A method and system for monitoring the posture of firefighters based on flexible bending sensors

Abstract

This invention discloses a method and system for monitoring firefighter posture based on a flexible bending sensor. The invention relates to the technical field of firefighter posture monitoring methods. It determines a bending change map of the flexible bending sensor based on changes in bending data. Abnormal posture nodes are identified based on this bending change map, the action path of the firefighter's actions, and a schematic diagram of the firefighter's current posture, improving the accuracy of abnormal posture node monitoring. Corresponding abnormal action paths are determined based on the bending data and action position of the abnormal posture nodes. Abnormal areas are identified based on multiple abnormal action paths, the corresponding action environment, and the firefighter's current position. Risky actions of the firefighter are determined based on the abnormal areas and corresponding firefighter actions. Posture correction measures are then determined based on these risky actions, the corresponding posture data, and the bending data of the flexible bending sensor, ensuring the accuracy of the firefighter's posture correction measures.

Description

Technical Field

[0001] This invention relates to the technical field of firefighter posture monitoring methods, and more particularly to a firefighter posture monitoring method and system based on a flexible bending sensor. Background Art

[0002] With the development of technology, firefighters are equipped with tight-fitting suits during training. These suits contain flexible bending sensors, which are arranged around various parts of the firefighter's body and flexibly change with the firefighter's movements to collect corresponding bending data. In existing technology, the firefighter's posture is determined based on the identification of bending data. However, it is impossible to know the abnormal posture nodes of the firefighter, which affects the accuracy of the firefighter's posture correction measures and makes it impossible to achieve dynamic adjustment of the firefighter's risky movements. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a method and system for monitoring the posture of firefighters based on a flexible bending sensor.

[0004] This invention provides a method for monitoring firefighter posture based on a flexible bending sensor, comprising: determining a schematic diagram of the firefighter's current posture based on the firefighter's current posture data and body shape data, and marking the bending data of a flexible bending sensor configured on the firefighter's clothing; determining a bending change diagram of the flexible bending sensor based on the changes in the bending data of the flexible bending sensor when the firefighter performs firefighting actions; determining abnormal posture nodes based on the bending change diagram, the action path of the firefighting action, and the schematic diagram of the firefighter's current posture; during the monitoring of the abnormal posture nodes, determining the corresponding abnormal action path based on the bending data and action position of the abnormal posture nodes, and determining abnormal areas based on multiple abnormal action paths, the corresponding action environment, and the firefighter's current position; determining the firefighter's risky actions based on the abnormal areas and the corresponding firefighting actions, and determining the firefighter's posture correction measures based on the risky actions, the corresponding posture data, and the bending data of the flexible bending sensor, and triggering the firefighter to dynamically adjust the risky actions.

[0005] This invention provides a firefighter posture monitoring system based on a flexible bending sensor. The system is applied to the aforementioned firefighter posture monitoring method based on a flexible bending sensor. The flexible bending sensor-based firefighter posture monitoring system includes:

[0006] The posture diagram module is used to determine the current posture diagram of the firefighter based on the firefighter's current posture data and body shape data, and to mark the bending data of the flexible bending sensor configured on the firefighter's clothing;

[0007] The bending change graph module is used to determine the bending change graph of the flexible bending sensor based on the changes in bending data of the flexible bending sensor when firefighters perform firefighting actions.

[0008] The abnormal posture node module is used to determine abnormal posture nodes based on the curvature change diagram, the action path of the firefighting action, and the current posture diagram of the firefighter.

[0009] The abnormal region module is used to determine the corresponding abnormal action path based on the bending data and action position of the abnormal posture node during the monitoring process of the abnormal posture node, and to determine the abnormal region based on multiple abnormal action paths, the corresponding action environment and the firefighter's current position.

[0010] The posture correction module is used to determine the firefighter's risky actions based on the abnormal area and the corresponding firefighting actions, and to determine the firefighter's posture correction measures based on the risky actions, the corresponding posture data and the bending data of the flexible bending sensor, and to trigger the firefighter to dynamically adjust the risky actions.

[0011] Compared with the prior art, the present invention has the following beneficial effects:

[0012] In this embodiment of the invention, the method described herein determines the bending change map of a flexible bending sensor based on the changes in bending data of the flexible bending sensor when a firefighter performs firefighting actions. Abnormal posture nodes are determined based on the bending change map, the action path of the firefighting action, and the current posture diagram of the firefighter. The introduction of the bending change map of the flexible bending sensor, combined with the overall consideration of the bending change map, the action path of the firefighting action, and the current posture diagram of the firefighter, improves the monitoring accuracy of abnormal posture nodes.

[0013] Therefore, during the monitoring of this abnormal posture node, the corresponding abnormal action path is determined based on the bending data and action position of the abnormal posture node. The abnormal area is determined based on multiple abnormal action paths, the corresponding action environment, and the firefighter's current position. The firefighter's risky action is determined based on the abnormal area and the corresponding firefighting action. The firefighter's posture correction measures are determined based on the risky action, the corresponding posture data, and the bending data of the flexible bending sensor. This triggers the firefighter's dynamic adjustment of the risky action, achieving further control over the risky action. In turn, the overall consideration of the risky action, the corresponding posture data, and the bending data of the flexible bending sensor is realized, ensuring the accuracy of the firefighter's posture correction measures and enabling the firefighter's dynamic adjustment of the risky action. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the firefighter posture monitoring method based on a flexible bending sensor in an embodiment of the present invention.

[0015] Figure 2 This is a schematic diagram of the structural composition of a firefighter posture monitoring system based on a flexible bending sensor according to an embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0017] Please see Figure 1 and Figure 2 A method for monitoring firefighter posture based on a flexible bending sensor is proposed and applied to firefighter posture monitoring scenarios. The method includes:

[0018] Step S11: Determine the current posture diagram of the firefighter based on the firefighter's current posture data and body shape data, and mark the bending data of the flexible bending sensor configured on the firefighter's clothing;

[0019] Step S12: When firefighters perform firefighting actions, determine the bending change map of the flexible bending sensor based on the changes in bending data of the flexible bending sensor;

[0020] Step S13: Determine the abnormal posture nodes based on the bending change diagram, the action path of the firefighting action, and the current posture diagram of the firefighter;

[0021] Step S14: During the monitoring of the abnormal posture node, the corresponding abnormal action path is determined based on the bending data and action position of the abnormal posture node, and the abnormal area is determined based on multiple abnormal action paths, the corresponding action environment and the current position of the firefighter.

[0022] Step S15: Based on the abnormal area and the corresponding firefighting action, determine the firefighter's risky action, and determine the firefighter's posture correction measures according to the risky action, the corresponding posture data and the bending data of the flexible bending sensor, and trigger the firefighter to dynamically adjust the risky action.

[0023] In step S11, a current posture diagram of the firefighter is determined based on the firefighter's current posture data and body shape data, and the bending data of the flexible bending sensor configured on the firefighter's clothing is marked.

[0024] In the specific implementation of this invention, the specific steps are as follows:

[0025] S111: Collect the firefighter's current position, determine the corresponding target detection area based on the firefighter's current position and the surrounding cameras, determine the firefighter's current posture data based on the real-time detection of the target detection area, and construct a schematic diagram of the firefighter's current posture based on the current posture data, the corresponding time node and the firefighter's body contour image.

[0026] S112: Determine the key posture parts of the firefighter based on the detection of the firefighter's current posture diagram, and determine the flexible bending sensor configured on the firefighter's clothing based on the position detection of the key posture parts.

[0027] S113: Real-time monitoring of the flexible bending sensor; determining the bending data recorded by the flexible bending sensor as the firefighter's posture changes based on the real-time monitoring of the flexible bending sensor, which shows the degree of bending of the flexible bending sensor and the degree of posture change of the firefighter.

[0028] In the embodiments of this application, the current location of the firefighter is collected, and a positioning technology suitable for complex environments (such as indoor fire scenes) is selected. Common positioning technologies include GPS, UWB (Ultra-Wideband), Bluetooth beacons, etc. In indoor environments, GPS signals may be weak, so UWB or Bluetooth beacons may be better choices. Positioning modules are installed on the firefighter's equipment (such as fire helmets, fire suits, or belts), and positioning base stations (such as UWB base stations or Bluetooth beacons) are deployed at the fire scene to receive and process positioning signals. The positioning module collects the firefighter's location data in real time and transmits the data to the monitoring system via wireless communication (such as Bluetooth or Wi-Fi). The monitoring system receives the location data and stores it in a database.

