A fire door monitoring and linkage control system based on vision and pressure perception

Abstract

This invention relates to the field of fire protection and safety management in nuclear power plants, and in particular to a fire door monitoring and linkage control system based on vision and pressure perception. The system includes a perception layer that collects door images and closing pressure data, processes the image and pressure data to identify the fire door's status, and controls its opening and closing. A transmission layer transmits the data from the perception layer to a platform layer and forwards control commands. The platform layer aggregates and fuses the image and closing pressure data from the perception layer to comprehensively determine the status of each fire door and provides a visualization platform. The application layer displays the visualization platform to the user, allowing the user to remotely operate the fire doors. The perception layer of this system employs a dual-mode fusion technology of machine vision and pressure perception, achieving accurate identification and dual verification of the fire door's status. This overcomes the shortcomings of single sensors, which are prone to false alarms under complex lighting or obstruction conditions, significantly improving the reliability and accuracy of status judgment.

Description

Technical Field

[0001] This invention relates to the field of fire protection and safety management in nuclear power plants, and in particular to a fire door monitoring and linkage control system based on vision and pressure perception. Background Technology

[0002] Currently, fire doors in locations such as nuclear power plants primarily rely on mechanical door closers to maintain their status, and are managed through manual inspections. However, this traditional method has several drawbacks: mechanical door closers are prone to fatigue, corrosion, or damage from external forces after prolonged use, causing fire doors to fail to close properly; manual inspections are inefficient, have a high rate of missed inspections, and struggle to achieve real-time, comprehensive monitoring; the status information of each fire door is scattered, making centralized management impossible and leading to delayed emergency response; furthermore, existing systems lack linkage mechanisms with fire alarm and smoke control systems, failing to automatically close fire doors during a fire and delaying optimal response times. Therefore, a fire door status management system capable of real-time monitoring, centralized control, and intelligent linkage is needed. Summary of the Invention

[0003] This invention provides a fire door monitoring and linkage control system based on vision and pressure perception, which solves the technical problems of fire doors relying on manual inspection, delayed status judgment, inability to be centrally controlled, and lack of linkage response with fire protection systems in the prior art.

[0004] The technical solution of the present invention is as follows: This invention proposes a fire door monitoring and linkage control system based on vision and pressure perception. The system includes a perception layer, a transmission layer, a platform layer, and an application layer. The perception layer collects images of the door and closing pressure data, processes the image data and pressure data, identifies the status of the fire door, and controls the opening and closing of the fire door. The transmission layer transmits the data from the perception layer to the platform layer and forwards control commands. The platform layer aggregates and merges the image data and closing pressure data from the perception layer, comprehensively judges the status of each fire door, and provides a visualization platform. The application layer displays the visualization platform of the platform layer to the user, allowing the user to remotely operate the fire door.

[0005] In some embodiments, the sensing layer is deployed at the fire door site. The sensing layer includes a machine vision module, a pressure sensing module, a machine vision camera, and a pressure sensor. The machine vision camera and pressure sensor are deployed on the fire door frame and the upper part of the door body.

[0006] In some embodiments, the machine vision camera captures images of the door in real time, and the machine vision module performs edge computing through a lightweight AI model to identify the opening and closing status, opening angle, and obstruction of the fire door by foreign objects.

[0007] In some embodiments, the machine vision module identifies the opening and closing status, opening angle, and obstruction of fire doors. Specifically, this includes: the machine vision module preprocessing the image to enhance image contrast; the machine vision module using a lightweight AI model to detect the main body region of the fire door in the preprocessed image, outputting the detection box coordinates and confidence score to obtain the fire door coordinates; the machine vision module calculating the aspect ratio of the detected fire door frame and determining the state of the fire door based on the aspect ratio; for fire doors in a half-open state, the machine vision module extracting the feature vector of the fire door detection box, inputting it into a pre-trained support vector machine regression model, and outputting the opening angle of the door relative to the door frame; and the machine vision module inputting the fire door detection box region into a transfer learning model to identify the fire door region image, outputting the probability of the presence of foreign objects, the category of foreign objects, and pixel coordinates.

[0008] In some embodiments, the sensing layer is further provided with an intelligent control execution module, which is installed on the door hinge side, receives instructions from the platform layer to execute the opening and closing operation of the fire door, and provides feedback on the status through an angle sensor.

