A safe production hidden danger investigation intelligent management method and system

Abstract

The application discloses a safety production hidden danger investigation intelligent management method and system, and belongs to the technical field of engineering safety and environmental supervision. The technical problem to be solved is that the existing hidden danger investigation mode has the problems of serious information island, strong subjectivity of hidden danger identification, false closed loop in rectification process, extensive expert resource scheduling, and formality of one-line operation safety access control. The technical solution points are to construct a whole-process closed-loop management system coordinated by Web and mobile terminals, to realize intelligent assistance of expert hidden danger reporting through AI image recognition, to realize rectification rigid closed loop by adopting hidden danger intelligent disassembly and multi-department parallel rectification mechanism, and to realize one-line operation access forced control by relying on personalized safety briefing, AI hidden danger investigation and safety ticket hardware linkage.

Description

Technical Field

[0001] This invention belongs to the field of engineering safety and environmental supervision technology, specifically relating to an intelligent management method and system for identifying potential safety hazards in production. Background Technology

[0002] The current safety hazard investigation work mainly relies on manual on-site inspections and paper or simple electronic records, which has many pain points and shortcomings and cannot meet the core needs of modern engineering safety supervision. The specific defects are as follows: (1) Severe information silos and low collaboration efficiency: Basic project information, expert resources, hidden danger data, rectification feedback, etc. are scattered in different departments and systems, and there is a lack of a unified digital platform for integration. Experts cannot accurately obtain key information such as the distribution of project risk points and the qualifications of subcontractors, resulting in a lack of targeted investigation; the project department also cannot keep track of the experts' schedules and task assignments in real time, resulting in high communication costs; (2) High subjectivity in hazard identification and insufficient professional support: Hazard investigation relies heavily on the personal experience of experts. For non-professionals or those with insufficient experience, it is difficult to accurately identify complex hazards. At the same time, experts need to manually consult a large number of standard provisions on-site to verify their judgments. The process is cumbersome and prone to errors, affecting the efficiency and authority of the investigation. (3) The rectification process is not closed-loop, and it is difficult to trace responsibility: After the notice of rectification of hidden dangers is issued, it often leads to shirking responsibility and passing the buck due to unclear responsible departments and unclear rectification requirements. The rectification feedback is mostly reported once, and there is a lack of independent verification and summary mechanism for the completion of sub-tasks of each department. This easily leads to a false closed-loop phenomenon of "reporting after partial rectification", and it cannot ensure that all hidden dangers are effectively eliminated. (4) The allocation of expert resources is crude and the management efficiency is low: information such as the experts' professional fields, available time, and geographical location is not effectively utilized, and the task allocation lacks a scientific basis. Conflicts are prone to occur in the experts' schedules, and there is a lack of effective reminder and coordination mechanisms, which affects the timeliness and coverage of the investigation work; (5) Safety access for front-line workers is merely a formality: For front-line workers such as team members, safety briefings, qualification verification, and hazard self-inspection are often conducted verbally or in writing, making it difficult to guarantee their authenticity and effectiveness. The lack of technical means to forcibly link "safety conditions" with "work permits" leads to frequent violations such as "working without briefing" and "working with defects".

[0003] In view of this, the present invention is hereby proposed. Summary of the Invention

[0004] In order to solve the above-mentioned technical problems in the existing technology, the present invention provides an intelligent management method and system for safety production hazard investigation, which solves the pain points of the existing hazard investigation mode, such as serious information silos, strong subjectivity in hazard identification, easy occurrence of false closed loops in the rectification process, extensive expert resource allocation, and the perfunctory nature of front-line operation safety access control.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows: Firstly, a smart management method for identifying and managing potential safety hazards in production includes: S1. Construct a project lifecycle information database and an expert full-dimensional archive database on the Web management terminal, and complete the intelligent allocation of expert tasks and itinerary collaborative management based on the project lifecycle information database and the expert full-dimensional archive database; S2. Hazard reporting is conducted via the web interface. After uploading on-site hazard photos, the system automatically triggers the AI ​​image recognition engine to perform feature analysis on the hazard photos. The system also recommends safety regulations and typical cases that match the hazard features to experts in real time through a floating assistant pop-up window, assisting experts in completing the hazard reporting. S3. Based on the reported hazard type, the system calls the preset "hazard type - responsible department" mapping rule to automatically generate a main rectification notice and several sub-task sheets corresponding to each responsible department, and accurately pushes the main rectification notice and sub-task sheets to the corresponding responsible departments. S4. Each responsible department independently completes the rectification feedback of the corresponding sub-task order within the platform. The system will automatically summarize the rectification content of all sub-task orders and generate a complete rectification feedback form only after all sub-task orders have been confirmed to be rectified. The rectification feedback form will then be submitted to the approval process to complete the closed loop of hidden danger rectification. S5. Team members select the corresponding job type and work content via mobile device. The system generates a personalized "Safety Risk Notification" based on the selected job type and work content. The AI ​​hazard investigation process can only be started after all the participants have completed the electronic confirmation. S6. The mobile terminal records the entire process of identifying potential hazards at the work site and performs AI hazard analysis on the work site photos uploaded by team members. Only when all inspection points are determined to be safe and without hazards will the system issue a structured electronic "safety ticket". S7. The validity status of the "safety ticket" is synchronized to the on-site access control system or the operating equipment control system in real time, and the "safety ticket" is used as a mandatory permit for operators to enter the work area and start the operating equipment to carry out construction work.