[0029] Multiple cameras are deployed at the fire scene to ensure coverage of key areas (such as entrances, corridors, staircases, etc.). The cameras can be fixed or mobile (such as cameras mounted on drones). Based on the firefighter's current location, the nearest camera is selected as the target detection device to determine the scope of the target detection area and ensure that the firefighter's full-body posture can be captured completely. The cameras collect the firefighter's image data in real time and transmit the data to the monitoring system via wired or wireless network. The monitoring system processes the received image data and extracts the firefighter's posture information.

[0030] The image data captured by the camera is preprocessed, including noise reduction and contrast enhancement, to improve image quality. Background subtraction or frame differencing is used to remove background interference and highlight the firefighter's outline. Deep learning algorithms (such as OpenPose and YOLO) are used to perform pose detection on the processed images. The algorithm can identify the position and angle of key parts of the firefighter's body (such as head, shoulders, elbows, knees, etc.). The current pose data of the firefighter is extracted from the detection results, including the position coordinates and angle information of each key part. The pose data is stored in the database of the monitoring system for subsequent analysis.

[0031] While collecting posture data, the corresponding time points are recorded to ensure the synchronization of posture data with time. The time points can be used for subsequent posture change analysis and historical data backtracking. The body contour image of the firefighter is extracted from the images captured by the camera. The body contour is extracted using edge detection algorithms (such as Canny edge detection) or deep learning algorithms (such as semantic segmentation). Combining posture data, time points, and body contour images, a current posture diagram of the firefighter is constructed. The posture diagram can be a three-dimensional model or a two-dimensional image, which intuitively displays the firefighter's body posture and the relative positions of various parts.

[0032] Optionally, suppose a fire breaks out in a large warehouse, and firefighters enter the building to conduct firefighting and search and rescue operations. Each firefighter's helmet is equipped with a UWB positioning module, and multiple UWB base stations are deployed around the site. The UWB positioning module collects the firefighter's location data every 0.1 seconds and transmits the data to a monitoring system via Bluetooth. Upon receiving the location data, the monitoring system updates the firefighter's location information in real time. After the firefighter enters the building, the monitoring system selects the nearest fixed camera as the target detection device based on the firefighter's UWB positioning data. This camera covers the corridor area where the firefighter is located. The camera collects image data every 0.03 seconds and transmits the data to the monitoring system via a wired network. The monitoring system processes the received image data in real time to extract the firefighter's posture information.

[0033] After receiving image data from the camera, the monitoring system first performs noise reduction and contrast enhancement on the image. Then, it uses the OpenPose algorithm to perform pose detection on the processed image. The OpenPose algorithm identifies the position coordinates and angle information of key parts of the firefighter's body, such as the head, shoulders, elbows, and knees. For example, the detection results show that the firefighter's left shoulder is at (x1, y1), the right shoulder is at (x2, y2), the left elbow angle is 30°, and the right elbow angle is 45°. This pose data is stored in the monitoring system's database. The monitoring system records the corresponding time node (e.g., June 28, 2025, 14:30:00) while recording the firefighter's pose data. From the images captured by the camera, the Canny edge detection algorithm is used to extract the firefighter's body contour image. Then, combining the pose data (e.g., left shoulder position (x1, y1), right shoulder position (x2, y2), left elbow angle 30°, right elbow angle 45°) and body contour image, a two-dimensional pose diagram is constructed. The posture diagram shows the firefighter currently in a crouching, crawling position, with the left arm bent at a 30° angle and the right arm bent at a 45° angle. The monitoring system displays this posture diagram in real time on the screen in the command center, allowing commanders to quickly understand the firefighter's status.

[0034] Furthermore, the key posture parts of the firefighter are determined based on the detection of the firefighter's current posture diagram, and the flexible bending sensor installed on the firefighter's clothing is determined based on the position detection of the key posture parts. This takes into account the overall consideration of the position detection of the key posture parts and ensures the accuracy of the flexible bending sensor installed on the firefighter's clothing.

[0035] At this point, key posture areas refer to the parts of the body that change most significantly during a firefighter's movements and are most helpful in judging posture. These typically include the shoulders, elbows, wrists, waist, hips, knees, and ankles. The bending angles and positional changes of these areas can reflect the firefighter's main movement states, such as crawling, squatting, bending over, and lifting objects. By analyzing the firefighter's current posture diagram, the position and angle of key posture areas can be identified.

[0036] The installation location of the flexible bending sensor should be determined based on the position of the key posture areas. The sensor should be installed in a location that can accurately capture bending changes in the key areas. For example, for the elbows and knees, the sensor can be installed on the sleeves and trouser legs of the fire suit; for the waist, the sensor can be installed at the waist position of the fire suit. A suitable fixing device should be designed to ensure that the sensor does not loosen or shift during the firefighter's movements. Elastic bandages, Velcro, or built-in pockets can be used to fix the sensor to the fire suit.

[0037] After firefighters don their gear, initial sensor calibration is performed. The sensors are calibrated using known bending angles (e.g., 0° when the arm is straight and 90° when bent at 90°) to ensure the accuracy of their measurements. Simple calibration maneuvers (e.g., arm straight, bent at 90°, knee straight, bent at 90°, etc.) can be used. The sensor response is tested while firefighters perform basic movements (e.g., raising their hands, bending over, squatting, etc.) to ensure accurate capture of bending changes. If inaccurate sensor responses are found, the sensor's position can be adjusted or it can be recalibrated.

[0038] Optionally, assuming the firefighter is climbing a ladder, the monitoring system constructs a schematic diagram of the firefighter's current posture using cameras and posture detection algorithms. The posture diagram shows the firefighter's left shoulder position as (x1, y1), right shoulder position as (x2, y2), left elbow angle as 30°, right elbow angle as 45°, waist flexion angle as 15°, left knee angle as 90°, and right knee angle as 100°. By analyzing this data, the key posture areas can be identified as the left and right elbows, waist, and both knees, because changes in the flexion angles of these areas are most crucial for judging the ladder-climbing action.

[0039] During the aforementioned ladder-climbing mission, the monitoring system identified the key posture areas as the left elbow, right elbow, waist, and both knees. Based on these locations, the firefighters' equipment was fitted with the following flexible bending sensors: Left elbow sensor: installed at the elbow of the left sleeve of the fire suit. Right elbow sensor: installed at the elbow of the right sleeve of the fire suit. Waist sensor: installed at the waist of the fire suit, close to the spine. Left knee sensor: installed at the knee of the left trouser leg of the fire suit. Right knee sensor: installed at the knee of the right trouser leg of the fire suit. These sensors were secured to the fire suit with elastic bandages to ensure stable capture of bending changes during the firefighter's ladder climb.

[0040] After firefighters donned their gear and installed the sensors, initial calibration was performed. With the firefighter's arm extended, the monitoring system recorded initial values ​​of 0° for the left and right elbow sensors; then, the firefighter bent their arm 90°, and the system recorded 90° for both elbow sensors. This process calibrated all sensors. Next, the firefighter performed some basic movement tests. For example, when the firefighter bent over, the waist sensor recorded a bending angle of 30°; when the firefighter squatted, the knee sensors recorded bending angles of 90° and 100° respectively. The monitoring system confirmed that the sensor responses were accurate and that the sensor placement perfectly matched the key posture areas.

[0041] Therefore, the flexible bending sensor is monitored in real time; based on the real-time monitoring of the flexible bending sensor, the bending data recorded by the flexible bending sensor as the firefighter's posture changes is determined. This bending data shows the degree of bending of the flexible bending sensor and the degree of posture change of the firefighter, which is compatible with the overall consideration of real-time monitoring of the flexible bending sensor and ensures the accuracy of the bending data recorded by the flexible bending sensor as the firefighter's posture changes.

[0042] At this point, after the firefighters put on their equipment and completed sensor calibration, they activate the real-time monitoring function of the sensors to ensure that the sensors are connected to the monitoring system (such as the server or mobile terminal in the command center) via wireless communication (such as Bluetooth, Wi-Fi) or wired communication. The sensors need to have low power consumption, high sampling rate and low latency to ensure the real-time performance and accuracy of the data. The sensors collect bending data in real time at a set sampling rate (such as 50Hz or 100Hz). The bending data includes information such as the bending angle, bending speed and bending direction of the sensors. The collected data is transmitted to the monitoring system in real time through the wireless communication module.

