Bid evaluation base total-factor real-time linkage supervision method and system

Abstract

The invention provides a total-factor real-time linkage supervision method and system for a bid evaluation base, and relates to the technical field of bid evaluation management. Comprising the steps of collecting and associating multi-source data in real time; constructing a three-dimensional model, and dynamically mapping the video monitoring data, the business management data and the Internet of Things equipment data to the three-dimensional model to form a three-dimensional digital twin model; performing intelligent analysis on the multi-source data through AI, and identifying abnormal behaviors; after behavior abnormity is recognized through the AI, an alarm event is triggered, and related video monitoring fragments, personnel historical tracks and equipment logs within a preset time period before and after the alarm event occurs are pushed to management personnel; and a manager locates an alarm event in the three-dimensional digital twinborn model, verifies and marks a processing result, and forms closed-loop feedback. According to the invention, the normal form upgrade of bid evaluation supervision from manual screen staring and post tracing to intelligent early warning and real-time intervention is realized, and a technical benchmark is provided for construction of a green digital intelligent bid evaluation base.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of bid evaluation, and particularly relates to a bid evaluation base supervision method and system. BACKGROUND

[0002] The existing bid evaluation base supervision system mostly adopts independently deployed monitoring devices (such as cameras and access control systems) and basic business systems (such as video storage systems and personnel check-in systems) to perform on-site supervision through manual inspection and scattered data.

[0003] 1. Video monitoring system: multiple cameras are installed in the bid evaluation base to realize full-coverage recording of the physical space, support real-time picture viewing and historical video playback, but its function is limited to passive recording and lacks intelligent analysis of video content (such as personnel behavior recognition or abnormal event detection).

[0004] 2. Basic business management system: responsible for the electronic management of the bid evaluation process, including bid evaluation meeting information input, expert allocation, permission configuration and other functions, but the system is closed and has no data interaction with video monitoring and Internet of Things devices (such as access control and storage cabinets), resulting in manual input of personnel check-in information and inability to link with the access control.

[0005] 3. Manual supervision process: relies on management personnel to regularly patrol the site, manually view monitoring pictures or handle device exception reports, and manually records and responds to alarm events, which is low in efficiency and prone to omissions due to negligence.

[0006] Overall, the existing technology takes "passive monitoring + manual management" as the core, which meets the basic supervision needs, but lacks data fusion, intelligent analysis and three-dimensional visualization capabilities, resulting in low supervision efficiency, high cost and blind spots. SUMMARY

[0007] Therefore, the technical problem to be solved by the application is to provide a bid evaluation base full-element real-time linkage supervision method and system, aiming to improve the efficiency and accuracy of bid evaluation base supervision.

[0008] The application provides a bid evaluation base full-element real-time linkage supervision method, which comprises the following steps:

[0009] Step 1: Real-time acquisition of multi-source data

[0010] Real-time acquisition of multi-source data, wherein the multi-source data comprises video monitoring data, business management data and Internet of Things device data, and the video monitoring data, the business management data and the Internet of Things device data are associated.

[0011] Step 2: Establishing a three-dimensional digital twin model

[0012] constructing a three-dimensional model, dynamically mapping the video monitoring data, the business management data and the Internet of Things equipment data to the three-dimensional model to form a three-dimensional digital twin model;

[0013] Step 3: Abnormal behavior identification

[0014] Intelligently analyzing the video monitoring data, the business management data and the Internet of Things equipment data through AI to identify abnormal behavior;

[0015] Step 4: Abnormal behavior processing

[0016] After identifying abnormal behavior through AI, triggering an alarm event, and pushing the relevant video monitoring segments, personnel historical trajectories and equipment logs within a preset time period before and after the alarm event to the management personnel; the management personnel locates the alarm event in the three-dimensional digital twin model, verifies and marks the processing results to form a closed-loop feedback.

[0017] Further, in step 1, the Internet of Things equipment data is pushed to the edge proxy server through MQTT / HTTP protocol, and the business management data is synchronized to the edge proxy server through API interface; the edge proxy server associates and preprocesses the video monitoring data, the business management data and the Internet of Things equipment data.