[0009] In some embodiments, a pressure sensor is integrated into the fire door frame sealing strip. The pressure sensing module reads the data from the pressure sensor to monitor the contact pressure when the door is closed in real time, so as to determine whether the fire door is completely closed.

[0010] In some embodiments, the transport layer uses a high-speed bus or wireless communication method to transmit data from the perception layer to the platform layer and forward control commands from the platform layer.

[0011] In some embodiments, the centralized management platform controls the intelligent control execution module to perform fire door opening and closing operations based on the fire door status or user instructions; the platform layer includes the centralized management platform, which receives and fuses image data and closing pressure data from the perception layer to determine the fire door status; the centralized management platform includes a data fusion analysis module, which uses a weighted confidence fusion algorithm to comprehensively determine the door status, specifically including the centralized management platform receiving visual recognition results and pressure sensing data from the perception layer, aligning the visual recognition results and pressure sensing data in time, and calculating a normalized pressure value; the centralized management platform defines and calculates the comprehensive door status function, specifically as shown in formula (1). F(S_v,P_norm, θ_v) =w_1·C_v(S_v)+ w_2·P_norm+w_3·(1-θ_v / 120) (1) Where w_1, w_2, and w_3 are weight coefficients; C_v(S_v) is the visual state encoding value; P_norm is the normalized stress value; θ_v is the opening angle; and S_v is the opening / closing state. The centralized management platform determines the fault status of fire doors, specifically including: when the visual recognition confidence score is below 0.75 and the pressure data exceeds 30% of the historical average, the centralized management platform determines it to be a sensor fault state; when the visual fire door opening / closing status and pressure status are inconsistent more than 5 times within 10 consecutive seconds, the centralized management platform determines it to be a door mechanical fault state; when the deviation between the angle sensor feedback value of the intelligent control execution module and the visual prediction angle is greater than 15° for more than 5 seconds, it is determined to be an angle sensor offset fault; the centralized management platform periodically calculates the independent judgment accuracy of the vision module and the pressure module based on historical data and dynamically adjusts the weight coefficients w_1, w_2, and w_3.

[0012] In some embodiments, a visualization platform is set up at the platform layer. The visualization platform supports the visualization display of fire doors, status query, abnormal alarm, historical records, and data report generation.

[0013] In some embodiments, the application layer provides web and mobile interfaces, allowing administrators to monitor in real time, receive alarms, and remotely and manually issue fire door control commands to the platform layer.

[0014] The implementation of this invention has the following beneficial effects: 1. This invention proposes a fire door monitoring and linkage control system based on vision and pressure perception. The system's perception layer employs a dual-mode fusion technology of machine vision and pressure perception, achieving accurate identification and dual verification of fire door status. This overcomes the shortcomings of single sensors, which are prone to false alarms under complex lighting or obstruction conditions, significantly improving the reliability and accuracy of status judgment. By deploying an edge computing AI model at the perception layer, local real-time identification of fire door status and on-site alarm for anomalies are achieved, greatly reducing data transmission latency and network bandwidth pressure, and solving the response lag problem caused by traditional solutions relying on cloud processing.

[0015] 2. This invention proposes a fire door monitoring and linkage control system based on vision and pressure perception. The system's platform layer constructs a centralized plant-wide management platform and achieves "one-map" visualization, changing the traditional management model of scattered manual inspections and information silos. It realizes unified monitoring, historical tracing, and alarm management of the status of thousands of fire doors, greatly improving emergency command efficiency and overall control capabilities. The platform layer establishes intelligent linkage with fire alarm systems and smoke extraction systems, enabling automatic closure of fire doors in relevant areas and real-time status feedback after fire confirmation. This compensates for the shortcomings of existing systems that lack linkage mechanisms and rely on manual intervention, buying valuable time for fire prevention and control.