[0043] The monitoring system processes the received bending data in real time, including filtering, noise reduction, and data correction. It extracts key features of the bending data, such as the range of bending angle changes and the peak bending speed. The processed data is combined with a diagram of the firefighter's current posture to analyze the changes in the firefighter's posture. The processed bending data is stored in the monitoring system's database, recording the data's timestamp, sensor location, and bending status. The data storage format can be tabular for easy subsequent analysis and backtracking.

[0044] The processed bending data is displayed in real time on the monitoring system interface to present the firefighter's posture changes in an intuitive way. Charts (such as line graphs and bar charts) or animations (such as dynamic displays of 3D models) can be used to display the data. Based on the range and speed of the bending data changes, the degree of the firefighter's posture change can be judged. For example, a rapid change in bending angle may indicate that the firefighter is performing a violent action, such as climbing a ladder quickly or avoiding danger, while a slow change in bending angle may indicate that the firefighter is performing a more stable action, such as walking slowly or carrying equipment.

[0045] Optionally, suppose a firefighter is on a firefighting mission, equipped with multiple flexible bending sensors located at the left elbow, right elbow, waist, and both knees. The sensors connect to the command center's monitoring system via Bluetooth. The sensors have a sampling rate of 50Hz, meaning they collect data 50 times per second. As the firefighter moves and adjusts their posture during firefighting, the sensors collect bending data in real time and transmit the data to the monitoring system via Bluetooth. In this firefighting mission, after receiving the bending data from the sensors, the monitoring system first performs filtering to remove potential noise interference. For example, the bending angle data of the left elbow sensor is as follows: Timestamp: 2025-06-28 14:30:00, bending angle: 30°; Timestamp: 2025-06-28 14:30:01, bending angle: 45°; Timestamp: 2025-06-28 14:30:02, bending angle: 60°; Timestamp: 2025-06-28 14:30:03, bending angle: 75°; Timestamp: 2025-06-28 14:30:04, bending angle: 90°. The monitoring system analyzes this data and finds that the bending angle of the left elbow increased from 30° to 90° in a short period, indicating that the firefighter may be performing an action requiring a significant bending of the left arm, such as lifting a fire extinguisher. The processed data is stored in the database, and a bending angle matching table is collected, as shown in Table 1.

[0046] Table 1 Bending Angle Matching Table

[0047]

[0048] The monitoring system interface displays a real-time curve showing the change in the flexion angle of the firefighter's left elbow. The curve shows that between 14:30:00 and 14:30:04 on June 28, 2025, the flexion angle of the left elbow rapidly increased from 30° to 90°, at a rate of 15° / s. Analyzing this data, the monitoring system determines that the firefighter may be performing an action requiring a rapid lifting of their left arm, such as raising a fire extinguisher to put out a fire. Simultaneously, the system also displays flexion data from other sensors. For example, the waist sensor recorded a flexion angle increasing from 15° to 30°, indicating the firefighter may be bending over; the knee sensors recorded flexion angles of 90° and 100° respectively, indicating the firefighter may be in a squatting position. By comprehensively analyzing this data, the monitoring system can understand the firefighter's overall posture changes in real time.

[0049] In step S12, when the firefighter performs firefighting actions, the bending change map of the flexible bending sensor is determined based on the change in bending data of the flexible bending sensor;

[0050] In the specific implementation of this invention, the specific steps are as follows:

[0051] S121: Collect the firefighters' fire training subjects, and determine the firefighters' fire training content based on the analysis of the firefighters' fire training subjects. Match the corresponding firefighting actions based on the firefighters' fire training content, and then the firefighters perform the firefighting actions.

[0052] S122: Real-time monitoring of the firefighter's execution of the firefighting action and marking the corresponding action time period. Based on the database of the flexible bending sensor and the corresponding action time period, multiple bending data of the flexible bending sensor are determined. The multiple bending data change in time sequence along the action time period.

[0053] S123: As the bending data of the flexible bending sensor changes, corresponding bending data markers are gradually constructed based on each bending data and the corresponding time node. The bending change map of the flexible bending sensor is dynamically constructed according to the presentation of multiple bending data markers in the same coordinate system.

[0054] In the embodiments of this application, training subjects can be obtained through a fire training management system (such as an electronic training record system) or manually entered by the instructor or firefighter. The training subjects are usually described in text form, such as "indoor fire extinguishing training", "high-rise building rescue training", "physical training", etc. The training management system parses the specific training content based on the name and description of the training subject. For example, "indoor fire extinguishing training" may include the following: using fire extinguishers to extinguish indoor fires; checking fire sources and assessing the fire situation; quickly evacuating the fire scene; rescuing trapped personnel.

[0055] Establish a motion library containing various firefighting actions, each with a detailed description and corresponding sensor data characteristics. The motion library can include the following: raising hands to operate a fire extinguisher; bending over to check for fire sources; running quickly; climbing ladders; carrying equipment; crouching down to rescue trapped personnel. Based on the training content, match the specific actions that firefighters need to perform from the motion library. For example, for "using a fire extinguisher to extinguish an indoor fire," the matched actions might be "raising hands to operate a fire extinguisher" and "bending over to check for fire sources."

[0056] The system guides firefighters to perform matched firefighting actions through voice prompts, screen displays, or AR devices. For example, the system can provide a voice prompt such as "Please raise your hand to operate the fire extinguisher" and display the correct posture for the action. During the firefighter's execution of the action, the system monitors the execution of the action in real time through cameras and flexible bending sensors to ensure that the firefighter's actions meet the training requirements.

[0057] Optionally, assume a firefighter is participating in "indoor fire extinguishing training." The training management system records his training subject as "indoor fire extinguishing training" and stores it in the database. The training management system analyzes the "indoor fire extinguishing training" subject and determines its training content as: using a fire extinguisher to extinguish an indoor fire; checking the fire source and assessing the fire; quickly evacuating the fire scene; rescuing trapped personnel. Based on the training content "using a fire extinguisher to extinguish an indoor fire," the system matches the following fire-fighting actions from the action library: Action 1: Raise your hand to operate the fire extinguisher (left elbow bent at 90°, right elbow bent at 90°); Action 2: Bend over to check the fire source (waist bent at 30°). The system guides the firefighter to perform the "raise your hand to operate the fire extinguisher" action through voice prompts. Following the prompts, the firefighter bends his left and right elbows to 90° respectively, simulating the action of operating the fire extinguisher. Simultaneously, the system monitors the firefighter's action execution in real time through a camera and a flexible bending sensor to ensure that his actions meet the requirements.

[0058] Furthermore, the execution process of the firefighter's firefighting action is monitored in real time, and the corresponding action time period is marked. Based on the database of the flexible bending sensor and the corresponding action time period, multiple bending data of the flexible bending sensor are determined. The multiple bending data change in chronological order along the action time period, which is compatible with the overall consideration of the database of the flexible bending sensor and the corresponding action time period, and ensures the accuracy of the multiple bending data of the flexible bending sensor.

[0059] At this point, cameras or other visual monitoring equipment are used to capture the firefighters' movements in real time. The cameras can be fixed in suitable locations on the training ground to ensure complete capture of the firefighters' full-body movements. Flexible bending sensors collect bending data of various parts of the firefighters' bodies in real time. The sensors transmit data to the monitoring system via wireless communication (such as Bluetooth). The start and end times of the firefighters' movements are determined using the monitoring equipment (such as cameras) or sensor data. The start and end times of movements can be automatically detected by setting thresholds (such as changes in bending angle). The start and end times of the movements are recorded in the monitoring system to form a time period. For example, the start time of movement 1 is 14:30:00, and the end time is 14:30:05.

[0060] Bending data for the corresponding action time period is extracted from the database of flexible bending sensors. The database stores the bending angle of each sensor at different time points. This time series reflects the bending changes of various parts of the firefighter's body during the action. The extracted bending data is arranged in chronological order to form a time series. Through time series analysis, the bending changes of various parts of the firefighter's body during the action can be observed. The time series data can be visualized, for example, by plotting a line graph.