[0018] Further, in step 1, the Internet of Things equipment data includes face recognition cameras, person certificate reporting machines, intelligent storage cabinets, face recognition access control, intelligent door locks, and recording telephones; the business management data includes bid evaluation meeting information and personnel white lists.

[0019] Further, in step 1, the Internet of Things equipment uses ultra-wideband or Bluetooth beacons to locate personnel in real time.

[0020] Further, in step 2, the three-dimensional digital twin model is used for panoramic tracing and spatial query through an interactive interface.

[0021] Further, in step 2, the three-dimensional digital twin model is constructed based on UE5 or Unity engine.

[0022] Further, in step 2, the three-dimensional digital twin model uses WebGL rendering to realize three-dimensional visualization based on a browser.

[0023] Further, in step 3, the abnormal behavior identification is performed through boundary detection of computer vision, combined with personnel permissions and location information for cross verification.

[0024] Further, in step 3, the video stream data is divided in time dimension and space dimension to obtain a plurality of personnel behavior feature sets; the plurality of personnel behavior feature sets are input into a trained abnormal behavior recognition model; the multi-modal feature coding layer of the abnormal behavior recognition model performs cross-modal correlation fusion on personnel category features, personnel permission features and isolation area features; in real-time monitoring of the video stream data, when the target behavior feature is identified as an abnormal behavior, an alarm event is triggered.

[0025] The application also provides an evaluation base full-element real-time linkage supervision system, comprising:

[0026] A multi-source data real-time acquisition module acquires multi-source data in real time, wherein the multi-source data comprises video monitoring data, business management data and Internet of Things equipment data, and the video monitoring data, the business management data and the Internet of Things equipment data are associated;

[0027] A three-dimensional digital twin model constructs a three-dimensional model, dynamically maps the video monitoring data, the business management data and the Internet of Things equipment data to the three-dimensional model, and forms a three-dimensional digital twin model;

[0028] An abnormal behavior recognition module intelligently analyzes the video monitoring data, the business management data and the Internet of Things equipment data through AI to identify abnormal behaviors;

[0029] An abnormal behavior processing module triggers an alarm event after identifying an abnormal behavior through AI, and pushes video monitoring segments, personnel historical trajectories and equipment logs within a preset time period before and after the alarm event to a management personnel; the management personnel locates the alarm event in the three-dimensional digital twin model, verifies and marks a processing result, and forms a closed-loop feedback.

[0030] Beneficial effects:

[0031] The evaluation base full-element real-time linkage supervision method provided by the application systematically solves the core problems of data islands, response lag and poor interactivity in traditional evaluation supervision technology. Through digital twin technology, multi-source data is integrated, AI algorithm realizes automatic abnormal identification and alarm linkage, and three-dimensional visualization technology converts complex data into intuitive spatial information. Through the above technical means, the accuracy, real-time performance and efficiency of supervision are significantly improved, and a full-process automatic management from abnormal discovery to closed-loop processing is formed, which provides a reusable technical paradigm and practical reference for the construction of the green and digital evaluation base of State Grid.

[0032] The evaluation base full-element real-time linkage supervision system provided by the application has the same technical effects. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to make the content of the present application more easily understood, the present application will be further described in detail below according to specific embodiments of the present application and in conjunction with the accompanying drawings.

[0034] Figure 1 The flow chart of the full-factor real-time linkage supervision method for the bid evaluation base of embodiment 1 of the present application. DETAILED DESCRIPTION

[0035] The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments. The principles and characteristics of the present application are described below in conjunction with the accompanying drawings. It should be noted that the embodiments in the present application and the characteristics in the embodiments can be combined with each other without conflict. The embodiments are only used to explain the present application and are not used to limit the scope of the present application.