[0006] Furthermore, the project lifecycle information database is used for refined management of project type, project address, risk level, subcontractor qualifications, and project organizational structure information; The comprehensive expert archive records experts' professional fields, historical performance, and competency evaluation information. The system supports regionalized task allocation for experts based on a map view. Experts can set standardized travel statuses. The system has built-in travel conflict detection and calendar view display functions. Travel information is synchronized to the administrator console in real time, and task reminders are sent to experts at key time points through multiple channels.

[0007] Furthermore, the AI ​​image recognition engine includes a lightweight AI inference server deployed at the edge layer of the project site and a cloud AI service. The edge layer lightweight AI inference server is used to perform low-latency calculation tasks for edge AI recognition in high-frequency hidden danger scenarios, and the cloud AI service is used to perform AI model inference and standard clause matching calculations for complex hidden danger features. The floating assistant pop-up allows experts to evaluate the recommended safety guidelines and typical cases as "useful" or "useless," and the evaluation data is used to feed back into the continuous iterative optimization of the AI ​​matching and recognition models.

[0008] Furthermore, the "Hazard Type - Responsible Department" mapping rule base supports dynamic configuration and flexible adjustment in the background to adapt to the organizational structure and responsibility division system of different projects.

[0009] Furthermore, the personalized "Safety Risk Notification" is automatically generated by the system based on the built-in multi-dimensional knowledge graph of "job type-operation-risk-measures". The generated content includes common risk types, legal basis, safety protection measures and a list of prohibited behaviors for the corresponding operation.

[0010] Furthermore, the successful issuance of the "safety ticket" requires the simultaneous fulfillment of three prerequisites: electronic confirmation of safety briefing by all personnel, qualification of all inspection points in AI hazard investigation, and complete screen recording of the inspection process. If any one of the prerequisites is not met, the system will not be able to issue a valid "safety ticket".

[0011] Furthermore, the status data of the "safety ticket" is pushed to the on-site access control system or the control system of the work equipment in real time via the MQTT protocol, so as to realize the mandatory control of "no construction without a ticket and no work without a briefing".

[0012] Secondly, a safety hazard investigation and management system, used in any of the preceding descriptions of the safety hazard investigation and management method, includes: The terminal layer includes a web browser client and a mobile client. The web browser client is used by experts, administrators, and project team members to complete project information management, expert task allocation, hazard reporting, and rectification process control. The mobile client is used by team members to complete safety briefing confirmation, hazard investigation at the work site, and on-site audio and video acquisition. The edge layer includes lightweight AI inference servers deployed on-site for performing low-latency, high-privacy computing tasks such as face recognition liveness detection and edge AI recognition in high-frequency hidden danger scenarios. The cloud layer is a central management platform based on a microservice architecture, including user service units, project service units, expert service units, hidden danger investigation service units, rectification management service units, AI service units, file service units, and message service units. It is used to execute core business logic processing, big data storage, complex AI model inference, safety ticket status management, and linkage control with external hardware systems.

[0013] Furthermore, the rectification management service unit in the cloud layer has a built-in "hazard type - responsible department" mapping rule library, which is used to automatically generate and push the main rectification notice and sub-task order, as well as automatically summarize and close the loop judgment of sub-task rectification feedback. Furthermore, the AI ​​service unit in the cloud layer has a built-in structured safety specification database and a multi-dimensional knowledge graph of "job type-operation-risk-measures" for intelligent matching of hazard characteristics and specification clauses, automatic generation of personalized safety risk notifications, and AI hazard analysis of work surface images.