[0016] 3. This invention proposes a fire door monitoring and linkage control system based on vision and pressure perception. The perception layer of this system is equipped with an intelligent control execution module and supports remote operation and automatic return technology, realizing remote opening and closing control of fire doors and automatic reset in fault conditions, reducing the difficulty of on-site operation and maintenance and the risk of personnel exposure, and improving the overall intelligence and automation level of the system. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of a fire door monitoring and linkage control system based on vision and pressure perception proposed in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the core functions of a fire door monitoring and linkage control system based on vision and pressure perception, as proposed in an embodiment of the present invention. Figure 3 This invention presents a flowchart of a fire door fault closing process for a fire door monitoring and linkage control system based on vision and pressure perception, as proposed in an embodiment of the present invention. It illustrates the complete processing flow from data acquisition at the perception layer, fusion and judgment at the platform layer, to alarm push and remote control at the application layer. Figure 4 This invention presents a flowchart illustrating the fire door closing process in a fire situation using a fire door monitoring and linkage control system based on vision and pressure perception, as proposed in an embodiment of the present invention. The flowchart shows the fire emergency response process from the triggering of the fire alarm system signal, the execution of the linkage strategy at the platform layer, the batch door closing control at the perception layer, to the status display at the application layer. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] like Figures 1 to 4 As shown, this invention proposes a fire door monitoring and linkage control system based on vision and pressure perception, specifically including a perception layer, a transmission layer, a platform layer, and an application layer, to realize closed-loop management of data acquisition, transmission, processing, and control.

[0020] like Figure 1As shown, the perception layer is deployed at the fire door site and includes a machine vision module, a pressure sensing module, a machine vision camera, a pressure sensor, and an intelligent control execution module. It collects door images and closing pressure data, and performs local preprocessing and AI inference at the edge terminal. The transmission layer consists of a high-speed bus or wireless communication module, responsible for uploading the perceived data to the platform layer and forwarding control commands. The platform layer, or central control platform, aggregates and integrates visual and pressure data, using a weighted confidence fusion algorithm to comprehensively determine the status of each fire door, and also interfaces with third-party platforms such as fire alarm systems and smoke extraction systems. The application layer is for on-duty personnel, providing a web-based monitoring screen and a mobile app for visualized display and remote operation.

[0021] The perception layer deploys machine vision cameras and pressure sensors on the upper part of the fire door frame and body. The cameras acquire door images in real time at a rate of no less than 25 frames per second. The machine vision module incorporates a lightweight AI model based on an improved YOLOv8-nano. This model uses depthwise separable convolutions instead of standard convolutions for feature extraction, reduces the number of channels to 1 / 4 of YOLOv8-s, compresses the model parameters to 1.8M, and after INT8 quantization, the model size does not exceed 2.5MB, making it compatible with ARM Cortex-A53 and above edge computing devices. The machine vision module reads the video stream data from the vision cameras and performs local inference at the edge, with a single-frame inference latency of no more than 80ms, reducing transmission pressure and improving response speed. Specifically, the image processing and recognition algorithm of the machine vision module includes the following steps: Step 1: Image preprocessing. The machine vision module normalizes the size of the acquired RGB image to 640×640 pixels and uses adaptive histogram equalization (CLAHE, clipLimit=2.0, tileGridSize=8×8) to enhance image contrast in order to overcome the influence of the complex lighting environment of the nuclear power plant.

[0022] Step 2: Door detection and segmentation. The machine vision module uses a lightweight AI model to detect the main body area of ​​the fire door in the image, and outputs the coordinates of the detection box (x_min, y_min, x_max, y_max) and the confidence score. The detection confidence threshold is set to 0.75, and frames below this threshold are discarded.

[0023] Step 3: Open / Closed State Recognition. The machine vision module calculates the aspect ratio R of the detection box = (y_max - y_min) / (x_max - x_min). When R ≥ 3.0, it is determined to be in the closed state; when 1.2 ≤ R < 3.0, it is determined to be in the half-open state; and when R < 1.2, it is determined to be in the fully open state.

[0024] Step 4: Calculate the opening angle. In the half-open state, the machine vision module extracts the gradient histogram (HOG) feature vector of the door detection box and inputs it into a pre-trained support vector machine (SVM) regression model (kernel function is RBF, C=10, gamma=0.001). The output is the opening angle θ of the door relative to the door frame. The angle resolution is 1° and the effective range is 0°-120°. This SVM model is trained based on 5000 labeled samples, and the mean square error (MSE) of the angle prediction does not exceed 2.5°.