[0036] The present application first analyzes the causes of the core defects existing in the existing bid evaluation base supervision technology, traces the limitations of the technical architecture and design logic, and then proposes the following solutions:

[0037] 1. Serious data island problem, multiple systems cannot be linked

[0038] Defect performance: video monitoring, business management, and Internet of Things equipment (such as access control and storage cabinet) belong to independent systems, and data cannot be interchanged. For example, personnel reporting information needs to be manually entered into the business system and cannot be automatically synchronized to the access control system for permission verification. The out-of-bound behavior captured by the camera cannot be associated with the personnel role in the business system, resulting in isolated alarm information.

[0039] Cause analysis: The traditional system design does not adopt a unified data interface standard or middleware technology. Each subsystem is developed by different suppliers, and the protocols are not compatible, resulting in data fragmentation. In addition, there is no centralized data processing platform to realize multi-source data fusion.

[0040] 2. Low level of intelligence, relying on manual intervention

[0041] Defect performance: abnormal behavior (such as personnel out-of-bound, strangers) completely relies on manual patrol or post-event playback to find, and the response lags for several hours or even several days. For example, the camera can only record the picture and cannot automatically identify the violation behavior. The device state anomaly (such as the storage cabinet not closed) needs to be manually inspected to be aware.

[0042] Cause analysis: The existing technology does not integrate AI algorithms (such as computer vision and behavior analysis model). The video stream is only used as storage data and is not analyzed in real time. At the same time, the Internet of Things equipment only provides basic state reporting function and lacks intelligent threshold setting and linkage rule engine, resulting in that the data value has not been tapped.

[0043] 3. Weak visualization capability, lack of dynamic three-dimensional display

[0044] Defect performance: The monitoring screen is displayed in a two-dimensional plane, which cannot intuitively present the dynamic trajectory of personnel, the distribution of equipment and the real-time state. For example, the manager cannot quickly locate the current position of a certain expert or view the historical activity path.

[0045] Cause analysis: The traditional technology is limited by the problems of high cost of three-dimensional modeling and insufficient real-time rendering computing power, and does not introduce digital twin technology. The monitoring system only relies on camera screen splicing and does not construct a virtual model corresponding to the physical space, resulting in single data display dimension.

[0046] 4. Insufficient real-time performance, and lagging alarm mechanism

[0047] Defect performance: The alarm event relies on manual discovery or post-playback and cannot be pushed in real time. For example, when a person violates the rules and enters a restricted area, the system cannot immediately trigger an alarm and the manager needs to actively check the monitoring screen. Device abnormalities (such as network interruption) can only be discovered by periodic polling, which has a delay of several minutes to several hours.

[0048] Cause analysis: The existing system lacks real-time data processing capability, and video stream and IoT data are not preprocessed through edge computing nodes, resulting in high load and slow response of the central server. At the same time, an event-driven alarm triggering mechanism is not designed, relying on periodic data pulling, which cannot achieve a response within seconds.

[0049] Defect root summary: The traditional technology is designed with "independent development of function modules" as the core, lacks a top-level architecture for cross-system collaboration, and does not introduce key technologies such as digital twin, AI algorithm, and edge computing, resulting in fragmented data, passive analysis, single display, and slow response. These defects jointly restrict the efficiency and accuracy of the evaluation base supervision.

[0050] The above comprehensive analysis of the existing evaluation base supervision technology provides a clear improvement direction for the concept of the present invention, which will be described in detail through specific embodiments.

[0051] Embodiment 1

[0052] This embodiment first provides an evaluation base full-factor real-time linkage supervision method based on digital twin and Internet of Things. This method addresses the core defects of the existing evaluation base supervision technology and primarily solves the problem of full-factor real-time linkage supervision, i.e., through multi-source data fusion, AI intelligent analysis, and three-dimensional visualization technology, the dynamic correlation and real-time supervision of personnel, equipment, and events are achieved.

[0053] This is achieved through the following aspects:

[0054] ①Solve the data island problem of traditional systems and achieve multi-source data fusion;

[0055] ②Construct a three-dimensional digital twin model to support dynamic visualized supervision;

[0056] ③ Through AI algorithm to automatically analyze abnormal behavior, reduce manual intervention;

[0057] ④ Build an alarm from identification, traceability, processing of the whole process of closed-loop processing mechanism;

[0058] Data fusion is the basis of linkage, AI analysis is the core of decision-making, three-dimensional visualization is the window of interaction, and alarm closed loop is the guarantee of efficiency. By systematically solving these problems, the method ultimately promotes the paradigm upgrade of bid supervision from "manual screen monitoring, post-tracing" to "intelligent early warning, real-time intervention", providing a technical benchmark for the construction of green digital bid evaluation base.

