Lightning protection detection abnormity identification method and system

Abstract

The invention provides a lightning protection detection abnormity identification method and system, and relates to the technical field of lightning protection detection abnormity identification, and the method comprises the steps: responding to the starting of a lightning protection detection task, and obtaining pre-generated standard task data; actual detection track information of a detector is collected in real time through interaction between a trigger deployed at a detection point location and an intelligent helmet worn by the detector; collecting operation behavior data in real time through the intelligent helmet; comparing the actual detection track information with the standard task data, and judging whether space-time abnormity exists or not; based on the operation behavior data, judging whether an operation behavior is abnormal or not; and when the time-space abnormity or the operation behavior abnormity is identified, generating and outputting abnormity alarm information. According to the invention, the standardization level, the active safety prevention and control capability and the industry supervision efficiency of lightning protection detection operation can be comprehensively improved, and intelligent transformation of lightning protection detection from result supervision to process supervision is promoted.

Description

Technical Field

[0001] This invention relates to the field of lightning protection detection anomaly identification technology, and in particular to a lightning protection detection anomaly identification method and system. Background Technology

[0002] As lightning protection inspections shift from routine checks to intelligent and refined monitoring, the industry's need for real-time quality control of the inspection process is becoming increasingly urgent. Lightning protection involves high-risk scenarios such as high-rise buildings and power facilities, and the standardization of the inspection process is directly related to the timely detection of safety hazards. However, existing technologies mostly focus on the early planning of inspection tasks, lacking effective monitoring of the on-site operation behavior of inspection personnel, resulting in blind spots in process supervision and making it difficult to meet the requirements of modern disaster prevention for real-time performance and reliability.

[0003] Currently, the supervision of lightning protection testing mainly relies on manual spot checks or basic electronic methods. For example, arrival times of personnel are recorded using time clocks at testing sites, or on-site videos are captured by fixed cameras and then reviewed and analyzed by management personnel afterward. This method can only obtain limited spatiotemporal information (such as whether personnel were present), and cannot continuously track and intelligently judge the operational details of testing personnel (such as the standardization of tool use and data recording methods). The supervision efficiency is low and easily affected by subjective factors.

[0004] The core problem with existing technologies lies in their inability to achieve real-time and accurate identification of abnormal behavior during the detection process. Relying on discrete attendance data or non-intelligent video surveillance, the system struggles to automatically extract abnormal features (such as abnormal lingering at monitoring points or unauthorized handwritten records) from detection trajectories and operational behaviors, resulting in delayed anomaly detection and a high false alarm rate. This regulatory blind spot prevents timely intervention in detection quality risks, reducing the overall reliability of lightning protection detection. Summary of the Invention

[0005] Based on this, the present invention provides a lightning protection detection anomaly identification method and system, which aims to solve the technical problems described in the background art above.

[0006] A first aspect of the present invention is to provide a method for identifying anomalies in lightning protection detection, the method comprising: In response to the initiation of the lightning protection testing task, pre-generated standard task data is acquired, including the planned testing route, the planned time nodes for each testing point, and the testing sequence. Through the interaction between triggers deployed at the detection points and the smart helmets worn by the detection personnel, the actual detection trajectory information of the detection personnel is collected in real time. The actual detection trajectory information includes at least the actual time of arrival at the detection point, the duration of stay, and the movement path. The smart helmet collects operational behavior data in real time, including video stream data and motion posture data. The actual detection trajectory information is compared with the standard task data. If the deviation between the actual time and the planned time node exceeds the first preset threshold, or the deviation between the actual movement path and the planned detection route exceeds the second preset threshold, then a spatiotemporal anomaly is determined to exist. Based on the operational behavior data, the operation actions of the inspector are analyzed by a pre-trained behavior recognition model. If a preset violation is detected, it is determined that there is an abnormal operation behavior. When spatiotemporal anomalies or operational anomalies are detected, an anomaly alarm message is generated and output.

