Identity identification method for violators for field safety supervision of gas turbine power plant

Abstract

The invention discloses a violation person identity recognition method for gas turbine power plant on-site safety supervision, relates to the technical field of person identity recognition, and aims at recognizing specific violation behaviors, positioning the occurrence time and space of the violation behaviors and marking corresponding behavior subjects. The method comprises the following steps: acquiring operation image data, extracting behavior subject image features and equipment acquisition data, generating fusion identity information, matching violation identities, outputting structured records, evaluating violation risk levels, verifying the identities of persons in charge and filing violation records, and automatically identifying dangerous behaviors in a gas turbine power plant through the operation image data acquisition and violation behavior identification model. And in combination with the dynamic attitude vector sequence and equipment acquisition data, generating identity feature fusion data, performing map matching, identifying the identity of a violation person, labeling through a risk level label, pushing to a responsible person terminal, completing identity confirmation and violation behavior recording and archiving, and realizing whole-process supervision and closed-loop management.

Description

Technical Field

[0001] This invention relates to the field of personnel identification technology, and in particular to a method for identifying violators in on-site safety supervision of gas turbine power plants. Background Technology

[0002] As high-risk industrial sites, gas turbine power plants place extremely high demands on on-site operational safety management. Existing safety supervision methods primarily rely on video surveillance and manual inspections, which are significantly inadequate for real-time detection, precise location, and identification of responsible personnel for violations. On the one hand, traditional monitoring systems cannot automatically identify and classify worker behavior, leading to delayed detection or even missed detection of violations. On the other hand, even when suspected violations are identified, it is difficult to accurately match the specific responsible personnel in complex scenarios such as image obstruction or multiple workers operating simultaneously.

[0003] Currently, Chinese patent application number CN202210787096.5 discloses a method and system for identifying personnel identity and safety violations. The method includes: acquiring images of engineering personnel entering the construction area and their operational processes during construction; identifying the engineering personnel based on the acquired images and identifying violations based on the acquired operational processes; if the engineering personnel have violated regulations, generating a violation record, sending the violation record to the engineering personnel and the construction project management personnel, and simultaneously sending the correct operating procedure matching the violation record to the engineering personnel.

[0004] The relevant technologies are insufficient for accurately identifying violations and for real-time matching and closed-loop confirmation of the responsible parties in multi-person operations and complex scenarios. Summary of the Invention

[0005] The technical problem solved by this invention is that existing technologies are difficult to achieve accurate identification of violations and real-time matching and closed-loop confirmation of the responsible person's identity in multi-person operations and complex scenarios.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] A method for identifying violators in on-site safety supervision at gas turbine power plants includes the following steps:

[0008] Step S1: Collect operational image data of key areas of the gas turbine power plant and identify specific violations;

[0009] Step S2: Locate the time and place where the violation occurred, and mark the corresponding subject of the violation;

[0010] Step S3: Extract the image features of the subject of the behavior and the data collected by the device to generate fused identity information;

[0011] Step S4: Match the violator's identity based on the graph model and output a structured record;

[0012] Step S5: Assess the level of violation risk and push the information to the management platform and the responsible person's terminal;

[0013] Step S6: Verify the identity of the responsible person and file the violation record.

[0014] Preferably, step S1 includes the following sub-steps:

[0015] Step S101: Collect operation image data. The collection points for the operation image data include the main passage of the gas turbine power plant, entrances and exits, areas near high-voltage equipment, and designated hazardous operation points.

[0016] Step S102: After preprocessing the work image data in the terminal, the preset violation behavior recognition model is called for real-time recognition. The violation behavior recognition model is based on a multi-scale feature fusion network to identify high-risk behaviors, including not wearing a safety helmet, crossing the isolation line and carrying a fire source. The first recognition result is output.

[0017] Step S103: Calculate the confidence score of the first identification result. If the confidence score is higher than the preset threshold, mark it as a suspected violation event and store the image frame number and region label of the corresponding operation image data into the violation event cache queue.

[0018] Preferably, step S2 includes the following sub-steps:

[0019] Step S201: Extract the timestamps, anchor point positions, and background reference information of the image frames of the work image data marked as suspected violations. Through image feature comparison and calibration coordinate transformation algorithms, calculate the precise time node of the behavior and the spatial coordinates in the work site, and output the violation behavior location data.