[0061] Specifically, suppose a firefighter is performing the "raise hand to operate a fire extinguisher" maneuver. Multiple cameras are installed in the training area to monitor the firefighter's movements in real time. Simultaneously, the firefighter is equipped with multiple flexible bending sensors located at the left elbow, right elbow, waist, left knee, and right knee. The sensors collect bending data in real time at a sampling rate of 50Hz and transmit the data to the monitoring system via Bluetooth. When the firefighter performs the "raise hand to operate a fire extinguisher" maneuver, the monitoring system uses the camera and sensor data to determine the start time of the maneuver as 14:30:00 and the end time as 14:30:05. The monitoring system records this time period and marks it as the "raise hand to operate a fire extinguisher" maneuver period.

[0062] The monitoring system extracts bending data from the flexible bending sensor database for the period from 14:30:00 to 14:30:05. The extracted data is shown in Table 2.

[0063] Table 2 Data Table

[0064]

[0065]

[0066] As can be seen from the table above, when firefighters perform the "raise hand to operate fire extinguisher" action, the bending angle of the left and right elbows gradually increases from 0° to 90°, and remains at 90° after 14:30:00.200, while the bending angle of the waist and knees remains basically unchanged.

[0067] Therefore, as the bending data of the flexible bending sensor changes, corresponding bending data markers are gradually constructed based on each bending data and the corresponding time node. The bending change map of the flexible bending sensor is dynamically constructed based on the presentation of multiple bending data markers in the same coordinate system.

[0068] At this point, the bending data marker refers to the sensor bending angle and position information recorded at a specific time point. Each marker includes a timestamp, sensor position (e.g., left elbow, right elbow, waist, etc.), and the corresponding bending angle. As sensor data is acquired in real time, the system records the bending angle at each time point and generates the corresponding marker. For example, if the timestamp is 14:30:00.000 and the left elbow bending angle is 0°, then the marker (14:30:00.000, left elbow, 0°) is generated.

[0069] Define a coordinate system where the horizontal axis represents time and the vertical axis represents the bending angle. Data from each sensor is represented by different colors or lines on the coordinate system for easy differentiation. Plot the bending data at each time point on the coordinate system to create a line graph or curve. Dynamically update the graph over time to display the real-time changes in the bending angle.

[0070] Optionally, assuming firefighter Zhang is performing the action of "raising his hand to operate the fire extinguisher," the flexible bending sensor collects data at a sampling rate of 50Hz. The data-timestamp matching table is shown in Table 3:

[0071] Table 3 Data-Timestamp Matching Table

[0072]

[0073] The corresponding markings are as follows: (14:30:00.000, left elbow, 0°); (14:30:00.020, left elbow, 10°); (14:30:00.040, left elbow, 20°); (14:30:00.060, left elbow, 30°); (14:30:00.080, left elbow, 40°); (14:30:00.100, left elbow, 50°); (14:30:00.120, left elbow, 60°); (14:30:00.140, left elbow, 70°); (14:30:00.160, left elbow, 80°); (14:30:00.180, left elbow, 90°).

[0074] In step S13, abnormal posture nodes are determined based on the bending change diagram, the action path of the firefighting action, and the current posture diagram of the firefighter.

[0075] In the specific implementation of this invention, the specific steps are as follows:

[0076] S131: Collect the bending change map, identify multiple abnormal bending data based on the identification of the bending change map, and determine the abnormal bending area based on the value and location of the multiple abnormal bending data and the corresponding body position of the firefighter.

[0077] S132: Collect a schematic diagram of the firefighter's current posture, determine the corresponding matching area based on the matching of the firefighter's current posture schematic diagram and the abnormal bending area, and mark multiple first sub-abnormal posture nodes based on the autonomous recognition of the matching area;

[0078] S133: Perform online monitoring of the fire-fighting action, determine the action path of the fire-fighting action based on the online monitoring of the fire-fighting action, and determine multiple second sub-abnormal posture nodes based on the action path of the fire-fighting action and the current posture diagram of the firefighter. Determine each abnormal posture node based on the matching of multiple first sub-abnormal posture nodes and multiple second sub-abnormal posture nodes.

[0079] In embodiments of this application, a bending change graph generated during the firefighter's actions is acquired from a monitoring system. The bending change graph records the bending angle changes of each sensor at different time points. The bending change graph is typically stored in time series format, with the bending angle of each sensor recorded at each time point. The data format can be tabular for ease of subsequent analysis.

[0080] Abnormal bending data refers to bending angle data that deviates from the normal movement path. Abnormal data can be identified by setting thresholds (such as excessively rapid changes in bending angle or exceeding the normal range). Machine learning algorithms (such as anomaly detection algorithms) or simple threshold judgment methods can be used to identify abnormal data. For example, if the lumbar bending angle increases from 0° to 30° in a short period, it can be considered abnormal data. An abnormal bending region refers to an area where the bending angle of a part of the firefighter's body changes abnormally during the execution of a movement. This region can be determined by the value and location of the abnormal data. The abnormal bending region is determined based on the timestamp of the abnormal data and the sensor location. For example, if a lumbar sensor records abnormal data, the abnormal bending region is the lumbar region.

[0081] Optionally, assuming firefighter Zhang is performing the action of "raising his hand to operate the fire extinguisher," the monitoring system generates the following bending change diagram, as shown in Table 4:

[0082] Table 4 Bending Variation Diagram

[0083]

[0084]

[0085] Analysis of the bending change graph revealed that the lumbar bending angle suddenly increased to 30° between 14:30:00.200 and 14:30:00.300, which is inconsistent with the normal motion path and was therefore identified as abnormal bending data. The lumbar sensor recorded the abnormal data (lumbar bending angle increasing from 0° to 30°), thus the abnormal bending area was identified as the lumbar region. Specifically, the locations were the left and right sides of the lumbar region, as the bending angle changed significantly at these locations. This clearly demonstrates how the bending change graph was acquired, abnormal bending data was identified, and the abnormal bending region was determined. This process provides an important foundation for subsequent abnormal posture node identification.

[0086] Furthermore, a schematic diagram of the firefighter's current posture is collected. Based on the matching of the firefighter's current posture schematic diagram and the abnormal bending area, the corresponding matching area is determined. Based on the autonomous recognition of the matching area, multiple first sub-abnormal posture nodes are marked. This approach takes into account the overall consideration of matching the firefighter's current posture schematic diagram and the abnormal bending area, ensuring the accuracy of the corresponding matching area.

[0087] At this point, a current posture diagram of the firefighter is retrieved from the monitoring system. This posture diagram can be a two-dimensional image or a three-dimensional model, displaying the position and posture of various parts of the firefighter's body. It can be generated using image data captured by cameras, combined with deep learning algorithms (such as OpenPose). A posture diagram typically includes the following information: the coordinates of key body parts (such as the head, shoulders, elbows, waist, knees, and ankles); the bending angles of each body part; and a timestamp for synchronization with the bending change map. Abnormal bending areas (such as the waist) are matched with the current posture diagram to determine the corresponding matching region. The matching region refers to the body part in the current posture diagram that corresponds to the abnormal bending area.

[0088] Deep learning algorithms (such as convolutional neural networks, CNNs) are used to autonomously identify matching regions and label abnormal pose nodes. Abnormal pose nodes refer to the specific locations within the matching region where the pose is abnormal. The location and pose information of each abnormal pose node are labeled. For example, the bending angle and position coordinates of the left and right sides of the waist are labeled.

[0089] Optionally, assuming firefighter Zhang is performing the action of "raising his hand to operate the fire extinguisher," the monitoring system captures his posture image through the camera and generates a posture diagram, as shown in Table 5:

[0090] Table 5 Attitude Diagram

[0091]

[0092] The posture diagram shows that Xiao Zhang's left and right elbows are bent at 90°, his waist is bent at 30°, and his left and right knees are bent at 0°. In the above example, the abnormal bending area is the waist (the waist bending angle increases from 0° to 30°). By matching the current posture diagram, the matching area is determined to be the waist, specifically the left and right sides of the waist. Within the matching area (waist), the following first sub-abnormal posture nodes are marked by the autonomous recognition algorithm: Node 1: Left side of the waist, position (180, 300), bending angle 30°; Node 2: Right side of the waist, position (220, 300), bending angle 30°.