[0007] According to one aspect of the above technical solution, the step of comparing the actual detection trajectory information with the standard task data, and determining the existence of a spatiotemporal anomaly if the deviation between the actual time and the planned time node exceeds a first preset threshold, or the deviation between the actual movement path and the planned detection route exceeds a second preset threshold, includes: The actual time series and actual path series are extracted from the actual detection trajectory information, and the corresponding planned time series and planned path series are extracted from the standard task data, and the actual time series are aligned with the planned time series. Based on the aligned sequence, the time deviation and path deviation of each detection point are calculated, as well as the coefficient of variation of the moving speed between adjacent points, in order to identify abnormal acceleration or stagnation behavior. The time deviation, path deviation, and coefficient of variation of movement speed are compared with the first preset threshold, the second preset threshold, and the third preset threshold, respectively. If any indicator exceeds the corresponding threshold, it is determined that there is a spatiotemporal anomaly, and the anomaly is classified into time anomaly, path anomaly or behavioral rhythm anomaly according to the type of the exceeding indicator.

[0008] According to one aspect of the above technical solution, the steps of calculating the time deviation and path deviation of each detection point based on the aligned sequence, and calculating the coefficient of variation of the moving speed between adjacent points, to identify abnormal acceleration or stagnation behavior, include: For each detection point, the absolute difference between the actual arrival time and the planned arrival time is calculated, and the absolute difference is normalized relative to the total task duration to generate a dimensionless time deviation. For each detection point, the path similarity between the actual and planned geographic coordinates is calculated using a spatial geometric algorithm to generate a quantified path deviation. Based on the moving distance and time interval of adjacent points in the actual detection trajectory information, the moving speed of the detection personnel in each path segment is calculated, a moving speed sequence is constructed, and the coefficient of variation of the moving speed is calculated based on the speed sequence.

[0009] According to one aspect of the above technical solution, in the step of calculating the absolute difference between the actual arrival time and the planned arrival time for each detection point, and normalizing the absolute difference relative to the total task duration to generate a dimensionless time deviation, the calculation expression is as follows:

[0010] In the formula, Δ T deviation For time deviation, T actual The actual time when the testing personnel arrive at a specific testing location. T planned The time specified in the mission plan for arriving at this point. T total The total planned duration of the entire testing mission is the time span from the planned start time of the first testing point to the planned end time of the last testing point.

[0011] According to one aspect of the above technical solution, in the step of calculating the path similarity between the actual and planned geographic coordinates of each detection point using a spatial geometric algorithm to generate a quantified path deviation, the calculation expression is as follows:

[0012] In the formula, D deviation This refers to the path deviation. H ( P actual , P planned ) is the Hausdorff distance function, used to calculate the actual path point set. P actual With the planned path point set P planned The maximum and minimum distances between them L total This is the path normalization coefficient, which represents the total length of the planned detection route.

[0013] According to one aspect of the above technical solution, the expression for calculating the coefficient of variation of the moving speed based on the speed sequence is as follows:

[0014] In the formula, C v The coefficient of variation measures the dispersion of movement speed. σ v The standard deviation of the speed sequence of personnel moving between adjacent points. μ vThe mean of the speed sequence is the average moving speed.

[0015] According to one aspect of the above technical solution, based on the operational behavior data, the step of analyzing the operator's actions using a pre-trained behavior recognition model, and determining that there is an abnormal operational behavior if a preset violation is identified, includes: The video stream data and motion posture data collected in real time by the smart helmet are preprocessed; specifically, key frame extraction and image denoising are performed on the video stream data, and sliding window filtering and posture normalization are performed on the motion posture data. Hand motion trajectory features and tool grip features are extracted from the preprocessed video stream data, while head orientation angle features and body stability indicators are extracted from the motion posture data. The hand movement trajectory features, tool grip state features, head orientation angle features, and body stability indicators are fused at the feature level to generate a unified multimodal behavior feature vector. The multimodal behavior feature vector is input into a pre-trained behavior recognition model, which outputs the matching probability value between the current operation and various behaviors in the preset violation behavior database. If the matching probability value exceeds the preset probability threshold, it is determined that there is an abnormal operation behavior.

[0016] A second aspect of the present invention is to provide a lightning protection detection anomaly identification system, applied to the method described in the above-mentioned technical solution, the system comprising: The task acquisition module is used to acquire pre-generated standard task data in response to the start of the lightning protection detection task. The standard task data includes the planned detection route, the planned time nodes of each detection point, and the detection sequence. The trajectory acquisition module is used to collect the actual detection trajectory information of the inspectors in real time through the interaction between the triggers deployed at the detection points and the smart helmets worn by the inspectors. The actual detection trajectory information includes at least the actual time of arrival at the detection point, the duration of stay, and the movement path. The behavior acquisition module is used to collect operational behavior data in real time through the smart helmet. The operational behavior data includes video stream data and motion posture data. The spatiotemporal comparison module is used to compare the actual detection trajectory information with the standard task data. If the deviation between the actual time and the planned time node exceeds a first preset threshold, or the deviation between the actual movement path and the planned detection route exceeds a second preset threshold, then a spatiotemporal anomaly is determined to exist. The behavior comparison module is used to analyze the operator's actions based on the operation behavior data using a pre-trained behavior recognition model. If a preset violation is detected, it is determined that there is an abnormal operation behavior. The alarm output module is used to generate and output abnormal alarm information when spatiotemporal anomalies or abnormal operational behaviors are detected.

