CN-120632785-B - Power plant specific area personnel behavior recognition and early warning system

Abstract

The invention discloses a personnel behavior recognition and early warning system for a specific area of a power plant, which relates to the technical field of industrial safety intelligent monitoring and comprises a data acquisition module, an edge calculation module and a central processing platform, wherein the data acquisition module is responsible for acquisition and preliminary processing of multi-source heterogeneous data, including personnel identity, behavior, position and environmental parameters, the edge calculation module performs light weight processing and localization decision on original data, the load of the central processing platform is reduced, and the central processing platform realizes full-flow safety control of a high-risk area through multi-mode data fusion, dynamic authority management and track analysis. In the invention, light AI models are embedded in various cameras of the tablet authentication terminal and the visual perception unit, and preliminary reasoning is directly completed at the tablet authentication terminal and the camera end, so that the tablet authentication terminal and the cameras have edge computing capability, and an independent edge server or an industrial personal computer is deployed near each regional data source of the power plant and used as regional edge computing nodes.

Inventors

• LIN QUANWEI
• FANG HAIYAN

Assignees

• 安徽沃旭智能科技有限公司

Dates

Publication Date

20260505

Application Date

20250609

Claims (4)

1. 1. The system is characterized by comprising a data acquisition module, an edge calculation module and a central processing platform, wherein the data acquisition module is responsible for acquisition and preliminary processing of multi-source heterogeneous data, the acquisition and preliminary processing comprise personnel identity, behaviors, positions and environmental parameters, the edge calculation module performs light weight processing and localization decision on the original data to reduce the load of the central processing platform, and the central processing platform realizes the full-flow safety management and control of a high-risk area through multi-mode data fusion, dynamic authority management and track analysis; The visual perception unit captures personnel action and environment information through the thermal imaging camera, the wide-angle high-definition camera and the behavior analysis camera, the tablet authentication terminal reads tablet information by using RFID/NFC, integrated living body detection prevents identity impersonation, the wearable equipment unit reports personnel position and abnormal state in real time through the UWB positioning module and the acceleration sensor, and the environment perception unit acquires temperature and humidity, gas concentration and radiation dose and marks high-risk environment events; The central processing platform calibrates card punching time, video time stamp and UWB positioning coordinates, a global space consistency view is constructed, and D-S evidence theory is adopted to comprehensively judge risks; The central processing platform updates the authority based on the post-region-period matrix, generates a dynamic check log, the temporary authority is approved and issued through the mobile terminal, the region, the period and the operation type are limited, the electronic fence is drawn, the deviation or the stay timeout of the track is detected, the environment parameter optimization response strategy is linked, and the risk level is pushed to the early warning module; The system forms closed loop verification through four-dimensional data fusion of work card authentication, living body detection, track tracking and environment perception, a work card authentication terminal integrates RFID read-write and near infrared cameras, a ArcFace algorithm is adopted to compare the characteristics of the work card sheet and the face in real time, identity impersonation is judged if cosine similarity is lower than a threshold value, fake attack is resisted through random blink/turning motion detection, meanwhile, a UWB positioning module realizes centimeter-level real-time positioning, a thermal imaging camera is combined to generate human outline coordinates, smoke, dust and a low-illumination environment are effectively penetrated, and when the deviation between a work card authorization area and actual positioning exceeds the threshold value, override behavior judgment is triggered based on a D-S evidence theory.
2. 2. The system for identifying and early warning personnel behaviors in specific areas of a power plant according to claim 1, wherein the edge computing module is used for deploying a lightweight AI model to extract behavior characteristics, filtering invalid data, compressing and structuring the data, and adapting to a unified transmission protocol.
3. 3. The system for identifying and early warning personnel behaviors in specific areas of a power plant according to claim 1, wherein the edge calculation module presets a safety strategy library to trigger local response, cloud dependence is reduced, face and work card information of irrelevant personnel is blurred in real time, and only a risk event key frame after desensitization is uploaded.
4. 4. The system for identifying and early warning personnel behaviors in specific areas of a power plant according to any one of claims 1 to 3 is characterized in that the system flow is as follows: s1, multi-source data acquisition The visual perception unit is used for capturing the outline of a person, the wearing state of safety equipment and high-risk behaviors in real time by the thermal imaging camera, the wide-angle high-definition camera and the behavior analysis camera; The employee license information is forcefully checked at the entrance of the area, and the survival body detection is integrated to prevent the fraudulent use; the wearable equipment unit is used for reporting the positions of personnel and abnormal states of the intelligent safety helmet in real time; the environment sensing unit monitors environment risk parameters by using a temperature and humidity sensor, a gas concentration sensor and a radiation sensor; S2, edge computing pretreatment The lightweight AI reasoning comprises the steps of extracting behavior characteristics of a camera end deployment model in real time and filtering invalid data; The localization rule is executed, wherein a preset rule triggers a local response; Data desensitization and compression, namely carrying out fuzzy processing on the faces of irrelevant people in a video, only reserving key frames of risk events, and uploading after compression; S3, uploading data The pretreated standardized data are transmitted to a central processing platform through a 5G private network or Ethernet dual-channel redundancy, and the data are automatically cached and incrementally restored when the network is interrupted; s4, multi-mode data fusion Time-space alignment, namely calibrating card punching time, a camera time stamp and a time sequence from positioning data to millisecond level; the fusion decision is to comprehensively judge risks by adopting a D-S evidence theory and calculate the risk level by combining environmental parameter weighting; S5, dynamic authority verification The personnel permission is matched in real time based on the post-region-period permission matrix, the temporary permission is issued through mobile terminal approval, and automatic reset is achieved due to the temporary permission; S6, grading early warning triggering Primary early warning, namely, not wearing safety equipment and entering an unauthorized area by mistake; Responding to the action of on-site audible and visual alarm and pushing a reminder to a principal bracelet; second-level early warning, namely overtime override stay and illegal use of tools; responding to the action, namely locking the regional entrance guard, linking the DCS system to limit electricity, and pushing the work order to a manager; Three-stage early warning, namely personnel retention and high risk superposition in fire disaster/leakage; Responding to the action, namely starting the whole plant to stop in an emergency, broadcasting an evacuation instruction, pushing an AR escape path, and dispatching rescue workers; S7 edge response execution The early warning instruction is issued to the edge node through an increment protocol, directly triggers local action, is deeply integrated with the power plant DCS system, and automatically executes equipment linkage; S8, full link tracing Storing event-associated desensitization data; supporting multidimensional retrieval according to time, personnel and areas, and backtracking a complete event chain by one key; S9, compliance report generation And automatically extracting the safety index, generating a report conforming to the ISO45001 standard, and adapting to supervision and inspection.