[0014] The beneficial effects of this invention are as follows: (1) Significantly improve management standardization and collaboration efficiency: By breaking down information barriers through a unified platform, information is shared transparently among projects, experts, and rectification departments, and expert schedules and task allocation are scientific and efficient, greatly reducing communication and coordination costs; (2) Greatly enhances the professionalism and accuracy of hazard identification: AI intelligent assistant provides experts with timely and accurate legal basis and case references, combining expert experience with AI computing power, effectively making up for individual ability differences and improving the overall quality of investigation; (3) Thoroughly realize the rigid closed loop of the rectification process: the multi-department parallel rectification and automatic summary mechanism ensures that every hidden danger point is implemented, solves the industry's chronic problems of unclear rectification responsibilities and false closed loop from the source, and ensures the authenticity and effectiveness of rectification; (4) Comprehensively strengthen safety access control for front-line operations: Through multiple technical means such as mandatory briefing, AI self-inspection, and “safety ticket” linkage, the reach of safety management is extended to the last-line workers, realizing a fundamental shift from “human prevention” to “technical prevention”, and effectively preventing accidents caused by lack of briefing or unresolved hidden dangers. Attached Figure Description

[0015] Figure 1 A flowchart of the intelligent management method for identifying and managing safety hazards provided in an embodiment of the present invention. Detailed Implementation

[0016] The technical solution of the present invention will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are not all embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0017] It should be noted that, unless otherwise specifically stated, the relative arrangement and numerical expressions of the components and steps described in these embodiments should not be construed as limiting the scope of the invention.

[0018] The following description of exemplary embodiments is merely illustrative and is not intended to limit the invention or its application or use in any way. Techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail herein, but where applicable, such techniques, methods, and apparatus should be considered part of this specification.

[0019] Example 1 See Figure 1 , Figure 1 This is a flowchart of an intelligent management method for identifying and managing safety hazards proposed in this invention. Specific steps may include: S1. Information database construction and task collaboration: Construct a project lifecycle information database and an expert full-dimensional archive database on the Web management terminal, and complete the intelligent allocation of expert tasks and itinerary collaboration management based on the project lifecycle information database and the expert full-dimensional archive database; Among them, the project lifecycle information database is used for refined management of project type, project address, risk level, subcontractor qualifications and project organizational structure information; The expert full-dimensional archive is used to record experts' professional fields, historical performance, and ability evaluation information. The system supports regional task allocation for experts based on map view. Experts can set standardized travel status. The system has built-in travel conflict detection and calendar view display functions. Travel information is synchronized to the administrator console in real time, and task reminders are sent to experts at key time points through multiple channels.

[0020] S2. AI-assisted hazard reporting: Hazard reporting is conducted via the web interface. After uploading on-site hazard photos, the system automatically triggers the AI ​​image recognition engine to perform feature analysis on the hazard photos. The system also recommends safety regulations and typical cases that match the hazard features to experts in real time through a floating assistant pop-up window, assisting experts in completing the hazard reporting. The AI ​​image recognition engine includes a lightweight AI inference server deployed at the edge layer of the project site and a cloud AI service. The edge layer lightweight AI inference server is used to perform low-latency computing tasks for edge AI recognition in high-frequency hidden danger scenarios, while the cloud AI service is used to perform AI model inference and standard clause matching calculations for complex hidden danger features. The floating assistant pop-up allows experts to evaluate the recommended safety guidelines and typical cases as "useful" or "useless," and the evaluation data is used to feed back into the continuous iterative optimization of the AI ​​matching and recognition models.

[0021] S3. Hazard identification and task assignment: Based on the reported hazard type, the system calls the preset "hazard type - responsible department" mapping rule to automatically generate a main rectification notice and several sub-task sheets corresponding to each responsible department, and accurately pushes the main rectification notice and sub-task sheets to the corresponding responsible departments. Among them, the "Hazard Type - Responsible Department" mapping rule library supports dynamic configuration and flexible adjustment in the background to adapt to the organizational structure and responsibility division system of different projects.