[0025] Step 5: Obstruction Detection. The machine vision module crops the door detection box area from the original image and inputs it into a secondary classification network based on the MobileNetV3-small transfer learning model. This network performs transfer learning based on ImageNet pre-trained weights and outputs the probability of obstruction P_occ. When P_occ ≥ 0.6, obstruction is determined to exist. Simultaneously, the obstruction category (e.g., obstacle, pile, device) and its pixel coordinates within the door area are output for precise localization of the obstruction region. The above edge computing process runs on the NVIDIA Jetson Nano or Rockchip RK3588 edge computing unit in the perception layer, using TensorRT or RKNN to accelerate inference frameworks. GPU / NPU utilization is controlled below 80% to ensure real-time performance while reserving computing resources for data preprocessing.

[0026] A pressure sensor is integrated into the door frame sealing strip. The pressure sensing module reads the data from the pressure sensor to monitor the contact pressure when the door closes in real time, in order to determine whether the door is fully closed. Data from the machine vision module and the pressure sensing module are pre-processed locally and then uploaded through the transmission layer. The sensing layer also has an intelligent control execution module, which consists of an electric door closer, an electromagnetic release, a servo motor drive unit, an angle sensor (encoder type, 0.5° resolution), and a local controller (STM32F407 microcontroller, 168MHz). The intelligent control execution module is installed on the door hinge side, receives remote commands or linkage signals from the platform layer, executes the opening and closing operations of the fire door, and feeds back the status through the angle sensor, forming a closed-loop management system.

[0027] The transport layer uses a high-speed bus or wireless communication method to ensure low-latency and highly reliable data transmission, and supports the issuance of control commands.

[0028] The platform layer is a centralized control platform. This platform receives and integrates image data and closure pressure data from the sensing layer to make a comprehensive status judgment. Based on the fire door's status or user commands, the centralized control platform controls the intelligent control execution module to perform fire door opening and closing operations. The platform can also automatically issue control commands to the fire door according to preset strategies.

[0029] The centralized control platform includes a data fusion and analysis module. This module uses a weighted confidence fusion algorithm to comprehensively determine the door's status, fusing visual recognition results with pressure sensing data to eliminate misjudgments from a single sensor. The specific determination method is as follows: Step 1: Data Alignment and Normalization. The centralized management platform receives visual recognition results from the perception layer (including open / closed status S_v∈{closed, half-open, fully open}, opening angle θ_v, foreign object obstruction marker O_v∈{0,1}, and various confidence scores) and pressure sensing data (including multi-point pressure values ​​of the sealing strip P_1, P_2, ..., P_n, n≥3, and average pressure P_avg). First, time alignment is performed. Using the visual data timestamp t_v as a baseline, the nearest neighbor method is used to match the pressure data timestamp t_p, requiring |t_v - t_p| ≤ 500ms. Then, pressure normalization is performed, calculating the normalized pressure value P_norm = (P_avg - P_min) / (P_max - P_min), where P_max and P_min are the historical maximum and minimum pressure records for the fire door, respectively.

[0030] Step 2: State Comprehensive Judgment Algorithm: The centralized control platform defines and calculates the comprehensive state function of the gate, as shown in formula (1). F(S_v,P_norm, θ_v) =w_1·C_v(S_v)+ w_2·P_norm+w_3·(1-θ_v / 120) (1) Where w_1, w_2, and w_3 are weighting coefficients, satisfying w_1 + w_2 + w_3 = 1, with typical values ​​of w_1 = 0.4, w_2 = 0.35, and w_3 = 0.25; C_v(S_v) is the visual state encoding value, where C_v = 1.0 when S_v = off, C_v = 0.5 when S_v = half-open, and C_v = 0 when S_v = fully open. The judgment thresholds for the comprehensive state score F are as follows: when F ≥ 0.85, it is judged as a normal off state; when 0.50 ≤ F < 0.85, it is judged as a partially covered state (visual display is off but pressure is insufficient or angle is too large); when 0.20 ≤ F < 0.50, it is judged as a half-open state; and when F < 0.20, it is judged as a fully open state.

[0031] The centralized management platform determines the fault status of fire doors, specifically including: when the visual recognition confidence score is below 0.75 and the pressure data shows abnormal fluctuations (variance σ_p exceeds 30% of the historical average), the centralized management platform determines it to be a sensor fault state; when the visual state S_v and the pressure state S_p (pressure state S_p is derived from P_norm: P_norm≥0.8 indicates closed, 0.4≤P_norm<0.8 indicates partially closed, and P_norm<0.4 indicates open) are inconsistent more than 5 times within 10 consecutive seconds, it is determined to be a mechanical fault state of the door (such as door closer failure or door deformation); when the deviation |θ_a - θ_v| between the angle sensor feedback value θ_a and the visual predicted angle θ_v is greater than 15° for more than 5 seconds, it is determined to be an angle sensor offset fault. In case of a fault, the system automatically generates fault codes (E01 - vision sensor fault, E02 - pressure sensor fault, E03 - door closer mechanical fault, E04 - angle sensor offset) and pushes them to the application layer alarm.