[0093] Therefore, the firefighting action is monitored online, and the action path of the firefighting action is determined based on the online monitoring. Multiple second sub-abnormal posture nodes are determined according to the action path of the firefighting action and the current posture diagram of the firefighter. Each abnormal posture node is determined based on the matching of multiple first sub-abnormal posture nodes and multiple second sub-abnormal posture nodes. The bending change diagram of the flexible bending sensor is introduced. The overall consideration of the bending change diagram, the action path of the firefighting action and the current posture diagram of the firefighter is taken into account, which improves the monitoring accuracy of abnormal posture nodes.

[0094] At this point, cameras or other visual monitoring equipment are used to monitor the firefighters' firefighting actions in real time. The cameras can be fixed in suitable locations on the training ground to ensure complete capture of the firefighters' full-body movements. Image processing and posture recognition algorithms (such as OpenPose or YOLO) are used to extract the movement trajectories of various parts of the firefighter's body in real time, forming a motion path. The motion path includes the position coordinates of each key part at different points in time. The motion path is analyzed to identify parts that do not conform to the normal motion path. For example, if the waist position changes abnormally in the motion path, it may indicate an abnormal waist posture. The abnormal parts in the motion path are matched with the current posture diagram to determine the corresponding second sub-abnormal posture node. The second sub-abnormal posture node refers to the specific location of the posture abnormality in the motion path. The first and second sub-abnormal posture nodes are matched to determine the final abnormal posture node. The matching method can be based on position and posture information; for example, if the position and posture information of two nodes are similar, they are considered to be the same abnormal posture node.

[0095] Optionally, suppose firefighter Zhang is performing the action of "raising his hand to operate a fire extinguisher." Multiple cameras are installed in the training area to monitor Zhang's actions in real time. The monitoring system uses the image data collected by the cameras, combined with the OpenPose algorithm, to generate a motion path matching table, as shown in Table 6.

[0096] Table 6 Motion Path Matching Table

[0097]

[0098] The motion path shows an abnormal change in waist position between 14:30:00.200 and 14:30:00.300 (waist bending angle increased from 0° to 30°). By matching the current posture diagram, the second sub-abnormal posture node is determined as follows: Node 3: left side of waist, position (180, 300), bending angle 30°; Node 4: right side of waist, position (220, 300), bending angle 30°.

[0099] In the above example, the first and second sub-abnormal posture nodes are as follows: First sub-abnormal posture node: Node 1: Left side of waist, position (180, 300), bending angle 30°; Node 2: Right side of waist, position (220, 300), bending angle 30°; Second sub-abnormal posture node: Node 3: Left side of waist, position (180, 300), bending angle 30°; Node 4: Right side of waist, position (220, 300), bending angle 30°; Through matching, the final abnormal posture nodes are determined as follows: Abnormal posture node 1: Left side of waist, position (180, 300), bending angle 30°; Abnormal posture node 2: Right side of waist, position (220, 300), bending angle 30°.

[0100] In step S14, during the monitoring of the abnormal posture node, the corresponding abnormal action path is determined based on the bending data and action position of the abnormal posture node, and the abnormal area is determined based on multiple abnormal action paths, the corresponding action environment and the firefighter's current position.

[0101] In the specific implementation of this invention, the specific steps are as follows:

[0102] S141: Real-time monitoring of abnormal posture nodes, collection of bending data of abnormal posture nodes, marking of action positions of abnormal posture nodes, determination of corresponding multimodal data based on the fusion of bending data and action positions of abnormal posture nodes, and determination of corresponding abnormal action paths based on the identification of multimodal data.

[0103] S142: Detect the surrounding environment of the abnormal action path, and determine multiple environmental data along the detection of the surrounding environment of the abnormal action path. Determine the corresponding action environment based on the multiple environmental data and the abnormal action path, and determine the first abnormal part based on the multiple abnormal action paths and the corresponding action environment.

[0104] S143: Determine the second abnormal part based on multiple abnormal action paths and the firefighter's current location; determine the abnormal area based on the mapping relationship between the first abnormal part, the second abnormal part, and the abnormal area.

[0105] In the embodiments of this application, abnormal posture nodes are monitored in real time, and bending data of abnormal posture nodes are collected. The action positions of abnormal posture nodes are marked. Based on the fusion of bending data and action positions of abnormal posture nodes, corresponding multimodal data is determined. Based on the identification of the multimodal data, the corresponding abnormal action path is determined. This approach is compatible with the overall consideration of the fusion of bending data and action positions of abnormal posture nodes, ensuring the accuracy of the corresponding multimodal data.

[0106] At this point, monitoring devices such as flexible bending sensors and cameras are used to monitor the dynamic changes of abnormal posture nodes in real time. An abnormal posture node refers to the specific location where the posture is abnormal during the execution of an action. Bending angle data of the abnormal posture node is collected from the flexible bending sensor. Action position data of the abnormal posture node is collected from the camera. The flexible bending sensor collects bending angle data of the abnormal posture node in real time. The data includes a timestamp, sensor position, and the corresponding bending angle. The collected bending data is recorded in the monitoring system's database.

[0107] Image data acquired by cameras is used to mark the specific locations of abnormal posture nodes during the movement process. The movement location can be represented by the coordinates of key components. The marked movement location data is recorded in the monitoring system's database. Bending data and movement location data are fused to form multimodal data. Multimodal data can include timestamps, sensor locations, bending angles, and movement location coordinates.

[0108] Optionally, suppose that when firefighter Zhang is performing the action of "raising his hand to operate the fire extinguisher," the monitoring system detects abnormal posture nodes on the left and right sides of his waist. The monitoring system monitors these nodes in real time using flexible bending sensors and cameras. The bending data matching table collected by the monitoring system is shown in Table 7:

[0109] Table 7 Bending Data Matching Table

[0110]

[0111] The action location data matching table marked by the monitoring system is shown in Table 8:

[0112] Table 8. Action Position Data Matching Table

[0113]

[0114] The fused multimodal data are shown in Table 9:

[0115] Table 9. Action Position Data Matching Table

[0116]

[0117] By analyzing the multimodal data, the abnormal motion paths were determined as follows: Abnormal motion path 1: The left side of the waist (node ​​1) was fixed at (180, 300) with a bending angle of 30° from 14:30:00.200 to 14:30:00.240; Abnormal motion path 2: The right side of the waist (node ​​2) was fixed at (220, 300) with a bending angle of 30° from 14:30:00.200 to 14:30:00.240.

[0118] Furthermore, the surrounding environment of the abnormal action path is detected, and multiple environmental data are determined along the detection of the surrounding environment of the abnormal action path. The corresponding action environment is determined based on the multiple environmental data and the abnormal action path. The first abnormal part is determined based on the multiple abnormal action paths and the corresponding action environment. This approach takes into account the overall consideration of multiple environmental data and abnormal action paths, ensuring the accuracy of the corresponding action environment.

[0119] At this point, by analyzing the multimodal data, the abnormal action path was determined as follows:

[0120] Abnormal movement path 1: The left side of the waist (node ​​1) is fixed at (180, 300) with a bending angle of 30° between 14:30:00.200 and 14:30:00.240.

[0121] Abnormal movement path 2: The right side of the waist (node ​​2) is fixed at (220, 300) between 14:30:00.200 and 14:30:00.240, with a bending angle of 30°.

[0122] At this point, environmental sensors (such as temperature sensors, smoke sensors, and cameras) are used to detect the environment around the abnormal movement path. These sensors can be fixed in suitable locations within the training area or mounted on the firefighters' equipment. Multiple environmental data points are collected along the abnormal movement path. This data can include temperature, smoke concentration, and obstacle locations. The collected environmental data is then organized into a table containing timestamps, sensor locations, temperature, smoke concentration, and obstacle locations. Environmental data at each time point can be considered a single data point.

[0123] The action environment refers to the environmental characteristics surrounding an abnormal action path during the execution of an action. The action environment can include information such as temperature, smoke concentration, and obstacle location. By matching environmental data with abnormal action paths, the action environment corresponding to each abnormal action path is determined. For example, if the environmental data around an abnormal action path indicates high temperature, high smoke concentration, and the presence of obstacles, then the action environment for that path is "high temperature, high smoke concentration, and obstacles".