[0020] Step S202: Based on the location data of the violation, retrieve continuous image frames for the corresponding time period, use a multi-target tracking algorithm to extract the activity trajectories of all visible workers, and mark the person whose activity trajectory is closest to the location data of the violation in time and space as the subject of the behavior, and output the individual labeled image fragment.

[0021] Preferably, step S3 includes the following sub-steps:

[0022] Step S301: Locate the outline region of the main body from the individual labeled image fragments, extract the clothing color and texture features, body shape and action posture parameters of the main body outline region, and construct a dynamic posture vector sequence based on the time-continuous image.

[0023] Step S302: Read the identification code of the outline area of ​​the main body of the behavior, and at the same time collect the corresponding three-axis acceleration, angular velocity and attitude angle sequence data. Combine the identification code, three-axis acceleration, angular velocity and attitude angle sequence data and output them as the data collected by the device.

[0024] Step S303: The dynamic posture vector sequence and the data collected by the device are aligned on a unified time axis using a feature fusion algorithm, and multi-dimensional normalization is performed based on time consistency and action shape matching degree to generate single individual identity feature fusion data.

[0025] Preferably, step S4 includes the following sub-steps:

[0026] Step S401: Using the location data of the violation as a reference, the timestamp, spatial coordinates and individual identity features of the outline area of ​​the subject of the violation are associated to establish a three-element behavior node. By modeling the trend of individual state changes in the continuous operation image sequence, a behavior identity map is formed.

[0027] Step S402: Perform similarity measurement calculation and node matching on the behavioral identity graph. Combine time alignment, spatial overlap and action consistency scores, and use a weighted comprehensive mechanism to determine the optimal matching identity code. Bind the optimal matching identity code to the corresponding violation behavior and output a structured violation behavior record. The structured violation behavior record includes the identity code, violation time, violation location and violation behavior type.

[0028] Preferably, in step S4, when the similarity metric is lower than a set value, the trajectory reconstruction-assisted recognition process is initiated.

[0029] Preferably, step S5 includes the following sub-steps:

[0030] Step S501: Based on the type of violation, time of violation, and risk level of the location of violation in the structured violation record, and combined with the power plant safety management rule base, execute the risk level assessment algorithm to generate risk level labels for the violation. The risk level labels are divided into general risk, major risk, and serious risk.

[0031] Step S502: The structured violation record with risk level label attached, along with the corresponding identification code, violation time and location, is uploaded to the power plant safety management platform through a data encryption mechanism.

[0032] Step S503: Simultaneously, the summary information of the violation record is pushed to the mobile terminal of the person in charge through a preset communication interface. The summary information of the violation record includes the risk level, the type of violation, and the identity code, triggering the prompt and processing confirmation process, and building a rapid response mechanism for violation information.

[0033] Preferably, step S6 includes the following sub-steps:

[0034] Step S601: Receive the identity tag feedback information pushed by the responsible person's terminal, and verify the identity information by reading the identity information in the NFC chip of the work badge. The verification content includes the identity number, work group number and attendance record of the day. If the match is consistent, the identity confirmation is completed.

[0035] Step S602: The confirmed violation records are structured and archived, and the violation record archive data is output. The violation record archive data includes the identity identification code, violation type, violation time, risk level and location coordinates. At the same time, the work trajectory data of the day is called and the archived records are bound with the behavior trajectory data within the corresponding time period. The archived records are stored in the safety supervision database as personnel behavior audit records.

[0036] Preferably, the archived data of the violation records is synchronously transmitted to the safety training system and incorporated into the typical case feedback database.

[0037] Preferably, the fused data of individual identity features is used to construct a risk profile of the operator.

[0038] The beneficial effects of this invention are as follows: This invention automatically identifies dangerous behaviors in gas turbine power plants by collecting operational image data and using a violation behavior recognition model. By combining dynamic posture vector sequences and equipment data, it generates identity feature fusion data and performs map matching to identify the violators. The data is then labeled with risk level tags and pushed to the responsible person's terminal to complete identity confirmation and violation behavior record archiving, thereby achieving full-process supervision and closed-loop management. Attached Figure Description

[0039] Figure 1 The flowchart illustrates the steps of a method for identifying violators in on-site safety supervision of gas turbine power plants, as provided in one embodiment of the present invention. Detailed Implementation

[0040] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0041] Example, refer to Figure 1 This paper provides a method for identifying violators in on-site safety supervision of gas turbine power plants, comprising the following steps:

[0042] Step S1: Collect operational image data of key areas of the gas turbine power plant and identify specific violations.