[0032] The centralized management platform integrates weights and adjusts them adaptively. Based on historical data, the platform periodically (weekly) calculates the independent judgment accuracy rates Acc_v and Acc_p of the vision module and the pressure module, and dynamically adjusts the weight coefficients w_1 = Acc_v / (Acc_v + Acc_p + Acc_θ) and w_2 = Acc_p / (Acc_v + Acc_p + Acc_θ), where Acc_θ is the historical effective data rate of the angle sensor. This ensures that the fusion algorithm prioritizes and trusts sensor data with higher reliability, continuously improving the accuracy of the comprehensive judgment.

[0033] The platform layer includes a visualization platform developed using a B / S architecture. The front-end is based on the Vue.js 3.0 framework combined with the ECharts 5.0 charting library, while the back-end uses a Spring Boot microservice architecture. The database uses InfluxDB time-series database to store historical status data and PostgreSQL relational database to store device configurations and alarm records. The specific functions of the visualization platform include: (1) Visualization: Using the nuclear power plant building floor plan as the base map, the icons of fire door locations are rendered using SVG vector graphics. The icon color code indicates the real-time status (green - normally closed, yellow - partially closed, red - open / faulty, gray - offline). It supports 7 levels of scaling from 1:500 to 1:50. Clicking the icon will pop up a detailed information floating window displaying the door number, location, current status, most recent detection time, pressure value, opening angle, obstruction information, and real-time video preview link.

[0034] (2) Status query: Supports querying by combination of conditions such as fire door number, area (factory / floor / room number), status type (closed / ajar / open / fault), and time range. The query results are displayed in a table format with 50 records per page. It can be exported to an Excel file. The table columns include fields such as door number, area location, current status, status duration, average pressure, angle value, most recent alarm time, and processing records.

[0035] (3) Abnormal alarm: When the centralized control platform determines that the door status is ajar, open or faulty, the visualization display platform triggers a three-level alarm mechanism. Level 1 alarm (ajar for more than 5 minutes) is prompted by a yellow flashing icon on the platform interface and the log is recorded. Level 2 alarm (open or ajar for more than 15 minutes) pushes a real-time notification to the duty terminal via WebSocket and plays a prompt sound. Level 3 alarm (fire linkage failure or fault status) triggers an audible and visual alarm and automatically sends an SMS and pushes the alarm to the designated management personnel via APP. The alarm information includes the door location, abnormality type, real-time screenshot and suggested handling measures.

[0036] (4) Historical records: The system automatically stores each status change record, including the status before the status change, the status after the change, the change time, the triggering reason (automatic detection / manual operation / linkage control), and the operator ID (if applicable). The record retention period is no less than 3 years, and it supports playback of the status change process of a specified fire door by time axis.

[0037] (5) Data report generation: Supports automatic generation of fire door status statistical reports by day / week / month / quarter. The report content includes total number of inspected doors, normal closing rate, number of times the door is ajar and the average duration, number of opening events, number of failures and MTBF (mean time between failures), ranking of the status qualification rate of each area, and year-on-year and month-on-month data compared with the previous period. The report format supports PDF and Excel export and can be automatically sent to the management personnel via email on a regular basis.

[0038] The centralized control platform supports cross-system linkage with fire alarm systems and smoke control systems. Specifically, the platform interfaces with the fire alarm system (FAS) via the standard Modbus TCP / IP protocol or OPC UA interface, receiving signals from fire detectors in real time. When the FAS sends a fire alarm confirmation signal (a valid fire alarm lasting more than 2 seconds, not a false alarm), the platform parses the area code in the signal, matches all fire door IDs within that area, generates a batch of door closing command queues, and distributes them to each sensing layer via the transport layer with a priority QoS=1 (at least one delivery guarantee). Simultaneously, the platform sends a linkage request to the smoke control system via a RESTful API. The request includes the fire alarm area code, the fire door closing status query interface address, and the expected complete closure timestamp. The smoke control system dynamically adjusts the start / stop sequence of smoke exhaust fans and the opening strategy of smoke exhaust vents based on the progress of fire door closure, achieving coordinated action between fire door closure and the smoke exhaust system. During the coordinated operation, the visualization platform displays a progress bar on the command screen, showing the number of instructions issued, the number of confirmed closures, the number of execution failures, and the estimated completion time. After the coordinated operation ends, a coordinated execution report is automatically generated and archived.