[0059] For example, the method uses multi-source data fusion, AI intelligent analysis and three-dimensional visualization technology to realize the whole process supervision from data collection to alarm closed loop through the following steps.

[0060] Step 1: Real-time collection and transmission of multi-source data

[0061] The Internet of Things equipment (such as face recognition camera, person certificate reporting machine, intelligent storage cabinet, face recognition access control, intelligent door lock, voice recording telephone) pushes the Internet of Things data to the edge server through MQTT / HTTP protocol in real time. At the same time, the bid evaluation meeting information, personnel white list and other data provided by the business system are synchronized to the server through the API interface. This step solves the problem of data isolation in traditional systems, ensuring that business information, Internet of Things data, personnel permissions, equipment status and other data are associated in real time.

[0062] Optionally, use UWB (Ultra Wide Band) positioning technology or RFID area scanning to replace face recognition camera to realize real-time positioning of personnel. Combined with Bluetooth beacon (Beacon), improve the positioning accuracy in complex environment.

[0063] In terms of data integration capability, the existing technology of video monitoring, business management and Internet of Things equipment belongs to independent systems, and the data formats and protocols are not compatible (such as video stream using RTSP protocol, business system relying on HTTP interface), which leads to the fact that business data and Internet of Things equipment data cannot be associated in real time.

[0064] This embodiment realizes multi-source data fusion through edge server and standardized interface (such as MQTT, API) to build a unified data pool. For example, when a person crosses the boundary, the system can synchronize to call its permission information and access control record to comprehensively determine whether it is a violation, avoiding the misjudgment or omission caused by data isolation in traditional technology.

[0065] Step 2: Dynamic mapping of three-dimensional digital twin model

[0066] Based on UE5 engine, high-precision three-dimensional model is built, real-time data in physical environment (such as personnel location, equipment information) is dynamically mapped, and panoramic tracing and space query (such as clicking the interaction button to show the floor structure, equipment layout, four isolation areas, and clicking the alarm to view the associated video) are supported through the interactive interface.

[0067] Optionally, a three-dimensional digital twin model is built using UE, Unity, Blender, or other engines, or a three-dimensional visualization is realized based on a browser using WebGL lightweight rendering technology.

[0068] In terms of visualization interaction capability, the existing technology mainly uses two-dimensional plane monitoring pictures, which cannot intuitively display personnel trajectories, equipment distribution, and spatial relationships, and management personnel need to switch between multiple systems to query information.

[0069] This embodiment converts abstract data into spatialized and dynamic visual display through a three-dimensional digital twin model. For example, an alarm event can be automatically located to a specific location in a three-dimensional scene, and clicking can view associated videos, trajectory routes, etc., significantly improving information acquisition and decision-making efficiency.

[0070] Step 3: Multi-source data fusion and AI intelligent analysis

[0071] The edge server cleanses and standardizes the collected data and uses AI image recognition algorithms to identify abnormal behaviors (such as rule violations and strangers) in real time. After identifying abnormal events, the system outputs the identification results (event type, event, location, and personnel).

[0072] AI intelligent analysis can obtain a set of multi-frame personnel behavior features by dividing video stream data from time and space dimensions; input the set of multi-frame personnel behavior features into a trained abnormal behavior recognition model; the multi-modal feature encoding layer of the abnormal behavior recognition model performs cross-modal association and fusion on personnel category features, personnel permission features, and isolation area features; during real-time monitoring of video stream data, when the target behavior feature is identified as an abnormal behavior, an alarm event is triggered.