[0124] The first anomalous region refers to the anomalous area in the action environment associated with the anomalous action path. The first anomalous region can be a specific area or a set of features. It is determined by combining multiple anomalous action paths with their corresponding action environments. For example, if multiple anomalous action paths indicate that a certain area has a high temperature, high smoke concentration, and obstacles, then that area can be considered the first anomalous region.

[0125] Optionally, suppose that when firefighter Zhang is performing the action of "raising his hand to operate the fire extinguisher," the monitoring system detects abnormal movement paths on the left and right sides of his waist. The monitoring system uses environmental sensors to detect the environment surrounding these paths. The data matching table collected by the environmental sensors is shown in Table 10:

[0126] Table 10. Action Position Data Matching Table: Data Matching Acquired by Environmental Sensors

[0127]

[0128] The resulting environmental data matching table is shown in Table 11:

[0129] Table 11 Environmental Data Matching Table

[0130]

[0131] Environmental data is matched with abnormal action paths to determine the action environment corresponding to each abnormal action path. For example, if the environmental data around an abnormal action path indicates high temperature, high smoke concentration, and the presence of obstacles, then the action environment for that path is "high temperature, high smoke concentration, and obstacles". By analyzing the environmental data, the action environment is determined as follows:

[0132] Action Environment 1: Around the left side of the waist (node ​​1), the temperature is 50-54°C, the smoke concentration is 10-15%, and there is an obstacle at location (160, 310). Action Environment 2: Around the right side of the waist (node ​​2), the temperature is 50-54°C, the smoke concentration is 10-15%, and there is an obstacle at location (240, 310). The first anomaly refers to the anomaly in the action environment. Through analysis of environmental data, the action environment is determined as follows: Action Environment 1: Around the left side of the waist (node ​​1), the temperature is 50-54°C, the smoke concentration is 10-15%, and there is an obstacle at location (160, 310). Action Environment 2: Around the right side of the waist (node ​​2), the temperature is 50-54°C, the smoke concentration is 10-15%, and there is an obstacle at location (240, 310). This can be a feature set. By comprehensively analyzing the abnormal action path and action environment, the first abnormal part was identified as follows: The first abnormal part is the area around the left and right sides of the waist, with a high temperature (50-54°C), a high smoke concentration (10-15%), and obstacles at (160, 310) and (240, 310).

[0133] Therefore, the second abnormal part is determined based on multiple abnormal action paths and the firefighter's current location; the abnormal area is determined based on the mapping relationship between the first abnormal part, the second abnormal part, and the abnormal area, which takes into account the overall consideration of the mapping relationship between the first abnormal part, the second abnormal part, and the abnormal area, and ensures the accuracy of the abnormal area.

[0134] At this point, positioning technologies (such as GPS, UWB, Bluetooth beacons, etc.) are used to obtain the firefighter's current location in real time. Location information includes coordinates (x, y) and a timestamp. Multiple abnormal movement paths are analyzed to determine their relationship to the firefighter's current location. An abnormal movement path refers to an abnormal movement trajectory of a part of the firefighter's body during the execution of an action. A second abnormal region refers to an abnormal area near the firefighter's current location related to the abnormal movement paths. By analyzing the abnormal movement paths and the firefighter's current location, the specific location and extent of the second abnormal region are determined.

[0135] The first abnormal part refers to the abnormal area related to the abnormal action path in the action environment. The specific location and characteristics of the first abnormal part have been determined through step S142. A mapping relationship is established between the first and second abnormal parts to determine the final abnormal area. An abnormal area refers to the area jointly defined by the abnormal action path of a certain part of the firefighter's body and the characteristics of the surrounding environment during the execution of the action. The final abnormal area is determined by combining the first and second abnormal parts. The abnormal area can be a specific region or a set of features.

[0136] Optionally, assume the first anomalous area is: the area around the left and right sides of the waist, with high temperature (50-54°C), high smoke concentration (10-15%), and obstruction locations (160, 310) and (240, 310). The second anomalous area is: the area around the firefighter's current position (200, 300), particularly the area around the left and right sides of the waist. By combining the first and second anomalous areas, the final anomalous area is determined as follows: Assume the first anomalous area is: the area around the left and right sides of the waist, with high temperature (50-54°C), high smoke concentration (10-15%), and obstruction locations (160, 310) and (240, 310). The second anomalous area is: the area around the firefighter's current position (200, 300), particularly the area around the left and right sides of the waist. By combining the first and second abnormal parts, the final abnormal area is determined as follows: Abnormal area: the area around the left and right sides of the waist, and the area around the firefighter's current position (200, 300), with high temperature (50-54°C), high smoke concentration (10-15%), and obstacles at (160, 310) and (240, 310).

[0137] In step S15, the firefighter's risky actions are determined based on the abnormal area and the corresponding firefighting actions. The firefighter's posture correction measures are determined based on the risky actions, the corresponding posture data, and the bending data of the flexible bending sensor. The firefighter's dynamic adjustment of the risky actions is then triggered.

[0138] In the specific implementation of this invention, the specific steps are as follows:

[0139] S151: Collect abnormal areas, locate corresponding fire actions based on the abnormal areas, the action paths and action times of fire actions, determine multiple action nodes based on the detection of fire actions, determine the risk coefficient based on the location, action range and action environment of the action node, determine the risk node of the firefighter based on the comparison of the risk coefficient with the corresponding risk coefficient threshold, and output the risk action corresponding to the risk node.

[0140] S152: Determine the corresponding posture data based on the posture detection of the risk node, determine the first correction range based on the risk action and the corresponding posture data, and determine the second correction range based on the risk action and the bending data of the flexible bending sensor.

[0141] S153: Determine the firefighter's posture correction measures based on the mapping relationship between the first correction range, the second correction range, and the posture correction measures, and mark the correction action steps corresponding to the posture correction measures to trigger the firefighter's dynamic adjustment of risky actions.

[0142] In the embodiments of this application, abnormal areas are collected, and corresponding fire-fighting actions are located based on the abnormal areas, the action paths of the fire-fighting actions, and the action times. Multiple action nodes are determined based on the detection of the fire-fighting actions, and a risk coefficient is determined based on the location, action range, and action environment of the action node. The risk node of the firefighter is determined by comparing the risk coefficient with the corresponding risk coefficient threshold, and the risk action corresponding to the risk node is output. This approach takes into account the overall consideration of comparing the risk coefficient with the corresponding risk coefficient threshold, ensuring the accuracy of the firefighter's risk node.

[0143] At this point, the abnormal zone refers to the area defined by the abnormal movement path of a certain part of the firefighter's body and the characteristics of the surrounding environment during the execution of the action. Characteristics of the abnormal zone include location, temperature, smoke concentration, and obstacle location. Information about the abnormal zone, including location, temperature, and smoke concentration, is obtained from the monitoring system. This data is collected in real time using environmental sensors (such as temperature sensors, smoke sensors, and cameras).

[0144] Action path refers to the movement trajectory of various parts of a firefighter's body during the execution of an action. Action time refers to the time period during which the action is performed. Historical data and trained models are used to analyze the relationship between abnormal areas, action paths, and action times. Machine learning algorithms (such as decision trees and support vector machines) are used to locate the corresponding firefighting actions.

[0145] Action nodes refer to the position and posture information of key parts during the execution of an action. Action nodes can be detected through camera and sensor data. Using cameras and flexible bending sensors, the position and posture of key parts of the firefighter are detected in real time. The timestamp, position coordinates, and bending angle of each action node are recorded.

[0146] The risk coefficient is a quantitative indicator used to assess the risk level of an action node. It is calculated using weighted scores, considering factors such as location, action range, and action environment. Weights are assigned to each factor, adjusting their contribution to the total score according to their importance. The risk coefficient for each action node is calculated. A preset risk coefficient threshold is used to determine whether an action node is a risk node. If the risk coefficient exceeds the threshold, the node is considered a risk node. The risk coefficient of each action node is compared to the risk coefficient threshold to identify risk nodes.