[0017] A third aspect of the present invention is to provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the methods described in the above-described technical solutions.

[0018] A fourth aspect of the present invention is to provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the methods described in the above-described technical solutions.

[0019] Compared with existing technologies, the lightning protection detection anomaly identification method and system shown in this invention have the following advantages: This invention, based on the collaborative working mechanism of triggers and smart helmets, achieves seamless, end-to-end data collection of the spatiotemporal trajectories and operational behaviors of inspection personnel. It overcomes the shortcomings of traditional time clocks and fixed cameras, such as blind spots and data fragmentation, laying a solid foundation for precise supervision. By performing multi-dimensional quantitative comparisons of actual inspection trajectories and standard task data, and combining this with in-depth analysis using a pre-trained behavior recognition model, it effectively solves the problems of high false alarm rates and delayed detection caused by traditional reliance on manual judgment. This shifts anomaly identification from experience-based judgment to data-driven scientific decision-making, significantly improving the accuracy and timeliness of detecting violations. In summary, this invention comprehensively enhances the standardization level, proactive safety control capabilities, and industry regulatory efficiency of lightning protection inspection operations, promoting the intelligent transformation of lightning protection inspection from result-based supervision to process-based supervision. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the implementation of a lightning protection detection anomaly identification method provided in an embodiment of the present invention. Figure 2 This is a structural block diagram of a lightning protection detection anomaly identification system provided in an embodiment of the present invention. Detailed Implementation

[0021] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0022] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0024] Example 1 Please see Figure 1 The first embodiment of the present invention provides a method for identifying anomalies in lightning protection detection, the method comprising steps S10-S60: Step S10: In response to the start of the lightning protection detection task, obtain the pre-generated standard task data, which includes the planned detection route, the planned time nodes of each detection point, and the detection sequence.

[0025] In this embodiment, by responding to the initiation of the lightning protection testing task, pre-generated standard task data is automatically acquired. This transforms the abstract testing task into a quantifiable execution standard, ensuring that the subsequent anomaly identification process has an objective basis for comparison. Without this benchmark, it would be impossible to effectively distinguish between normal operating procedures and genuine violations, such as determining whether testing personnel have arbitrarily changed the order of testing points or significantly deviated from the planned timeline.

[0026] The above relies on the system's dynamic response capability to the start signal of the lightning protection detection task. When the system detects the command issued by the task management platform or the activation of devices such as the smart helmet worn by the testing personnel, it will automatically trigger the data acquisition process. Standard task data is usually called from a preset cloud server or local database. Its source is often an independent lightning protection detection task generation system, which ensures the authority and rationality of the data.

[0027] The standard task data itself is a multi-dimensional, structured collection of information, rather than a simple list of locations. It comprises three key elements: the planned inspection route, the planned timeframes for each inspection location, and a strict inspection sequence. The planned inspection route is an optimized spatial sequence generated based on electronic maps and path planning algorithms, comprehensively considering distance, efficiency, and risk avoidance. The planned timeframes set the expected arrival and operation time for each location, and their calculation incorporates historical data and operational complexity analysis. The inspection sequence defines the access logic for each location, which is crucial for ensuring the rigor of the inspection process. These three elements together constitute a complete standard operating procedure.

[0028] Step S20: Through the interaction between the trigger deployed at the detection point and the smart helmet worn by the detection personnel, the actual detection trajectory information of the detection personnel is collected in real time. The actual detection trajectory information includes at least the actual time of arrival at the detection point, the duration of stay, and the movement path.