[0073] In terms of abnormal detection methods, traditional technologies rely on manual patrols or post-event playback to discover abnormalities, which have a lagging response and are easily influenced by subjective factors. For example, personnel rule violations can only be identified by continuous screen monitoring by management personnel, and equipment failures can only be discovered through regular inspections.

[0074] This embodiment introduces AI algorithms to analyze video streams and IoT data in real time, enabling automatic identification and classification of abnormal behaviors. For example, based on computer vision, boundary crossing detection not only reduces human intervention but also improves alarm accuracy through multi-dimensional data cross-validation (such as combining personnel permissions and location information).

[0075] Step 4: Real-time alarm triggering and closed-loop management

[0076] After AI identifies the anomaly, the system automatically triggers an alarm and pushes it to the manager, while associating video clips, personnel trajectories, and other data before and after the alarm occurs. The manager locates the event through the three-dimensional model and marks the processing result, forming a closed-loop feedback.

[0077] The voice broadcast warning function can be added to remind the violator in real time through the on-site broadcasting equipment, or other alarm message pushing mechanisms can be added, integrating third-party alarm platforms to support multi-terminal collaborative processing.

[0078] In terms of alarm management mechanism, traditional alarms only record events, and the processing process relies on manual coordination, lacking closed-loop feedback. For example, after the alarm occurs, the video needs to be manually retrieved and the on-site personnel needs to be contacted for verification, resulting in processing delay.

[0079] This embodiment realizes automatic alarm triggering, data association and processing closed loop through event-driven mechanism. For example, the alarm event can automatically associate the 15-second video clips before and after the event, the historical trajectory of personnel and the device log. The manager can locate and mark the processing result through the three-dimensional model, forming a complete supervision evidence chain to ensure that the event is traceable and the responsibility is accountable.

[0080] This embodiment realizes a number of innovations in the evaluation base supervision field through the deep integration of digital twinning and Internet of Things technology, systematically solves the limitations of traditional solutions, and realizes significant technical breakthroughs and application value improvement.

[0081] 1. Full-factor digital twinning modeling and dynamic mapping

[0082] Based on the UE5 engine, a high-precision three-dimensional model is built to digitize and restore the floor structure, device layout, and four-isolation areas of the physical evaluation base. Personnel location, device status, activity data, etc. are dynamically mapped to the virtual space through real-time data interface.

[0083] Traditional systems rely on flat monitoring screens and require manual switching between multiple systems to query information, which is inefficient. This technical solution converts abstract data into spatial display, allowing managers to quickly understand various information in complex scenarios. This solution supports panoramic tracking and interactive query through the three-dimensional model, such as clicking on an alarm event to view associated video, display real-time location or historical trajectory of personnel, and display device distribution for Internet of Things data. The solution highlights the four-isolation control areas on site, enabling managers to quickly understand the activity dynamics on site and perform efficient and accurate management.

[0084] 2. Real-time fusion of multi-source data

[0085] Through the edge server integration of business systems, Internet of Things equipment (such as personnel certificate reporting machines, intelligent storage cabinets, face recognition access control, intelligent door locks, and voice recording telephones), and video stream data, a unified data pool is formed.

[0086] In traditional technologies, video monitoring, business management, and Internet of Things equipment operate independently, and data protocols are not compatible (such as video stream using RTSP and business systems relying on HTTP), resulting in scattered data and the inability to effectively integrate for business supervision applications. The technical solution effectively integrates various data through standardized interfaces and data fusion technology, such as camera shooting of face information, AI service for face recognition to determine personnel identity, and business service to define whether personnel are in violation based on personnel identity and permission. In summary, multiple data sources such as Internet of Things equipment and business services are integrated to comprehensively determine abnormal events.

[0087] After data fusion, AI algorithms cross-verify multidimensional data (such as combining personnel permissions and location information), significantly improving the accuracy and timeliness of abnormal detection. Alarm events can automatically associate complete evidence chains (such as previous and subsequent video clips and device logs), achieving closed-loop management from discovery to processing, significantly reducing the need for manual intervention, and significantly improving supervision efficiency.