[0043] Step S2: Locate the time and place where the violation occurred, and mark the corresponding subject of the violation.

[0044] Step S3: Extract the image features of the subject of the behavior and the data collected by the device to generate fused identity information.

[0045] Step S4: Match the identity of the violator based on the graph model and output a structured record.

[0046] Step S5: Assess the level of violation risk and push the information to the management platform and the responsible person's terminal.

[0047] Step S6: Verify the identity of the responsible person and file the violation record.

[0048] Step S1 includes the following sub-steps:

[0049] Step S101: Collect operation image data. The collection points for operation image data include the main passage of the gas turbine power plant, entrances and exits, areas near high-voltage equipment, and designated hazardous operation points.

[0050] Step S101 completes the acquisition of operational image data for the main passage, entrances and exits, high-voltage equipment areas and hazardous operation points of the gas turbine power plant, ensuring coverage of key operation areas and providing clear and continuous image information sources.

[0051] Step S102: After preprocessing the operation image data in the terminal, the preset violation behavior recognition model is called for real-time recognition. The violation behavior recognition model identifies high-risk behaviors based on a multi-scale feature fusion network. High-risk behaviors include not wearing a safety helmet, crossing the isolation line, and carrying a fire source. The first recognition result is output.

[0052] Step S102 uses a preset multi-scale feature fusion recognition model to perform real-time analysis on the acquired images, identify typical high-risk behaviors such as not wearing a safety helmet, crossing the isolation line, and carrying a fire source, and initially screen out suspected violation image frames.

[0053] Step S103: Calculate the confidence score of the first identification result. If the confidence score is higher than the preset threshold, mark it as a suspected violation event and store the image frame number and region label of the corresponding operation image data into the violation event cache queue.

[0054] Step S103 assesses the confidence level of the recognition results, automatically marks high-confidence violations as suspected violations, and records the image frame number and region label to provide a basis for subsequent behavior localization and trajectory tracking.

[0055] Step S1 enables the continuous acquisition of operational image data in key areas of the gas turbine power plant and the real-time identification of violations, providing an image basis and initial event screening results for subsequent positioning and identification.

[0056] Step S2 includes the following sub-steps:

[0057] Step S201: Extract the timestamps, anchor point positions, and background reference information of the image frames of the work image data marked as suspected violations. Through image feature comparison and calibration coordinate transformation algorithms, calculate the precise time node of the behavior and the spatial coordinates in the work site, and output the violation behavior location data.

[0058] Step S201 extracts the timestamp, anchor point, and background reference information of the suspected violation image frame, and combines image feature comparison and coordinate transformation algorithms to calculate the time of occurrence of the violation and its specific spatial coordinates in the work site, thus forming violation location data.

[0059] Step S202: Based on the location data of the violation, retrieve continuous image frames for the corresponding time period, use a multi-target tracking algorithm to extract the activity trajectories of all visible workers, and mark the person whose activity trajectory is closest to the location data of the violation in time and space as the subject of the behavior, and output the individual labeled image fragment.

[0060] Step S202 retrieves continuous image frames based on the positioning data, uses a multi-target tracking algorithm to obtain the activity trajectory of all workers, and determines the person closest to the violation as the subject of the behavior through spatiotemporal matching, extracts and outputs their image fragments for subsequent identity recognition.

[0061] Step S2 achieves precise location of suspected violations in terms of time and space, and extracts the activity trajectory of the subject of the violation by combining the image sequence, thereby accurately marking the location of the violator in the image.

[0062] Step S3 includes the following sub-steps:

[0063] Step S301: Locate the outline region of the main body from the individual labeled image fragments, extract the clothing color and texture features, body shape and action posture parameters of the main body outline region, and construct a dynamic posture vector sequence based on the time-continuous image.

[0064] Step S301 extracts the clothing color, texture features, body shape and action posture parameters of the subject from the individual labeled image fragments, and generates a dynamic posture vector sequence based on continuous image sequence analysis to comprehensively characterize the image features of the individual.

[0065] Step S302: Read the identification code of the outline area of ​​the main body of the behavior, and at the same time collect the corresponding three-axis acceleration, angular velocity and attitude angle sequence data. Combine the identification code, three-axis acceleration, angular velocity and attitude angle sequence data and output them as the device acquisition data.