[0147] Optionally, assume the monitoring system identifies the following abnormal areas: Abnormal areas: the area around the left and right sides of the waist, with high temperature (50-54°C), high smoke concentration (10-15%), and obstacles at locations (160, 310) and (240, 310); assume that through training data, the monitoring system locates the corresponding fire-fighting action as "raising hand to operate the fire extinguisher". The action path and time are as follows: Action path: the positions of the left and right sides of the waist are (180, 300) and (220, 300) respectively, with a bending angle of 30°. Action time: 14:30:00.200 to 14:30:00.240. The monitoring system detects the following action nodes: Action node 1: left side of the waist, position (180, 300), bending angle 30°; Action node 2: right side of the waist, position (220, 300), bending angle 30°.

[0148] Assume the weighting is as follows: Location weight: 0.4; Movement range weight: 0.3; Movement environment weight: 0.3; The risk coefficient of movement node 1 is calculated as follows: Location score: (180, 300) is close to the obstacle (160, 310), score 80; Movement range score: Bending angle 30° exceeds the normal range, score 70; Movement environment score: Temperature 52°C, smoke concentration 12%, score 75; Risk coefficient = (80 × 0.4) + (70 × 0.3) + (75 × 0.3) = 32 + 21 + 22.5 = 75.5

[0149] Similarly, the risk coefficient for action node 2 is 74.5. Assume the risk coefficient threshold is 70. The risk coefficient for action node 1 is 75.5, exceeding the threshold of 70, and is therefore identified as a risk node. The risk coefficient for action node 2 is 74.5, exceeding the threshold of 70, and is therefore identified as a risk node. The fire-fighting action corresponding to risk nodes 1 and 2 is "excessive bending of the waist," which may lead to waist injury. The monitoring system outputs the following risk action: Risk Action: Excessive bending of the waist, which may lead to waist injury.

[0150] Furthermore, the corresponding attitude data is determined based on the attitude detection of the risk node, the first correction range is determined based on the risk action and the corresponding attitude data, and the second correction range is determined based on the risk action and the bending data of the flexible bending sensor. This takes into account the overall consideration of the risk action and the bending data of the flexible bending sensor, ensuring the accuracy of the second correction range.

[0151] At this point, posture detection refers to real-time monitoring of the position and posture of various parts of the firefighter's body using camera and sensor data. Posture data includes the position coordinates and bending angles of key body parts. Cameras and flexible bending sensors are used to collect posture data at risk points in real time. The timestamp, position coordinates, and bending angle of each risk point are recorded. The first correction range refers to the range of key body parts that need correction based on the posture data. This range is determined based on the risky actions and posture data to ensure that the firefighter's actions comply with safety standards. The risky actions and posture data are analyzed to determine the range of key body parts that need correction.

[0152] For example, if the angle of the waist bends too much, it needs to be adjusted to a safe range.

[0153] Optionally, assuming the risky action is "excessive lumbar flexion," the corresponding posture data is as follows: Risk node 1: Left side of the waist, position (180, 300), flexion angle 30°; Risk node 2: Right side of the waist, position (220, 300), flexion angle 30°; Based on the risky action and posture data, the first correction range is determined as follows: First correction range: Position range of the left and right sides of the waist from (170, 290) to (230, 310). Assuming the risky action is "excessive lumbar flexion," the corresponding flexion data is as follows: Risk node 1: Left side of the waist, flexion angle 30°; Risk node 2: Right side of the waist, flexion angle 30°; Based on the risky action and flexion data, the second correction range is determined as follows: Second correction range: Flexion angle range of the left and right sides of the waist from 0° to 15°.

[0154] Flexible bending sensors are used to collect the bending angles of key areas in real time. For example, the bending angle on the left side of the waist is 30°, and the bending angle on the right side of the waist is 30°. Risk node 1: Left side of the waist, position (180, 300), bending angle 30°; Risk node 2: Right side of the waist, position (220, 300), bending angle 30°; First correction range: Position range of the left and right sides of the waist (170, 290) to (230, 310); Risk node 1: Left side of the waist, bending angle 30°; Risk node 2: Right side of the waist, bending angle 30°; Second correction range: Bending angle range of the left and right sides of the waist from 0° to 15°.

[0155] Therefore, based on the mapping relationship between the first correction range, the second correction range, and the posture correction measures, the posture correction measures for firefighters are determined, and the corresponding correction action steps are marked to trigger the dynamic adjustment of firefighters to risky actions. This approach incorporates the overall consideration of the mapping relationship between the first correction range, the second correction range, and the posture correction measures, ensuring the accuracy of the firefighters' posture correction measures. At the same time, it enables further control of the risky action, thereby achieving an overall consideration of the risky action, the corresponding posture data, and the bending data of the flexible bending sensor, ensuring the accuracy of the firefighters' posture correction measures and enabling the dynamic adjustment of firefighters to risky actions.

[0156] At this point, the attitude correction measure mapping relationship refers to mapping the first and second correction ranges to specific correction measures. This mapping relationship can be achieved through preset rules or machine learning models. Based on the first correction range (position range) and the second correction range (bending angle range), combined with the attitude correction measure mapping relationship, specific correction measures are determined. Correction measures should include specific motion adjustment suggestions, such as adjusting the bending angle or adjusting the position.

[0157] Corrective action steps refer to the specific steps that need to be followed when implementing corrective measures. Each corrective measure can be broken down into multiple specific action steps. Based on the corrective measure, the specific steps of each corrective action are marked. The corrective action steps should describe in detail how to adjust posture, including the starting position, target position, and adjustment method. Dynamic adjustment refers to adjusting the firefighter's actions based on real-time data during the execution of the action. Dynamic adjustment can be achieved by using a monitoring system to provide real-time reminders to firefighters to adjust their posture. The monitoring system monitors the firefighter's posture data in real time. When a risky action is detected, corrective measures are triggered, and a real-time reminder is sent to the firefighter.

[0158] Specifically, assume the first and second correction ranges are as follows: First correction range: the position range of the left and right sides of the waist (170, 290) to (230, 310); Second correction range: the bending angle range of the left and right sides of the waist (0° to 15°); The posture correction measures are mapped as follows: Correction measure 1: left side of the waist, adjust the bending angle to 0° to 15°; Correction measure 2: right side of the waist, adjust the bending angle to 0° to 15°; The determined posture correction measures are as follows: Correction measure 1: left side of the waist, adjust the bending angle to 0° to 15°; Correction measure 2: right side of the waist, adjust the bending angle to 0° to 15°.

[0159] Assume the corrective actions are as follows: Corrective action 1: Left side of the waist, adjust the bending angle to 0° to 15°; Corrective action 2: Right side of the waist, adjust the bending angle to 0° to 15°; The marked corrective action steps are as follows: Corrective action step 1: Left side of the waist, slowly adjust the bending angle to 0° to 15°; Starting position: 30°; Target position: 0° to 15°; Adjustment method: Slowly relax the waist, avoid sudden movements.

[0160] Correction Step 2: Slowly adjust the bending angle of the right side of the waist to 0° to 15°; Starting position: 30°; Target position: 0° to 15°; Adjustment method: Slowly relax the waist, avoiding sudden movements; Assume the monitoring system detects in real-time that firefighter Zhang's left and right waist bending angles are both 30°, exceeding the second correction range (0° to 15°). The monitoring system triggers corrective measures and sends a real-time reminder to the firefighter: Real-time reminder: The bending angles of the left and right sides of the waist are too large. Please slowly adjust to 0° to 15° to avoid waist injury. Through the analysis of the above detailed steps and the illustration of practical examples, it is clear how to determine the firefighter's posture correction measures and mark the correction action steps through the mapping relationship between the first correction range, the second correction range, and posture correction measures, in order to trigger the firefighter's dynamic adjustment of risky movements.

[0161] Please see Figure 2 , Figure 2 This is a schematic diagram of the structural composition of a firefighter posture monitoring system based on a flexible bending sensor according to an embodiment of the present invention; the firefighter posture monitoring system based on a flexible bending sensor includes:

[0162] The posture diagram module 21 is used to determine the current posture diagram of the firefighter based on the firefighter's current posture data and body shape data, and to mark the bending data of the flexible bending sensor configured on the firefighter's clothing.

[0163] The bending change graph module 22 is used to determine the bending change graph of the flexible bending sensor based on the changes in bending data of the flexible bending sensor when firefighters perform firefighting actions.

[0164] Abnormal posture node module 23 is used to determine abnormal posture nodes based on the bending change diagram, the action path of the firefighting action, and the current posture diagram of the firefighter;

[0165] The abnormal area module 24 is used to determine the corresponding abnormal action path based on the bending data and action position of the abnormal posture node during the monitoring process of the abnormal posture node, and to determine the abnormal area based on multiple abnormal action paths, the corresponding action environment and the current position of the firefighter.