[0029] Specifically, low-power triggers, such as Bluetooth Low Energy beacons or RFID tags, are pre-deployed at each detection point, with each trigger having a unique location identification code. The smart helmets worn by the inspectors have built-in signal receiving modules (such as Bluetooth scanners), high-precision clock chips, and positioning modules (such as GPS / BeiDou). When an inspector arrives at a detection point and their helmet enters the signal coverage area of ​​a trigger, the trigger automatically broadcasts its location ID. The receiving module on the smart helmet captures this signal and simultaneously records the precise current timestamp.

[0030] Based on the discrete triggering events collected using the aforementioned mechanism, continuous actual detection trajectory information is generated through data fusion and trajectory reconstruction algorithms. First, by analyzing the timestamp sequence and location ID sequence, the dwell time of the inspector at each location can be accurately calculated, i.e., the difference between the departure and arrival timestamps at the current location. Second, by combining the geographical coordinates periodically reported by the smart helmet's built-in positioning module and performing path interpolation in chronological order, the movement path of the inspector between locations can be reconstructed. This path not only reflects the inspector's route but its coordinate sequence, combined with timestamps, can also be used to calculate movement speed, providing a data foundation for subsequent identification of abnormal acceleration or lingering behavior.

[0031] Step S30: Real-time collection of operation behavior data through the smart helmet, including video stream data and motion posture data.

[0032] The smart helmet, as the core carrier of data collection, relies on its integrated sensing system, which forms the technological foundation for this step. A high-definition wide-angle camera located at the front of the helmet continuously acquires video stream data from a first-person perspective, fully recording the operator's hand movements, instrument interface readings, and the surrounding environment. Simultaneously, a built-in high-precision inertial measurement unit continuously captures head motion data such as three-axis acceleration and angular velocity, accurately reflecting changes in head posture and movement frequency during the detection process. After preliminary preprocessing and time synchronization by the edge computing module, this raw data is transmitted in real-time to the backend analysis platform via a high-speed wireless network, ensuring data continuity and low latency.

[0033] Specifically, the behavioral data collected by smart helmets is a multi-dimensional and complementary dataset. Video stream data provides the most direct visual evidence, primarily aimed at identifying explicitly defined violations such as handwritten records. It can also be used to assist in determining whether tools are used correctly and whether safety procedures are followed. Motion posture data provides supplementary information from a behavioral dynamics perspective. For example, analyzing the frequency of head pitch changes can infer attention levels, and sudden turning movements can indicate potential abnormalities. In short, video data provides contextual explanations for posture changes, while posture data adds objective quantitative indicators to behavioral judgments within the video, effectively improving the accuracy and robustness of subsequent behavior recognition.

[0034] Step S40: Compare the actual detection trajectory information with the standard task data. If the deviation between the actual time and the planned time node exceeds the first preset threshold, or the deviation between the actual movement path and the planned detection route exceeds the second preset threshold, then it is determined that there is a spatiotemporal anomaly.

[0035] In this embodiment, the step of comparing the actual detection trajectory information with the standard task data, and determining the existence of a spatiotemporal anomaly if the deviation between the actual time and the planned time node exceeds a first preset threshold, or the deviation between the actual movement path and the planned detection route exceeds a second preset threshold, includes: The actual time series and actual path series are extracted from the actual detection trajectory information, and the corresponding planned time series and planned path series are extracted from the standard task data, and the actual time series are aligned with the planned time series. Based on the aligned sequence, the time deviation and path deviation of each detection point are calculated, as well as the coefficient of variation of the moving speed between adjacent points, in order to identify abnormal acceleration or stagnation behavior. The time deviation, path deviation, and coefficient of variation of movement speed are compared with the first preset threshold, the second preset threshold, and the third preset threshold, respectively. If any indicator exceeds the corresponding threshold, it is determined that there is a spatiotemporal anomaly, and the anomaly is classified into time anomaly, path anomaly or behavioral rhythm anomaly according to the type of the exceeding indicator.

[0036] Furthermore, based on the aligned sequence, the steps of calculating the time deviation and path deviation of each detection point, and calculating the coefficient of variation of movement speed between adjacent points, to identify abnormal acceleration or stagnation behavior, include: For each detection point, the absolute difference between the actual arrival time and the planned arrival time is calculated, and the absolute difference is normalized relative to the total task duration to generate a dimensionless time deviation. For each detection point, the path similarity between the actual and planned geographic coordinates is calculated using a spatial geometric algorithm to generate a quantified path deviation. Based on the moving distance and time interval of adjacent points in the actual detection trajectory information, the moving speed of the detection personnel in each path segment is calculated, a moving speed sequence is constructed, and the coefficient of variation of the moving speed is calculated based on the speed sequence.

