head behavior monitoring task, limb behavior monitoring task and eye movement trajectory monitoring task of the first operation function block, the abnormal behavior recognition of the head behavior video stream, the limb behavior video stream and the eye movement trajectory video stream is mapped and executed in parallel to obtain a first head abnormal sequence, a first limb abnormal sequence and a first eye movement abnormal sequence; Similarly, the first operation function block analyzes the N periodic video streams returned by the image acquisition sensor, and outputs N head abnormality sequences, N limb abnormality sequences, and N eye movement abnormality sequences; The first cheating alarm information is output by performing time series fusion analysis on the N head abnormality sequences, the N limb abnormality sequences and the N eye movement abnormality sequences.

5. The intelligent operating system for pre-job training and examination for electric power construction units as claimed in claim 4, characterized in that: include: Interactively obtaining a plurality of head violation image sets of a plurality of sample head violation types, and constructing a head violation recognition model using the plurality of head violation image sets; Perform frame extraction processing on the head behavior video stream at a preset frame extraction interval scale to obtain a plurality of head behavior frames, wherein the plurality of head behavior frames have a plurality of acquisition time identifiers; The head behavior monitoring task drives the head violation recognition model to recognize the multiple head behavior frames one by one, and obtains multiple violation recognition results, wherein the multiple violation recognition results have multiple head violation type identifiers; An inter-frame behavior consistency detection is performed on the multiple violation identification results, and a duration of the violation behavior is calculated based on the detection results, and the first head abnormal sequence is output.

6. The intelligent operating system for pre-job training and examination of electric power construction units as claimed in claim 5, characterized in that: include: Interactively obtaining a plurality of limb violation image sets of a plurality of sample limb violation types, and constructing a limb violation recognition model using the plurality of limb violation image sets; The limb behavior monitoring task drives the limb violation recognition model to perform abnormality recognition of the limb behavior video stream to obtain the first limb abnormality sequence.

7. The intelligent operating system for pre-job training and examination of electric power construction units as claimed in claim 6, characterized in that: By performing time series fusion analysis on the N head abnormality sequences, the N limb abnormality sequences, and the N eye movement abnormality sequences, the first cheating alarm information is output, including: splicing the N head abnormality sequences, the N limb abnormality sequences and the N eye movement abnormality sequences according to the data acquisition timing to obtain an examination head abnormality sequence, an examination limb abnormality sequence and an examination eye movement abnormality sequence; Interactively obtain a correlation abnormality scoring table, wherein the correlation abnormality scoring table records multiple cheating quantitative scores for multiple correlation abnormal behaviors; The multiple associated abnormal behaviors are used to traverse the test head abnormality sequence, the test limb abnormality sequence, and the test eye movement abnormality sequence, and after locating M associated abnormal behaviors, the M associated abnormal behaviors are removed from the test head abnormality sequence, the test limb abnormality sequence, and the test eye movement abnormality sequence to obtain a single-dimensional head abnormality sequence, a single-dimensional limb abnormality sequence, and a single-dimensional eye movement abnormality sequence; According to the M associated abnormal behaviors, the associated abnormal scoring table is traversed to calculate the associated abnormal cheating coefficient; Interactively obtain the head abnormality scoring table, limb abnormality scoring table and eye movement abnormality scoring table; The head abnormality scoring table, the limb abnormality scoring table and the eye movement abnormality scoring table are traversed by using the single-dimensional head abnormality sequence, the single-dimensional limb abnormality sequence and the single-dimensional eye movement abnormality sequence mapping, and the single-dimensional abnormal cheating coefficient is obtained based on the sum of the traversal scoring results; Adding the associated abnormal cheating coefficient and the single-dimensional abnormal cheating coefficient to obtain a first comprehensive cheating coefficient; The first comprehensive cheating coefficient, the abnormal head sequence of the examination, the abnormal limb sequence of the examination, and the abnormal eye movement sequence of the examination are stored in association as the first cheating alarm information.

8. The intelligent operating system for pre-job training and examination for electric power construction units as claimed in claim 7, characterized in that: After updating the computing resource configuration of the second operating function block based on the first cheating alarm information, scheduling the second operating function block from the function block scheduling sequence includes: calling the first comprehensive cheating coefficient from the first cheating alarm information; Using the first comprehensive cheating coefficient to traverse the computing power allocation adjustment table to obtain first computing power update information; After the computing power resource configuration of the second operating functional block is updated by using the first computing power update information, the second operating functional block is scheduled from the functional block scheduling sequence.

9. The intelligent operating system for pre-job training and examination of electric power construction units as claimed in claim 7, characterized in that: Taking the multiple scheduling time windows as scheduling duration constraints and the functional block scheduling sequence as scheduling order constraints, monitoring the test user's test behavior of performing the multiple construction operation tasks, and outputting fusion cheating alarm information, including: Taking the multiple scheduling time windows as scheduling duration constraints and the functional block scheduling sequence as scheduling order constraints, real-time monitoring of the test user's test behavior in performing the multiple construction operation tasks is performed to obtain multiple comprehensive cheating coefficients, multiple test head abnormality sequences, multiple test limb abnormality sequences, and multiple test eye movement abnormality sequences; Performing weight configuration according to the execution difficulty of the multiple construction operation tasks to obtain multiple construction difficulty weights; Using the multiple construction difficulty weights to perform weighted calculation of the multiple comprehensive cheating coefficients to obtain a fusion cheating coefficient; According to the multiple test order identifier mappings, the multiple test head abnormality sequences, the multiple test limb abnormality sequences, and the multiple test eye movement abnormality sequences are spliced ​​to obtain a fused head abnormality sequence, a fused limb abnormality sequence, and a fused eye movement abnormality sequence; The fused cheating coefficient, the fused head abnormal sequence, the fused limb abnormal sequence and the fused eye movement abnormal sequence are stored in association to obtain the fused cheating alarm information.

Images