[0022] S4. Multi-department rectification and closed-loop approval: Each responsible department independently completes the rectification feedback of the corresponding sub-task order within the platform. The system automatically summarizes the rectification content of all sub-task orders and generates a complete rectification feedback form only after all sub-task orders have been confirmed for rectification. The rectification feedback form is then submitted to the approval process to complete the closed loop of hidden danger rectification. S5. Personalized Safety Briefing Confirmation: Team members select their corresponding job type and work content via mobile device. The system generates a personalized "Safety Risk Notification" based on the selected job type and work content. The AI ​​hazard investigation process can only be started after all participants have completed the electronic confirmation. Among them, the personalized "Safety Risk Notification" is automatically generated by the system based on the built-in multi-dimensional knowledge graph of "job type-operation-risk-measure". The generated content includes common risk types, legal basis, safety protection measures and a list of prohibited behaviors for the corresponding operation.

[0023] S6. AI Hazard Investigation and Safety Ticket Issuance: The mobile device records the entire process of hazard investigation on the work surface and performs AI hazard analysis on the work surface pictures uploaded by team members. Only when all investigation points are determined to be hazard-free and qualified will the system issue a structured electronic "safety ticket". Specifically, the hazard image recognition adopts a method based on... The deep learning model trained by the framework is specifically trained and optimized for high-frequency hidden danger scenarios in engineering such as electrical, high-altitude, scaffolding, and edge protection. For edge deployment scenarios, the lightweight YOLOv5s model is used and deployed to the edge AI inference server at the project site to achieve low-latency edge hidden danger identification.

[0024] The successful issuance of the "safety ticket" requires the simultaneous fulfillment of three prerequisites: electronic confirmation of safety briefing by all personnel, qualification of all checkpoints in AI hazard investigation, and complete screen recording of the investigation process. If any one of the prerequisites is not met, the system will not be able to issue a valid "safety ticket".

[0025] S7. Safety Ticket Hardware Linkage Control: The validity status of the "safety ticket" is synchronized to the on-site access control system or the operating equipment control system in real time, and the "safety ticket" serves as a mandatory permit for operators to enter the work area and start the operating equipment to carry out construction work.

[0026] The status data of the "safety ticket" is pushed to the on-site access control system or the control system of the work equipment in real time via the MQTT protocol, so as to realize the mandatory control of "no construction without a ticket and no work without a briefing".

[0027] Example 2 This invention proposes an intelligent management system for identifying potential safety hazards in production, comprising: M1, the terminal layer, includes a web browser client and a mobile client. The web browser client is used by experts, administrators, and project team members to complete project information management, expert task allocation, hazard reporting, and rectification process control. The mobile client is used by team members to complete safety briefing confirmation, hazard investigation at the work site, and on-site audio and video acquisition. M2, the edge layer, includes lightweight AI inference servers deployed on-site for performing low-latency, high-privacy computing tasks such as face recognition liveness detection and edge AI recognition in high-frequency hidden danger scenarios. M3, the cloud layer, is a central management platform based on a microservice architecture. It includes user service units, project service units, expert service units, hidden danger investigation service units, rectification management service units, AI service units, file service units, and message service units. It is used to perform core business logic processing, big data storage, complex AI model inference, safety ticket status management, and linkage control with external hardware systems.

[0028] The cloud-based rectification management service unit has a built-in "hazard type - responsible department" mapping rule library, which is used to automatically generate and push the main rectification notice and sub-task order, as well as automatically summarize and close the loop judgment of sub-task rectification feedback. Among them, the cloud-based AI service unit has a built-in structured safety specification database and a multi-dimensional knowledge graph of "job type-operation-risk-measures" for intelligent matching of hazard characteristics and specification clauses, automatic generation of personalized safety risk notifications, and AI hazard analysis of work surface images.

[0029] Example 3 This embodiment uses the safety hazard investigation and management of a new energy power plant construction project as an example to illustrate the method of the present invention in detail. The specific steps are as follows: B1. Expert Task Allocation and Preparation The administrator viewed the expert map view on the web and found that Expert A (specializing in electrical safety) was currently in an "idle standby" state and located near the project. Based on the project information (including a high-risk point: installation of electrical equipment at the substation), the system automatically recommended Expert A. The administrator issued an investigation task, and Expert A received both an in-system message and an SMS notification on their mobile device.