[0037] In the step of calculating the absolute difference between the actual arrival time and the planned arrival time for each detection point, and then normalizing the absolute difference relative to the total task duration to generate a dimensionless time deviation, the calculation expression is as follows:

[0038] In the formula, Δ T deviation For time deviation, T actual The actual time when the testing personnel arrive at a specific testing location. T planned The time specified in the mission plan for arriving at this point. T total The total planned duration of the entire testing mission is the time span from the planned start time of the first testing point to the planned end time of the last testing point.

[0039] In the step of calculating the path similarity between the actual and planned geographic coordinates of each detection point using a spatial geometric algorithm to generate a quantified path deviation, the calculation expression is as follows:

[0040] In the formula, D deviation This refers to the path deviation. H ( P actual , P planned) is the Hausdorff distance function, used to calculate the actual path point set. P actual With the planned path point set P planned The maximum and minimum distances between them L total This is the path normalization coefficient, which represents the total length of the planned detection route.

[0041] The expression for calculating the coefficient of variation of the moving speed based on the speed sequence is as follows:

[0042] In the formula, C v The coefficient of variation measures the dispersion of movement speed. σ v The standard deviation of the speed sequence of personnel moving between adjacent points. μ v The mean of the speed sequence is the average moving speed.

[0043] Specifically, the method shown in this embodiment establishes a comparable data foundation through sequence alignment and feature extraction. The actual time series and actual path series extracted from the actual detection trajectory information are not simply compared directly with the planned sequence in the standard task data. By employing algorithms such as dynamic time warping to align the time series data, the minor time scaling effects caused by fluctuations in the normal work rhythm are eliminated, ensuring the fairness of subsequent comparisons.

[0044] At the indicator calculation level, the method shown in this embodiment employs a multi-dimensional, normalized scientific measurement approach. The time deviation metric uses a relative percentage algorithm to convert the absolute time difference into a dimensionless ratio related to the total task duration, eliminating the differences in judgment criteria between long and short tasks. The path deviation metric introduces the Hausdorff distance function, which effectively captures the similarity of the overall path shape. Compared to simple point-to-point Euclidean distance, it is better able to identify hidden anomalies such as "traveling all points but taking a completely wrong route." The coefficient of variation of movement speed measures the dispersion of speed fluctuations using statistical methods. Its dimensionless nature allows it to effectively identify abnormal acceleration or stagnation behaviors across tasks and personnel.

[0045] The final anomaly determination employs dynamic threshold comparison and classification. The calculated indicators are not compared to fixed thresholds, but rather to preset thresholds (first, second, and third preset thresholds) dynamically adjusted based on factors such as task environment complexity and historical data. This adaptive mechanism significantly improves the accuracy of the determination. When any indicator exceeds its corresponding threshold, a spatiotemporal anomaly is determined, and the anomaly can be further subdivided into subcategories such as time anomalies, path anomalies, or behavioral rhythm anomalies based on the type of the exceeding indicator (e.g., only time deviation exceeds the standard, or multiple indicators exceed the standard simultaneously). This classification result not only triggers alarms but also provides targeted decision support for subsequent management interventions. For example, time anomalies may only require recording, while path anomalies, if occurring in dangerous areas, require immediate intervention.

[0046] Based on the multi-dimensional quantitative analysis process, the automatic and accurate identification of spatiotemporal anomalies in lightning protection testing operations has been achieved, elevating the traditional qualitative supervision that relies on human experience to a new level of data-driven intelligent supervision.

[0047] Step S50: Based on the operation behavior data, the operation actions of the detection personnel are analyzed by a pre-trained behavior recognition model. If a preset violation is detected, it is determined that there is an abnormal operation behavior.

[0048] In this embodiment, based on the operational behavior data, the step of analyzing the operator's actions using a pre-trained behavior recognition model, and determining that there is an abnormal operational behavior if a preset violation is identified, includes: The video stream data and motion posture data collected in real time by the smart helmet are preprocessed; specifically, key frame extraction and image denoising are performed on the video stream data, and sliding window filtering and posture normalization are performed on the motion posture data. Hand motion trajectory features and tool grip features are extracted from the preprocessed video stream data, while head orientation angle features and body stability indicators are extracted from the motion posture data. The hand movement trajectory features, tool grip state features, head orientation angle features, and body stability indicators are fused at the feature level to generate a unified multimodal behavior feature vector. The multimodal behavior feature vector is input into a pre-trained behavior recognition model, which outputs the matching probability value between the current operation and various behaviors in the preset violation behavior database. If the matching probability value exceeds the preset probability threshold, it is determined that there is an abnormal operation behavior.