[0088] 3. AI intelligent analysis and violation supervision

[0089] Based on AI image recognition, real-time analysis of bid evaluation site video streams is performed to automatically identify personnel violations (such as strangers and violations of crossing zones), and to trigger alarms to prompt relevant supervisors to locate, verify, and handle violation events.

[0090] Traditional technologies rely on simple video monitoring or manual patrols for violation event supervision, which is prone to supervision gaps due to insufficient or negligent human resources. The technical solution analyzes and processes personnel activities on site through AI, and outputs alarms when abnormalities are identified. Alarm events can automatically associate complete evidence chains (such as previous and subsequent video clips and device logs), achieving closed-loop management from discovery to processing, significantly reducing the need for manual intervention, and significantly improving the accuracy, timeliness, and efficiency of violation event monitoring.

[0091] This embodiment systematically solves the core problems of data silos, delayed responses, and poor interactivity in traditional bid evaluation supervision technologies. Through digital twin technology to integrate multiple data sources, AI algorithms to achieve automated abnormality identification and alarm linkage, and three-dimensional visualization technology to convert complex data into intuitive spatial information. These innovations not only significantly improve the accuracy, real-time performance, and efficiency of supervision, but also form a full-process automated management from abnormality discovery to closed-loop processing, providing reusable technical paradigms and practical references for the construction of State Grid green digital evaluation bases.

[0092] Embodiment 2

[0093] The embodiment is a full-factor real-time linkage supervision system for bid evaluation ground, comprising:

[0094] A multi-source data real-time acquisition module acquires multi-source data in real time, wherein the multi-source data comprises video monitoring data, business management data and Internet of Things equipment data, and the video monitoring data, the business management data and the Internet of Things equipment data are associated.

[0095] A three-dimensional digital twin model constructs a three-dimensional model, dynamically maps the video monitoring data, the business management data and the Internet of Things equipment data to the three-dimensional model, and forms a three-dimensional digital twin model.

[0096] An abnormal behavior recognition module intelligently analyzes the video monitoring data, the business management data and the Internet of Things equipment data through AI, and recognizes abnormal behavior.

[0097] An abnormal behavior processing module triggers an alarm event after recognizing abnormal behavior through AI, and pushes video monitoring clips, personnel historical trajectories and equipment logs within a preset time period before and after the alarm event to a management personnel; the management personnel locates the alarm event in the three-dimensional digital twin model, verifies and marks a processing result, and forms a closed-loop feedback.

[0098] The embodiment adopts multi-source data real-time fusion and edge band collaborative processing technology as a bottom foundation, integrates a business system (bid evaluation meeting and meeting personnel), Internet of Things equipment (personnel check-in machine, intelligent storage cabinet, face recognition access control, intelligent door lock and recording telephone) and video stream data through an edge band server, realizes real-time cleaning, alignment and association of heterogeneous data through a standardized interface (such as MQTT / API), completely breaks the data island of a traditional system, and provides a unified data pool for subsequent AI analysis and three-dimensional visualization.

[0099] The embodiment constructs a high-precision three-dimensional model relying on UE5 engine, maps personnel positions, equipment states and activity factors in a physical environment to a virtual space in real time, and supports interactive operation (equipment viewing, positioning query and alarm tracing) to convert abstract data into intuitive spatialized display, significantly improves information understanding and decision-making efficiency in a complex scene, is a core interface of user interaction, and is a key point of three-dimensional visualization supervision.

[0100] Through intelligent analysis model, the monitoring pictures, the equipment state, the personnel track are scanned in real time, the potential risk is automatically identified and the alarm is immediately triggered. For example, when detecting that a certain bid evaluation personnel crosses the isolation area, the system not only pushes the alarm information in real time, but also synchronously associates the identity permission, the historical activity track and the monitoring video of the previous and the next 10 seconds of the personnel, forms a complete evidence chain. The management personnel can quickly locate the event position, verify the details and feed back the processing result online through the three-dimensional interface, realizes the whole-process closed-loop management from "risk discovery-precise positioning-collaborative disposal-result tracing". This capability upgrades the traditional passive and fragmented supervision mode to the active early warning and collaborative intelligent management, greatly reduces the manual intervention cost, and improves the supervision response speed and standardization.