[0039] The application layer provides web and mobile interfaces, making it convenient for administrators to monitor in real time, receive alarms, and perform remote manual control.

[0040] This invention proposes a fire door monitoring and linkage control system based on vision and pressure perception. The specific implementation of this system is as follows: like Figure 2 As shown, the core functional modules of the system work collaboratively. The machine vision module incorporates a lightweight AI model based on an improved YOLOv8-nano, combining HOG feature extraction and SVM regression to identify the opening and closing status, opening angle, and obstruction of fire doors in real time at the edge. The pressure sensing module is integrated into the door frame sealing strip, using multi-point pressure sensors to monitor the door closing pressure in real time. The data fusion module uses a weighted confidence fusion algorithm at the platform layer to align, normalize, and comprehensively judge the visual results and pressure values, effectively avoiding false alarms from single sensors. The intelligent control execution module consists of an electric door closer, an electromagnetic release, a servo motor drive unit, an angle sensor, and a local controller, installed on the door hinge side. It receives commands from the platform layer to execute opening and closing operations and provides status feedback through the angle sensor. The centralized management platform integrates all data and functions, supporting status monitoring, anomaly alarms, linkage strategy configuration, hierarchical access control, and visual display.

[0041] like Figure 3As shown, in daily monitoring scenarios, the perception layer continuously collects visual and pressure data and uploads it to the platform layer via the transmission layer for fusion and judgment. When the centralized control platform detects that the door is not fully closed (e.g., visual recognition indicates it is closed but the pressure value is below the closure threshold), the platform layer triggers an alarm and pushes it to the on-duty personnel via the application layer. After real-time video verification by the on-duty personnel through the application layer, they can remotely issue a door-closing command. This command reaches the intelligent control execution module of the perception layer via the transmission layer to execute the door-closing action. The platform layer receives feedback from the angle sensor, updates the status, and deactivates the alarm. All data throughout the process is automatically recorded in the platform layer database.

[0042] like Figure 4 As shown, in a fire linkage scenario, after confirming a fire, the fire alarm system sends a signal to the centralized control platform at the platform layer. Based on a preset linkage strategy, the centralized control platform sends closing commands in batches through the transmission layer to the sensing layers corresponding to all fire doors in the relevant area. Upon receiving the commands, the intelligent control execution modules of each door drive the fire doors to close and provide real-time feedback on the closing status to the platform layer via angle sensors. The visualization display platform at the platform layer dynamically displays the status of each door on the command screen. For doors that fail to close, an alarm is triggered again, and manual intervention is prompted through the application layer. All process data is automatically recorded and archived to the platform layer database.

[0043] Through the above-mentioned architecture design and module collaboration, this invention achieves real-time and accurate perception of fire door status, intelligent alarm for abnormalities, reliable remote control, and seamless linkage with the fire protection system, effectively solving the problems of existing technologies that rely on manual inspection, scattered information, and inability to respond automatically.

[0044] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A fire door monitoring and linkage control system based on vision and pressure perception, characterized in that, The system comprises a sensing layer, a transmission layer, a platform layer, and an application layer. The sensing layer collects images of the fire doors and closure pressure data, processes the image and pressure data, identifies the status of the fire doors, and controls their opening and closing. The transmission layer transmits the data from the sensing layer to the platform layer and forwards control commands. The platform layer aggregates and merges the image and closure pressure data from the sensing layer to comprehensively determine the status of each fire door and provides a visualization platform. The application layer displays the visualization platform of the platform layer to the user, allowing the user to remotely operate the fire doors.

2. The fire door monitoring and linkage control system based on vision and pressure perception according to claim 1, characterized in that, The sensing layer is deployed at the fire door site. The sensing layer includes a machine vision module, a pressure sensing module, a machine vision camera, and a pressure sensor. The machine vision camera and pressure sensor are deployed on the fire door frame and the upper part of the door body.