[0049] Specifically, the aforementioned behavior recognition model is based on a deep learning architecture, such as a combination of convolutional neural networks (CNN) and long short-term memory networks (LSTM). Before implementation, the model needs to be trained on a large amount of labeled data, including normal operation video clips and various violation behavior samples, such as handwritten records and unauthorized touching of devices. The training process enables the model to learn the spatiotemporal features of the operation.

[0050] In practical applications, the aforementioned behavior recognition model performs multimodal fusion analysis on the input video stream data and motion posture data. First, visual features such as hand key point trajectories and tool holding states are extracted from the video stream; simultaneously, dynamic features such as head orientation and body stability are analyzed from the motion posture data; then, through feature-level or decision-level fusion, the model comprehensively determines whether the current operation sequence matches a preset violation behavior pattern.

[0051] The definition of pre-defined violations is based on safety regulations and standard operating procedures for lightning protection testing. Typical violations include, but are not limited to: handwritten recording of test data (instead of using electronic devices for input), failure to operate testing instruments according to procedures, failure to wear safety equipment in hazardous areas, and prolonged absence from the work station. In terms of specific judgment logic, the behavior recognition model outputs a matching probability value between the current operation and the violation database. If the probability value exceeds a pre-defined confidence threshold, an abnormal operation is determined.

[0052] Step S60: When a spatiotemporal anomaly or operational anomaly is detected, generate and output an anomaly alarm message.

[0053] In this embodiment, once a spatiotemporal anomaly or operational anomaly is confirmed, the core metadata of the anomaly event is automatically extracted to construct a complete alarm information object containing fields such as anomaly identifier, type, level, details, associated context, and suggested measures. This not only ensures the integrity of the information but also provides ample data support for subsequent source tracing analysis and processing decisions. For example, for high-risk anomalies, visual evidence such as on-site collected trajectory screenshots or behavioral recognition keyframes will be dynamically attached, making the alarm content more intuitive and credible.

[0054] The alarm information output adopts an intelligent routing strategy, selecting the optimal push channel based on the anomaly level and preset rules. For high-risk anomalies requiring immediate response, alarm information is pushed in real time to the large screen of the back-end monitoring center and the mobile terminals of on-site management personnel through message queues or instant messaging interfaces; while all anomaly records are synchronized to the cloud database to form audit logs, providing a data foundation for quality traceability and statistical analysis.

[0055] Compared with existing technologies, the lightning protection detection anomaly identification method shown in this embodiment has the following advantages: This embodiment, based on the collaborative working mechanism of triggers and smart helmets, achieves seamless, end-to-end data collection of the spatiotemporal trajectories and operational behaviors of inspection personnel. It overcomes the shortcomings of traditional time clocks and fixed cameras, such as blind spots and data fragmentation, laying a solid foundation for precise supervision. By performing multi-dimensional quantitative comparisons of actual inspection trajectories and standard task data, and combining this with in-depth analysis using a pre-trained behavior recognition model, it effectively solves the problems of high false alarm rates and delayed detection caused by traditional reliance on manual judgment. This shifts anomaly identification from experience-based judgment to data-driven scientific decision-making, significantly improving the accuracy and timeliness of detecting violations. In summary, this embodiment comprehensively enhances the standardization level, proactive safety control capabilities, and industry regulatory efficiency of lightning protection inspection operations, promoting the intelligent transformation of lightning protection inspection from result-based supervision to process-based supervision.