[0101] Obviously, the above embodiments are only examples for clearly illustrating but not limiting the embodiments. Based on the above description, other different forms of changes or variations can be made by those skilled in the art. Here, all the embodiments are not required to be exhausted. The obvious changes or variations derived therefrom are still within the protection scope of the present application.

Claims

1. A real-time linkage supervision method for all elements of a bid evaluation base, characterized by: The steps include: Step 1: Real-time collection of multi-source data Collect multi-source data in real time, the multi-source data including video surveillance data, business management data, and IoT device data, and associate the video surveillance data, the business management data, and the IoT device data; Step 2: Build a 3D digital twin model Constructing a three-dimensional model, dynamically mapping the video surveillance data, the business management data, and the IoT device data to the three-dimensional model to form a three-dimensional digital twin model; Step 3: Abnormal behavior identification Performing intelligent analysis of the video surveillance data, the business management data, and the IoT device data through AI to identify abnormal behavior; Step 4: Abnormal behavior handling After AI identifies abnormal behavior, it triggers an alarm event and pushes relevant video surveillance clips, personnel history tracks, and device logs within a preset time period before and after the alarm event to management personnel; The manager locates the alarm event in the three-dimensional digital twin model, verifies and marks the processing results, and forms a closed-loop feedback.

2. The method according to claim 1, characterized in that In step 1, the IoT device data is pushed to the edge proxy server via the MQTT / HTTP protocol, and the business management data is synchronized to the edge proxy server via the API interface; the edge proxy server associates and preprocesses the video surveillance data, the business management data, and the IoT device data.

3. The method according to claim 1, characterized in that In step 1, the IoT device data includes facial recognition cameras, ID check-in machines, smart lockers, facial recognition access control, smart door locks, and recording phones; the business management data includes bid evaluation meeting information and a personnel whitelist.

4. The method according to claim 1, wherein In step 1, the IoT device uses ultra-wideband or Bluetooth beacons to locate personnel in real time.

5. The method according to claim 1, wherein In step 2, the three-dimensional digital twin model performs panoramic tracing and spatial query through an interactive interface.

6. The method according to claim 1, characterized in that In step 2, the three-dimensional digital twin model is built based on UE5 or Unity engine.

7. The method according to claim 1, characterized in that In step 2, the three-dimensional digital twin model is rendered using WebGL and three-dimensional visualization is achieved based on the browser.

8. The method according to claim 1, characterized in that In step 3, the abnormal behavior identification is performed through cross-verification by combining the personnel authority and location information through computer vision cross-border detection.

9. The method according to claim 8, characterized in that In step 3, the video stream data is segmented from the time dimension and the space dimension to obtain a multi-frame personnel behavior feature set; the multi-frame personnel behavior feature set is input into a trained abnormal behavior recognition model; the multimodal feature encoding layer of the abnormal behavior recognition model performs cross-modal correlation and fusion on personnel category features, personnel authority features, and isolation area features; in real-time monitoring of the video stream data, when the target behavior feature is identified as abnormal behavior, an alarm event is triggered.

10. A real-time linkage supervision system for all elements of a bid evaluation base, characterized by: include: A multi-source data real-time acquisition module collects multi-source data in real time, the multi-source data including video surveillance data, business management data and IoT device data, and associates the video surveillance data, the business management data and the IoT device data; A three-dimensional digital twin model is constructed, and the video surveillance data, the business management data, and the IoT device data are dynamically mapped to the three-dimensional model to form a three-dimensional digital twin model; An abnormal behavior identification module, which uses AI to intelligently analyze the video surveillance data, the business management data, and the IoT device data to identify abnormal behavior; The abnormal behavior processing module uses AI to identify abnormal behavior, trigger an alarm event, and push relevant video surveillance clips, personnel historical trajectory, and device logs within a preset time period before and after the alarm event to management personnel; The manager locates the alarm event in the three-dimensional digital twin model, verifies and marks the processing results, and forms a closed-loop feedback.

Images