[0066] Step S302 reads the identification code from the device carried by the subject and simultaneously collects sensor data such as triaxial acceleration, angular velocity and attitude angle sequence. The above content is integrated into complete device data to enhance the accuracy and robustness of identification.

[0067] Step S303: The dynamic posture vector sequence and the data collected by the device are aligned on a unified time axis using a feature fusion algorithm, and multi-dimensional normalization is performed based on time consistency and action shape matching degree to generate single individual identity feature fusion data.

[0068] Step S303 uses a feature fusion algorithm to perform time-axis alignment and normalization processing on image features and device-acquired data. Combining time consistency and action shape similarity, it outputs identity feature fusion data that can reflect individual characteristics for map matching analysis.

[0069] Step S3 involves the joint extraction and fusion of image features and data collected by the device from the subject of the violation, generating identity feature fusion data with unique identity and behavioral feature correlation, providing a foundation for subsequent identity matching.

[0070] Step S4 includes the following sub-steps:

[0071] Step S401: Using the location data of the violation as a reference, the timestamp, spatial coordinates and individual identity feature fusion data of the outline area of ​​the behavior subject are associated to establish a three-element behavior node. By modeling the trend of individual state changes in the continuous operation image sequence, a behavior identity map is formed. The individual identity feature fusion data is used to construct a risk feature profile of the operator.

[0072] Step S401 establishes a connection between the location data of the violation and the fusion data of the timestamp, spatial coordinates and identity features of the subject of the violation, forming a three-element behavior node, and combines the modeling of the individual state change trend in the operation image sequence to construct a complete behavior identity map structure.

[0073] Step S402: Perform similarity measurement calculation and node matching on the behavior identity graph. Combine time alignment, spatial overlap and action consistency scores, and use a weighted comprehensive mechanism to determine the optimal matching identity code. Bind the optimal matching identity code to the corresponding violation behavior and output a structured violation behavior record. The structured violation behavior record includes the identity code, violation time, violation location and violation behavior type.

[0074] In step S4, when the similarity metric is lower than the set value, the trajectory reconstruction-assisted recognition process is initiated.

[0075] Step S402 performs node matching on the constructed behavioral identity graph, and measures similarity through time alignment, spatial overlap and action consistency scores. The matching results are output using a weighted mechanism, and the optimal matching identity code is bound to the violation behavior to form a structured violation behavior record. The record includes key elements such as identity code, violation time, violation location and violation behavior type. When the similarity measurement of the graph matching is lower than the set threshold, the trajectory reconstruction process is automatically started, and the original data is supplemented by the spatiotemporal changes in the preceding and following image frames to improve the accuracy and completeness of identity matching.

[0076] Step S4 associates the violation with the identity feature fusion data to construct a behavior identity graph, and identifies the violator's identity through multi-dimensional similarity matching, outputting a structured violation record to ensure that the identification results are accurate and traceable.

[0077] Step S5 includes the following sub-steps:

[0078] Step S501: Based on the type of violation, time of violation, and risk level of the preset location of violation in the structured violation record, and in conjunction with the power plant safety management rule base, execute the risk level assessment algorithm to generate risk level labels for the violation. The risk level labels are divided into general risk, major risk, and serious risk.

[0079] Step S501, based on the risk level of the behavior type, time of occurrence, and location of the violation in the structured violation record, and combined with the power plant safety management rule base, performs a risk level assessment and automatically generates clear risk level labels. The labels are divided into general risk, major risk, and serious risk, providing a graded basis for subsequent handling.

[0080] Step S502 involves uploading the structured violation record with the attached risk level label, along with the corresponding identification code, violation time, and violation location, to the power plant safety management platform using a data encryption mechanism.

[0081] Step S502 packages the violation records with attached risk level labels, identification codes, time and location information into data units, and securely uploads them to the power plant safety management platform through an encryption mechanism to achieve unified storage and backend processing and retrieval of information.

[0082] Step S503: Simultaneously, the summary information of the violation record is pushed to the responsible person's mobile terminal through a preset communication interface. The summary information of the violation record includes the risk level, the type of violation, and the identity code, triggering a prompt and processing confirmation process, and establishing a rapid response mechanism for violation information.