[0056] Example 2 Please see Figure 2 The second embodiment of the present invention provides a lightning protection detection anomaly identification system, applied to the method described in the above embodiments, the system comprising: The task acquisition module 10 is used to acquire pre-generated standard task data in response to the start of the lightning protection detection task. The standard task data includes the planned detection route, the planned time nodes of each detection point, and the detection sequence. The trajectory acquisition module 20 is used to collect the actual detection trajectory information of the inspector in real time through the interaction between the trigger deployed at the detection point and the smart helmet worn by the inspector. The actual detection trajectory information includes at least the actual time of arrival at the detection point, the duration of stay, and the movement path. The behavior acquisition module 30 is used to collect operation behavior data in real time through the smart helmet. The operation behavior data includes video stream data and motion posture data. The spatiotemporal comparison module 40 is used to compare the actual detection trajectory information with the standard task data. If the deviation between the actual time and the planned time node exceeds the first preset threshold, or the deviation between the actual movement path and the planned detection route exceeds the second preset threshold, then it is determined that there is a spatiotemporal anomaly. The behavior comparison module 50 is used to analyze the operation actions of the detection personnel based on the operation behavior data through a pre-trained behavior recognition model. If a preset violation behavior is detected, it is determined that there is an abnormal operation behavior. The alarm output module 60 is used to generate and output abnormal alarm information when spatiotemporal anomalies or abnormal operational behaviors are detected.

[0057] Compared with existing technologies, the lightning protection detection anomaly identification system shown in this embodiment has the following advantages: This embodiment, based on the collaborative working mechanism of triggers and smart helmets, achieves seamless, end-to-end data collection of the spatiotemporal trajectories and operational behaviors of inspection personnel. It overcomes the shortcomings of traditional time clocks and fixed cameras, such as blind spots and data fragmentation, laying a solid foundation for precise supervision. By performing multi-dimensional quantitative comparisons of actual inspection trajectories and standard task data, and combining this with in-depth analysis using a pre-trained behavior recognition model, it effectively solves the problems of high false alarm rates and delayed detection caused by traditional reliance on manual judgment. This shifts anomaly identification from experience-based judgment to data-driven scientific decision-making, significantly improving the accuracy and timeliness of detecting violations. In summary, this embodiment comprehensively enhances the standardization level, proactive safety control capabilities, and industry regulatory efficiency of lightning protection inspection operations, promoting the intelligent transformation of lightning protection inspection from result-based supervision to process-based supervision.

[0058] Example 4 The fourth embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the above-described technical solution.

[0059] The fifth embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the methods described in the above technical solutions.

[0060] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0061] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for identifying anomalies in lightning protection detection, characterized in that, The method includes: In response to the initiation of the lightning protection testing task, pre-generated standard task data is acquired, including the planned testing route, the planned time nodes for each testing point, and the testing sequence. Through the interaction between triggers deployed at the detection points and the smart helmets worn by the detection personnel, the actual detection trajectory information of the detection personnel is collected in real time. The actual detection trajectory information includes at least the actual time of arrival at the detection point, the duration of stay, and the movement path. The smart helmet collects operational behavior data in real time, including video stream data and motion posture data. The actual detection trajectory information is compared with the standard task data. If the deviation between the actual time and the planned time node exceeds the first preset threshold, or the deviation between the actual movement path and the planned detection route exceeds the second preset threshold, then a spatiotemporal anomaly is determined to exist. Based on the operational behavior data, the operation actions of the inspector are analyzed by a pre-trained behavior recognition model. If a preset violation is detected, it is determined that there is an abnormal operation behavior. When spatiotemporal anomalies or operational anomalies are detected, an anomaly alarm message is generated and output.

2. The lightning protection detection anomaly identification method according to claim 1, characterized in that, The step of comparing the actual detection trajectory information with the standard task data, and determining that a spatiotemporal anomaly exists if the deviation between the actual time and the planned time node exceeds a first preset threshold, or the deviation between the actual movement path and the planned detection route exceeds a second preset threshold, includes: The actual time series and actual path series are extracted from the actual detection trajectory information, and the corresponding planned time series and planned path series are extracted from the standard task data, and the actual time series are aligned with the planned time series. Based on the aligned sequence, the time deviation and path deviation of each detection point are calculated, as well as the coefficient of variation of the moving speed between adjacent points, in order to identify abnormal acceleration or stagnation behavior. The time deviation, path deviation, and coefficient of variation of movement speed are compared with the first preset threshold, the second preset threshold, and the third preset threshold, respectively. If any indicator exceeds the corresponding threshold, it is determined that there is a spatiotemporal anomaly, and the anomaly is classified into time anomaly, path anomaly or behavioral rhythm anomaly according to the type of the exceeding indicator.