[0030] B2. Expert on-site investigation and intelligent assistance Expert A arrived at the site and filled out the hazard report using the web-based, split-screen interface. After uploading a photo of the haphazardly laid temporary cables, the AI ​​image recognition engine immediately extracted hazard features from the image, and the AI ​​assistant simultaneously recommended: "Article 7.2.3 of the 'Technical Specification for Safety of Temporary Power Supply at Construction Sites' JGJ46-2005: Cable lines should be laid underground or overhead, and it is strictly prohibited to lay them openly along the ground. Mechanical damage and media corrosion should be avoided," along with typical rectification cases of similar hazards. Expert A adopted the recommended content and quickly completed the problem description and rectification suggestions.

[0031] B3. Intelligent dismantling and issuance of rectification tasks The system identifies the hazard type as "temporary power supply". Based on the preset "hazard type - responsible department" mapping rules, it automatically generates a main rectification notice and breaks it down into two sub-task orders, which are issued to "Engineering Technology Department" (responsible for preparing compliant cable laying technical solutions) and "Subcontractor B" (responsible for the rectification and construction of on-site cable laying) respectively.

[0032] B4. Parallel rectification and feedback by multiple departments The Engineering and Technology Department submitted a compliant cable laying plan online within the system, completing the feedback for the corresponding sub-task. Subcontractor B uploaded photos of the rectified site and construction records, which were electronically signed and confirmed by the project manager, completing the feedback for the corresponding sub-task. The operations of the two departments did not interfere with each other, and the rectification work was completed in parallel.

[0033] B5. Automatic Summarization and Closed-Loop Approval The system detects in real time that both sub-tasks have completed rectification feedback, automatically summarizes the rectification content and supporting materials of the two sub-tasks, generates a complete "Rectification Feedback Form," and automatically submits it to the project safety director for review. After the safety director's approval, the hazard is officially closed and archived.

[0034] B6. Linking team operations with "safety tickets" The following day, the electrician team was required to perform wiring work in the rectification area. The team leader selected the "Electrician - Temporary Power Wiring" job and work content on the mobile device. The system then called up the built-in "Job Type - Work - Risk - Measures" knowledge graph and automatically generated a personalized "Safety Risk Notification" for the corresponding work.

[0035] After all work team members read and completed the electronic confirmation, the system unlocked the hazard inspection function. Team members took photos of the work area and uploaded them. The system used AI to analyze the photos, confirming that all inspection points were free of hazards. It also verified the completeness of the inspection process recording and finally issued a valid electronic "safety ticket." The validity status of this "safety ticket" was synchronized in real time to the work area access control system via the MQTT protocol. After team members verified their identity and valid safety ticket by facial recognition, they successfully entered the work area to begin construction.

[0036] The above specific 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 examples, 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 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 intelligent management of safety hazard investigation, characterized in that, include: S1. Construct a project lifecycle information database and an expert full-dimensional archive database on the Web management terminal, and complete the intelligent allocation of expert tasks and itinerary collaborative management based on the project lifecycle information database and the expert full-dimensional archive database; S2. Hazard reporting is conducted via the web interface. After uploading on-site hazard photos, the system automatically triggers the AI ​​image recognition engine to perform feature analysis on the hazard photos. The system also recommends safety regulations and typical cases that match the hazard features to experts in real time through a floating assistant pop-up window, assisting experts in completing the hazard reporting. S3. Based on the reported hazard type, the system calls the preset "hazard type - responsible department" mapping rule to automatically generate a main rectification notice and several sub-task sheets corresponding to each responsible department, and accurately pushes the main rectification notice and sub-task sheets to the corresponding responsible departments. S4. Each responsible department independently completes the rectification feedback of the corresponding sub-task order within the platform. The system will automatically summarize the rectification content of all sub-task orders and generate a complete rectification feedback form only after all sub-task orders have been confirmed to be rectified. The rectification feedback form will then be submitted to the approval process to complete the closed loop of hidden danger rectification. S5. Team members select the corresponding job type and work content via mobile device. The system generates a personalized "Safety Risk Notification" based on the selected job type and work content. The AI ​​hazard investigation process can only be started after all the participants have completed the electronic confirmation. S6. The mobile terminal records the entire process of identifying potential hazards at the work site and performs AI hazard analysis on the work site photos uploaded by team members. Only when all inspection points are determined to be safe and hazard-free will the system issue a structured electronic "safety ticket". S7. The validity status of the "safety ticket" is synchronized to the on-site access control system or the operating equipment control system in real time, and the "safety ticket" is used as a mandatory permit for operators to enter the work area and start the operating equipment to carry out construction work.