[0083] Step S503 pushes the summary information of the violation record to the mobile terminal of the person in charge through a preset communication interface, triggering an immediate prompt and processing confirmation operation, thus establishing a rapid response mechanism for violation information.

[0084] Step S5 assesses the risk level of the identified violations and uploads the records of the violations with risk tags to the platform and pushes them to the terminal, thereby realizing the linkage between risk-based response and rapid processing.

[0085] Step S6 includes the following sub-steps:

[0086] Step S601: Receive the identity tag feedback information pushed by the responsible person's terminal, and verify the identity information by reading the identity information in the NFC chip of the work badge. The verification content includes the identity number, work group number and attendance record for the day. If the match is consistent, the identity confirmation is completed.

[0087] Step S601 receives the identity tag information fed back by the responsible person's terminal, and reads the identity number, work group number and daily sign-in record from the NFC chip of the work badge to complete the identity information matching and verification, ensuring the accurate confirmation of the identity of the subject of the violation.

[0088] Step S602: The confirmed violation records are structured and archived, and the violation record archive data is output. The violation record archive data includes the identity identification code, violation type, violation time, risk level and location coordinates. At the same time, the work trajectory data of the day is called and the archived records are bound with the behavior trajectory data within the corresponding time period. The archived records are stored in the safety supervision database as personnel behavior audit records.

[0089] The records of violations are archived and simultaneously transmitted to the safety training system, and included in the typical case feedback database.

[0090] Step S602 involves structuring and organizing the confirmed violation records to create archived data containing identification codes, violation types, times, risk levels, and coordinates. This data is then combined with the day's work trajectory data for storage, constructing a complete personnel behavior audit record. This record is uniformly stored in the safety supervision database, and the archived data is simultaneously transmitted to the safety training system to supplement the typical case feedback database for subsequent training optimization and risk prevention. Simultaneously, the identity feature fusion data contributes to building risk profiles of workers, providing support for safety management strategies.

[0091] Step S6 verifies and confirms the identity of the violator at the terminal and archives the complete information on the violation in a structured manner. This information is then used for security audits, training feedback, and risk profile construction, forming a closed-loop supervision and management system.

[0092] This invention identifies violations such as not wearing a safety helmet, crossing a safety barrier, and carrying a source of ignition by acquiring work image data and using a violation recognition model. It combines the timestamp and spatial coordinates of the violation to pinpoint its location and extracts image features such as clothing color and texture, body shape, and dynamic posture vector sequences. These features are then fused with identification codes and data collected from devices such as three-axis acceleration and angular velocity to generate identity feature fusion data. The responsible party's identity is then determined through image matching. The system supports risk level label generation, encrypted data upload, terminal prompts for responsible parties, identity verification, and violation record archiving, improving recognition accuracy and closed-loop management capabilities.

[0093] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0094] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for identifying violators in on-site safety supervision of gas turbine power plants, characterized in that, Includes the following steps: Step S1: Collect operational image data of key areas of the gas turbine power plant and identify specific violations; Step S2: Locate the time and place where the violation occurred, and mark the corresponding subject of the violation; Step S3: Extract the image features of the subject of the behavior and the data collected by the device to generate fused identity information; Step S4: Match the violator's identity based on the graph model and output a structured record; Step S5: Assess the level of violation risk and push the information to the management platform and the responsible person's terminal; Step S6: Verify the identity of the responsible person and file the violation record.

2. The method for identifying violators in on-site safety supervision of gas turbine power plants as described in claim 1, characterized in that, Step S1 includes the following sub-steps: Step S101: Collect operation image data. The collection points for the operation image data include the main passage of the gas turbine power plant, entrances and exits, areas near high-voltage equipment, and designated hazardous operation points. Step S102: After preprocessing the work image data in the terminal, the preset violation behavior recognition model is called for real-time recognition. The violation behavior recognition model is based on a multi-scale feature fusion network to identify high-risk behaviors, including not wearing a safety helmet, crossing the isolation line and carrying a fire source. The first recognition result is output. Step S103: Calculate the confidence score of the first identification result. If the confidence score is higher than the preset threshold, mark it as a suspected violation event and store the image frame number and region label of the corresponding operation image data into the violation event cache queue.