3. The lightning protection detection anomaly identification method according to claim 2, characterized in that, Based on the aligned sequence, the steps for identifying abnormal acceleration or stagnation behavior include calculating the time deviation and path deviation of each detection point, and calculating the coefficient of variation of movement speed between adjacent points. For each detection point, the absolute difference between the actual arrival time and the planned arrival time is calculated, and the absolute difference is normalized relative to the total task duration to generate a dimensionless time deviation. For each detection point, the path similarity between the actual and planned geographic coordinates is calculated using a spatial geometric algorithm to generate a quantified path deviation. Based on the moving distance and time interval of adjacent points in the actual detection trajectory information, the moving speed of the detection personnel in each path segment is calculated, a moving speed sequence is constructed, and the coefficient of variation of the moving speed is calculated based on the speed sequence.

4. The lightning protection detection anomaly identification method according to claim 3, characterized in that, In the step of calculating the absolute difference between the actual arrival time and the planned arrival time for each detection point, and normalizing the absolute difference relative to the total task duration to generate a dimensionless time deviation, the calculation expression is as follows: In the formula, Δ T deviation For time deviation, T actual The actual time when the testing personnel arrive at a specific testing location. T planned The time specified in the mission plan for arriving at this point. T total The total planned duration of the entire testing mission is the time span from the planned start time of the first testing point to the planned end time of the last testing point.

5. The lightning protection detection anomaly identification method according to claim 3, characterized in that, In the step of calculating the path similarity between the actual and planned geographic coordinates of each detection point using a spatial geometric algorithm to generate a quantified path deviation, the calculation expression is as follows: In the formula, D deviation This represents the path deviation. H ( P actual , P planned ) is the Hausdorff distance function, used to calculate the actual path point set. P actual With the planned path point set P planned The maximum and minimum distances between them L total This is the path normalization coefficient, which represents the total length of the planned detection route.

6. The lightning protection detection anomaly identification method according to claim 3, characterized in that, The expression for calculating the coefficient of variation of the moving speed based on the aforementioned speed sequence is as follows: In the formula, C v The coefficient of variation measures the dispersion of movement speed. σ v The standard deviation of the speed sequence of personnel moving between adjacent points. μ v The mean of the speed sequence is the average moving speed.

7. The lightning protection detection anomaly identification method according to any one of claims 1-6, characterized in that, Based on the operational behavior data, the steps of analyzing the operator's actions using a pre-trained behavior recognition model, and determining that there is an abnormal operational behavior if a preset violation is identified, include: The video stream data and motion posture data collected in real time by the smart helmet are preprocessed; specifically, key frame extraction and image denoising are performed on the video stream data, and sliding window filtering and posture normalization are performed on the motion posture data. Hand motion trajectory features and tool grip features are extracted from the preprocessed video stream data, while head orientation angle features and body stability indicators are extracted from the motion posture data. The hand movement trajectory features, tool grip state features, head orientation angle features, and body stability indicators are fused at the feature level to generate a unified multimodal behavior feature vector. The multimodal behavior feature vector is input into a pre-trained behavior recognition model, which outputs the matching probability value between the current operation and various behaviors in the preset violation behavior database. If the matching probability value exceeds the preset probability threshold, it is determined that there is an abnormal operation behavior.

8. A lightning protection detection anomaly identification system, characterized in that, The system, applicable to the method of any one of claims 1-7, comprises: The task acquisition module is used to acquire pre-generated standard task data in response to the start of the lightning protection detection task. The standard task data includes the planned detection route, the planned time nodes of each detection point, and the detection sequence. The trajectory acquisition module is used to collect the actual detection trajectory information of the inspectors in real time through the interaction between the triggers deployed at the detection points and the smart helmets worn by the inspectors. The actual detection trajectory information includes at least the actual time of arrival at the detection point, the duration of stay, and the movement path. The behavior acquisition module is used to collect operational behavior data in real time through the smart helmet. The operational behavior data includes video stream data and motion posture data. The spatiotemporal comparison module is used to compare the actual detection trajectory information with the standard task data. If the deviation between the actual time and the planned time node exceeds a first preset threshold, or the deviation between the actual movement path and the planned detection route exceeds a second preset threshold, then a spatiotemporal anomaly is determined to exist. The behavior comparison module is used to analyze the operator's operation actions based on the operation behavior data using a pre-trained behavior recognition model. If a preset violation behavior is identified, it is determined that there is an abnormal operation behavior. The alarm output module is used to generate and output abnormal alarm information when spatiotemporal anomalies or abnormal operational behaviors are detected.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.

10. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any one of claims 1-7.

Images