Description

Power plant specific area personnel behavior recognition and early warning system Technical Field The invention relates to the technical field of industrial safety intelligent monitoring, in particular to a system for identifying and early warning personnel behaviors in specific areas of a power plant. Background The industrial safety intelligent monitoring technology is a core field for guaranteeing the safety of personnel and equipment in high-risk industrial environments. The field focuses on realizing active safety control on personnel behaviors, environmental risks and equipment states through multi-mode sensing, real-time data analysis and intelligent decision, and a power plant is used as a typical high-risk scene, has risks of high-temperature and high-pressure equipment, radiation areas, toxic gas leakage and the like, and has more severe requirements on a monitoring system. The prior art generally relies on manual inspection and a fixed alarm device to form a passive monitoring mode, when personnel perform illegal operation or enter a forbidden zone by mistake, the system cannot actively identify real-time risks, response delay is caused, hysteresis makes accident prevention capability weak, illegal actions are usually found only in postretrospective tracing, anti-jamming capability is insufficient due to the singleness of a sensing mode, the traditional scheme adopts an isolated sensor, each system data is split and cannot be subjected to cross verification, the single-mode system is difficult to adapt to complex working conditions of high temperature, high pressure and high dust of a power plant, authority configuration relies on manual updating of a scheduling table, change delay often exceeds 2 hours, actual moving paths of personnel in an authorized zone cannot be tracked, so that actions such as override stay, mistaken entering a radiation zone and the like cannot be blocked in real time. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a power plant specific area personnel behavior recognition and early warning system. In order to achieve the purpose, the system for identifying and early warning personnel behaviors in the specific area of the power plant adopts the following technical scheme that the system comprises a data acquisition module, an edge calculation module and a central processing platform, wherein the data acquisition module is responsible for acquisition and preliminary processing of multi-source heterogeneous data, the acquisition and preliminary processing of the multi-source heterogeneous data comprise personnel identities, behaviors, positions and environmental parameters, the edge calculation module performs light weight processing and localization decision on original data, the load of the central processing platform is reduced, and the central processing platform realizes full-flow safety control of a high-risk area through multi-mode data fusion, dynamic authority management and track analysis. As a further description of the above technical solution: The data acquisition module comprises: the visual perception unit captures personnel action and environmental information through the thermal imaging camera, the wide-angle high-definition camera and the behavior analysis camera, the tablet authentication terminal reads tablet information through RFID/NFC, integrated living body detection prevents identity impersonation, the wearable equipment unit reports personnel position and abnormal state in real time through the UWB positioning module and the acceleration sensor, the environmental perception unit collects temperature and humidity, gas concentration and radiation dose, and high risk environmental event is marked. As a further description of the above technical solution: The edge computing module deploys a lightweight AI model to extract behavior characteristics, filters invalid data, compresses and constructs data, and adapts a unified transmission protocol. As a further description of the above technical solution: The edge calculation module presets a security policy library to trigger local response, cloud dependence is reduced, face and work card information of irrelevant personnel is blurred in real time, and only a risk event key frame after desensitization is uploaded. As a further description of the above technical solution: the central processing platform calibrates card punching time, video time stamp and UWB positioning coordinates, constructs global space consistency view, and comprehensively judges risks by adopting D-S evidence theory. As a further description of the above technical solution: The central processing platform updates the authority based on the post-region-period matrix, generates a dynamic check log, the temporary authority is approved and issued through the mobile terminal, the region, the period and the operation type are limited, the electronic fence is drawn, the deviation of the detectio