3. The method for identifying violators in on-site safety supervision of gas turbine power plants as described in claim 2, characterized in that, Step S2 includes the following sub-steps: Step S201: Extract the timestamps, anchor point positions, and background reference information of the image frames of the work image data marked as suspected violations. Through image feature comparison and calibration coordinate transformation algorithms, calculate the precise time node of the behavior and the spatial coordinates in the work site, and output the violation behavior location data. Step S202: Based on the location data of the violation, retrieve continuous image frames for the corresponding time period, use a multi-target tracking algorithm to extract the activity trajectories of all visible workers, and mark the person whose activity trajectory is closest to the location data of the violation in time and space as the subject of the behavior, and output the individual labeled image fragment.

4. The method for identifying violators in on-site safety supervision of gas turbine power plants as described in claim 3, characterized in that, Step S3 includes the following sub-steps: Step S301: Locate the outline region of the main body from the individual labeled image fragments, extract the clothing color and texture features, body shape and action posture parameters of the main body outline region, and construct a dynamic posture vector sequence based on the time-continuous image. Step S302: Read the identification code of the outline area of ​​the main body of the behavior, and at the same time collect the corresponding three-axis acceleration, angular velocity and attitude angle sequence data. Combine the identification code, three-axis acceleration, angular velocity and attitude angle sequence data and output them as the data collected by the device. Step S303: The dynamic posture vector sequence and the data collected by the device are aligned on a unified time axis using a feature fusion algorithm, and multi-dimensional normalization is performed based on time consistency and action shape matching degree to generate single individual identity feature fusion data.

5. The method for identifying violators in on-site safety supervision of gas turbine power plants as described in claim 4, characterized in that, Step S4 includes the following sub-steps: Step S401: Using the location data of the violation as a reference, the timestamp, spatial coordinates and individual identity features of the outline area of ​​the subject of the violation are associated to establish a three-element behavior node. By modeling the trend of individual state changes in the continuous operation image sequence, a behavior identity map is formed. Step S402: Perform similarity measurement calculation and node matching on the behavioral identity graph. Combine time alignment, spatial overlap and action consistency scores, and use a weighted comprehensive mechanism to determine the optimal matching identity code. Bind the optimal matching identity code to the corresponding violation behavior and output a structured violation behavior record. The structured violation behavior record includes the identity code, violation time, violation location and violation behavior type.

6. The method for identifying violators in on-site safety supervision of gas turbine power plants as described in claim 5, characterized in that, In step S4, when the similarity metric is lower than a set value, the trajectory reconstruction-assisted recognition process is initiated.

7. The method for identifying violators in on-site safety supervision of gas turbine power plants as described in claim 6, characterized in that, Step S5 includes the following sub-steps: Step S501: Based on the type of violation, time of violation, and risk level of the location of violation in the structured violation record, and combined with the power plant safety management rule base, execute the risk level assessment algorithm to generate risk level labels for the violation. The risk level labels are divided into general risk, major risk, and serious risk. Step S502: The structured violation record with risk level label attached, along with the corresponding identification code, violation time and location, is uploaded to the power plant safety management platform through a data encryption mechanism. Step S503: Simultaneously, the summary information of the violation record is pushed to the mobile terminal of the person in charge through a preset communication interface. The summary information of the violation record includes the risk level, the type of violation, and the identity code, triggering the prompt and processing confirmation process, and building a rapid response mechanism for violation information.

8. The method for identifying violators in on-site safety supervision of gas turbine power plants as described in claim 7, characterized in that, Step S6 includes the following sub-steps: Step S601: Receive the identity tag feedback information pushed by the responsible person's terminal, and verify the identity information by reading the identity information in the NFC chip of the work badge. The verification content includes the identity number, work group number and attendance record of the day. If the match is consistent, the identity confirmation is completed. Step S602: The confirmed violation records are structured and archived, and the violation record archive data is output. The violation record archive data includes the identity identification code, violation type, violation time, risk level and location coordinates. At the same time, the work trajectory data of the day is called and the archived records are bound with the behavior trajectory data within the corresponding time period. The archived records are stored in the safety supervision database as personnel behavior audit records.

9. The method for identifying violators in on-site safety supervision of gas turbine power plants as described in claim 8, characterized in that, The recorded violation data is synchronously transmitted to the safety training system and incorporated into the typical case feedback database.

10. The method for identifying violators in on-site safety supervision of gas turbine power plants as described in claim 9, characterized in that, The fused data of individual identity features is used to construct a risk profile of the workers.

Images