3. The fire door monitoring and linkage control system based on vision and pressure perception according to claim 2, characterized in that, The machine vision camera captures images of the door in real time, and the machine vision module performs edge computing through a lightweight AI model to identify the opening and closing status, opening angle, and obstruction of the fire door by foreign objects.

4. The fire door monitoring and linkage control system based on vision and pressure perception according to claim 3, characterized in that, The machine vision module identifies the opening and closing status, opening angle, and obstruction of fire doors. Specifically, this includes: preprocessing the image to enhance contrast; using a lightweight AI model to detect the main body area of ​​the fire door in the preprocessed image, outputting the detection box coordinates and confidence score to obtain the fire door coordinates; calculating the aspect ratio of the detected fire door frame to determine the fire door's state; for fire doors in a half-open state, extracting the feature vector of the fire door detection box, inputting it into a pre-trained support vector machine regression model, and outputting the opening angle of the door relative to the frame; and inputting the fire door detection box area into a transfer learning model to identify the fire door area image, outputting the probability of foreign object presence, the foreign object category, and pixel coordinates.

5. A fire door monitoring and linkage control system based on vision and pressure perception according to claim 4, characterized in that, The perception layer is also equipped with an intelligent control execution module, which is installed on the door hinge side. It receives instructions from the platform layer to execute the opening and closing operation of the fire door and feeds back the status through the angle sensor.

6. The fire door monitoring and linkage control system based on vision and pressure perception according to claim 1, characterized in that, The pressure sensor is integrated into the sealing strip of the fire door frame. The pressure sensing module reads the data from the pressure sensor to monitor the contact pressure when the door is closed in real time, so as to determine whether the fire door is completely closed.

7. A fire door monitoring and linkage control system based on vision and pressure perception according to claim 1, characterized in that, The transmission layer uses a high-speed bus or wireless communication method to transmit data from the perception layer to the platform layer and forward control commands from the platform layer.

8. A fire door monitoring and linkage control system based on vision and pressure perception according to claim 5, characterized in that, The centralized management and control platform controls the intelligent control execution module to perform fire door opening and closing operations according to the fire door status or user instructions; the platform layer includes a centralized management and control platform, which receives and integrates the image data and closing pressure data from the sensing layer to determine the fire door status; the centralized management and control platform includes a data fusion analysis module, which uses a weighted confidence fusion algorithm to comprehensively determine the door status, specifically including the centralized management and control platform receiving visual recognition results and pressure sensing data from the sensing layer, aligning the visual recognition results and pressure sensing data in time, and calculating a normalized pressure value; the centralized management and control platform defines and calculates a comprehensive door status function, specifically as shown in formula (1). F(S_v,P_norm, θ_v) =w_1·C_v(S_v)+ w_2·P_norm+w_3·(1-θ_v / 120) (1) Where w_1, w_2, and w_3 are weight coefficients; C_v(S_v) is the visual state encoding value; P_norm is the normalized stress value; θ_v is the opening angle; and S_v is the opening / closing state. The centralized management platform determines the fault status of fire doors in the following ways: when the visual recognition confidence score is below 0.75 and the pressure data exceeds 30% of the historical average, the centralized management platform determines it to be a sensor fault state; when the visual fire door opening / closing status and pressure status are inconsistent more than 5 times within 10 consecutive seconds, the centralized management platform determines it to be a door mechanical fault state; when the deviation between the angle sensor feedback value of the intelligent control execution module and the visual prediction angle is greater than 15° for more than 5 seconds, it is determined to be an angle sensor offset fault; the centralized management platform periodically calculates the independent judgment accuracy of the visual module and the pressure module based on historical data and dynamically adjusts the weight coefficients w_1, w_2, and w_3.

9. A fire door monitoring and linkage control system based on vision and pressure perception according to claim 8, characterized in that, The platform layer is equipped with a visualization display platform, which supports the visualization display of fire doors, status query, abnormal alarm, historical records, and data report generation.

10. A fire door monitoring and linkage control system based on vision and pressure perception according to claim 1, characterized in that, The application layer provides web and mobile interfaces, allowing administrators to monitor in real time, receive alarms, and remotely and manually send fire door control commands to the platform layer.

Images