2. The intelligent management method for identifying and managing safety hazards in production according to claim 1, characterized in that, In step S1, the project lifecycle information database is used for refined management of project type, project address, risk level, subcontractor qualifications and project organizational structure information; The comprehensive expert archive records experts' professional fields, historical performance, and competency evaluation information. The system supports regionalized task allocation for experts based on a map view. Experts can set standardized travel statuses. The system has built-in travel conflict detection and calendar view display functions. Travel information is synchronized to the administrator console in real time, and task reminders are sent to experts at key time points through multiple channels.

3. The intelligent management method for identifying and managing safety hazards in production according to claim 1, characterized in that, In step S2, the AI ​​image recognition engine includes a lightweight AI inference server deployed at the edge layer of the project site and a cloud AI service. The edge layer lightweight AI inference server is used to perform low-latency calculation tasks for edge AI recognition in high-frequency hidden danger scenarios, and the cloud AI service is used to perform AI model inference and standard clause matching calculations for complex hidden danger features. The floating assistant pop-up allows experts to evaluate the recommended safety guidelines and typical cases as "useful" or "useless," and the evaluation data is used to feed back into the continuous iterative optimization of the AI ​​matching and recognition models.

4. The intelligent management method for identifying and managing safety hazards according to claim 1, characterized in that, In step S3, the "Hazard Type - Responsible Department" mapping rule base supports dynamic configuration and flexible adjustment in the background to adapt to the organizational structure and responsibility division system of different projects.

5. The intelligent management method for identifying and managing safety hazards in production according to claim 1, characterized in that, In step S5, the personalized "Safety Risk Notification" is automatically generated by the system based on the built-in multi-dimensional knowledge graph of "job type-operation-risk-measure". The generated content includes common risk types, legal basis, safety protection measures and a list of prohibited behaviors for the corresponding operation.

6. The intelligent management method for identifying and managing safety hazards in production according to claim 1, characterized in that, In step S6, the successful issuance of the "safety ticket" requires the simultaneous fulfillment of three prerequisites: the completion of electronic confirmation of safety briefing by all personnel, the passing of all inspection points in the AI ​​hazard investigation, and the complete screen recording of the inspection process. If any of the prerequisites are not met, the system will not be able to issue a valid "safety ticket".

7. The intelligent management method for identifying and managing safety hazards in production according to claim 1, characterized in that, In step S7, the status data of the "safety ticket" is pushed to the on-site access control system or the control system of the work equipment in real time via the MQTT protocol, so as to realize the mandatory control of "no construction without ticket and no work without briefing".

8. A smart management system for identifying and managing potential safety hazards in production, characterized in that, The method for implementing the intelligent management method for identifying and managing safety hazards according to any one of claims 1-7 includes: The terminal layer includes a web browser client and a mobile client. The web browser client is used by experts, administrators, and project team members to complete project information management, expert task allocation, hazard reporting, and rectification process control. The mobile client is used by team members to complete safety briefing confirmation, hazard investigation at the work site, and on-site audio and video acquisition. The edge layer includes lightweight AI inference servers deployed on-site for performing low-latency, high-privacy computing tasks such as face recognition liveness detection and edge AI recognition in high-frequency hidden danger scenarios. The cloud layer is a central management platform based on a microservice architecture, including user service units, project service units, expert service units, hidden danger investigation service units, rectification management service units, AI service units, file service units, and message service units. It is used to execute core business logic processing, big data storage, complex AI model inference, safety ticket status management, and linkage control with external hardware systems.

9. The intelligent management system for identifying and managing potential safety hazards according to claim 8, characterized in that, The rectification management service unit in the cloud layer has a built-in "hazard type - responsible department" mapping rule library, which is used to automatically generate and push the main rectification notice and sub-task, as well as automatically summarize and close the loop judgment of sub-task rectification feedback.

10. The intelligent management system for identifying and managing potential safety hazards according to claim 8, characterized in that, The AI ​​service unit in the cloud layer has a built-in structured safety specification database and a multi-dimensional knowledge graph of "job type-operation-risk-measures" for intelligent matching of hazard characteristics and specification clauses, automatic generation of personalized safety risk notifications, and AI hazard analysis of work surface images.

Images