[0166] The posture correction module 25 is used to determine the firefighter's risky actions based on the abnormal area and the corresponding firefighting actions, and to determine the firefighter's posture correction measures based on the risky actions, the corresponding posture data and the bending data of the flexible bending sensor, and to trigger the firefighter to dynamically adjust the risky actions.

[0167] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for monitoring the posture of firefighters based on a flexible bending sensor, characterized in that, include: A diagram of the firefighter's current posture is determined based on the firefighter's current posture data and body shape data, and the bending data of the flexible bending sensor configured on the firefighter's clothing is marked. When firefighters perform firefighting actions, the bending change map of the flexible bending sensor is determined based on the changes in bending data of the flexible bending sensor; Based on the bending change diagram, the action path of the firefighting action, and the diagram of the firefighter's current posture, abnormal posture nodes are determined; During the monitoring of this abnormal posture node, the corresponding abnormal action path is determined based on the bending data and action position of the abnormal posture node, and the abnormal area is determined based on multiple abnormal action paths, the corresponding action environment and the current position of the firefighter. Based on the abnormal area and the corresponding firefighting action, the firefighter's risky actions are determined, and the firefighter's posture correction measures are determined based on the risky action, the corresponding posture data and the bending data of the flexible bending sensor, and the firefighter's dynamic adjustment of the risky action is triggered.

2. The firefighter posture monitoring method based on a flexible bending sensor according to claim 1, characterized in that, The process of determining a firefighter's current posture diagram based on the firefighter's current posture data and body shape data, and marking the bending data of the flexible bending sensor configured on the firefighter's clothing, includes: The system collects the firefighter's current location, determines the corresponding target detection area based on the firefighter's current location and surrounding cameras, determines the firefighter's current posture data based on real-time detection of the target detection area, and constructs a schematic diagram of the firefighter's current posture based on the current posture data, the corresponding time point, and the firefighter's body contour image. Based on the detection of the firefighter's current posture diagram, the key posture parts of the firefighter are determined, and the flexible bending sensor installed on the firefighter's clothing is determined based on the position detection of the key posture parts. The flexible bending sensor is monitored in real time; based on the real-time monitoring of the flexible bending sensor, the bending data recorded by the flexible bending sensor as the firefighter's posture changes is determined, and the bending data shows the degree of bending of the flexible bending sensor and the degree of posture change of the firefighter.

3. The firefighter posture monitoring method based on a flexible bending sensor according to claim 1, characterized in that, The method of determining the bending change map of the flexible bending sensor based on the change in bending data of the flexible bending sensor when firefighters perform firefighting actions includes: Collect firefighters' fire training subjects, and determine the firefighters' fire training content based on the analysis of the firefighters' fire training subjects. Match the corresponding firefighting actions based on the firefighters' fire training content, and then the firefighters perform the firefighting actions. The system monitors the firefighters' execution of the firefighting action in real time and marks the corresponding action time period. Based on the database of the flexible bending sensor and the corresponding action time period, it determines multiple bending data of the flexible bending sensor. The multiple bending data change in chronological order along the action time period. As the bending data of the flexible bending sensor changes, corresponding bending data markers are gradually constructed based on each bending data point and the corresponding time node. The bending change map of the flexible bending sensor is dynamically constructed based on the presentation of multiple bending data markers in the same coordinate system.

4. The firefighter posture monitoring method based on a flexible bending sensor according to claim 1, characterized in that, The process of determining abnormal posture nodes based on the curvature change diagram, the action path of the firefighting action, and the current posture diagram of the firefighter includes: The bending change map is collected, and multiple abnormal bending data are identified based on the identification of the bending change map. The abnormal bending area is determined based on the value and location of the multiple abnormal bending data and the corresponding body position of the firefighter.

5. The firefighter posture monitoring method based on a flexible bending sensor according to claim 4, characterized in that, The step of determining abnormal posture nodes based on the curvature change diagram, the action path of the firefighting action, and the current posture diagram of the firefighter also includes: Collect a schematic diagram of the firefighter's current posture, determine the corresponding matching area based on the matching of the firefighter's current posture schematic diagram and the abnormal bending area, and mark multiple first sub-abnormal posture nodes based on the autonomous recognition of the matching area; The firefighting action is monitored online. Based on the online monitoring of the firefighting action, the action path of the firefighting action is determined. Based on the action path of the firefighting action and the current posture diagram of the firefighter, multiple second sub-abnormal posture nodes are determined. Each abnormal posture node is determined based on the matching of multiple first sub-abnormal posture nodes and multiple second sub-abnormal posture nodes.

6. The firefighter posture monitoring method based on a flexible bending sensor according to claim 1, characterized in that, During the monitoring of the abnormal posture node, the corresponding abnormal motion path is determined based on the bending data and motion position of the abnormal posture node. The abnormal area is determined based on multiple abnormal motion paths, the corresponding motion environment, and the firefighter's current position, including: The system monitors abnormal posture nodes in real time, collects their bending data, and marks their action positions. Based on the fusion of the bending data and action positions of the abnormal posture nodes, it determines the corresponding multimodal data and identifies the corresponding abnormal action path based on the recognition of the multimodal data.

7. The firefighter posture monitoring method based on a flexible bending sensor according to claim 6, characterized in that, During the monitoring of the abnormal posture node, the corresponding abnormal motion path is determined based on the bending data and motion position of the abnormal posture node. The abnormal area is determined based on multiple abnormal motion paths, the corresponding motion environment, and the firefighter's current position. This also includes: The surrounding environment of the abnormal action path is detected, and multiple environmental data are determined along the detection of the surrounding environment of the abnormal action path. The corresponding action environment is determined based on the multiple environmental data and the abnormal action path. The first abnormal part is determined based on the multiple abnormal action paths and the corresponding action environment. The second abnormal part is determined based on multiple abnormal action paths and the firefighter's current location; the abnormal area is determined based on the mapping relationship between the first abnormal part, the second abnormal part, and the abnormal area.

8. The firefighter posture monitoring method based on a flexible bending sensor according to claim 1, characterized in that, The process involves determining the firefighter's risky actions based on abnormal areas and corresponding firefighting actions, and determining posture correction measures for the firefighter based on the risky actions, corresponding posture data, and bending data from a flexible bending sensor, triggering dynamic adjustments to the firefighter's risky actions, including: The system collects data on abnormal areas, locates corresponding firefighting actions based on the abnormal areas, action paths, and action times of the firefighting actions, determines multiple action nodes based on the detection of the firefighting actions, determines the risk coefficient based on the location, action range, and action environment of the action node, determines the risk node of the firefighter based on the comparison of the risk coefficient with the corresponding risk coefficient threshold, and outputs the risk action corresponding to the risk node.

9. The firefighter posture monitoring method based on a flexible bending sensor according to claim 8, characterized in that, The process of determining firefighters' risky actions based on abnormal areas and corresponding firefighting actions, and determining firefighter posture correction measures based on the risky actions, corresponding posture data, and bending data from flexible bending sensors, and triggering dynamic adjustments to the firefighters' risky actions, also includes: The posture data corresponding to the risk node is determined by posture detection. A first correction range is determined based on the risk action and the corresponding posture data. A second correction range is determined based on the risk action and the bending data of the flexible bending sensor. Based on the mapping relationship between the first correction range, the second correction range, and the posture correction measures, the posture correction measures for firefighters are determined, and the corresponding correction action steps are marked to trigger the dynamic adjustment of firefighters to risky actions.

10. A firefighter posture monitoring system based on a flexible bending sensor, characterized in that, The firefighter posture monitoring system based on a flexible bending sensor is applied to the firefighter posture monitoring method based on a flexible bending sensor as described in any one of claims 1-9, wherein the firefighter posture monitoring system based on a flexible bending sensor includes: The posture diagram module is used to determine the current posture diagram of the firefighter based on the firefighter's current posture data and body shape data, and to mark the bending data of the flexible bending sensor configured on the firefighter's clothing; The bending change graph module is used to determine the bending change graph of the flexible bending sensor based on the changes in bending data of the flexible bending sensor when firefighters perform firefighting actions. The abnormal posture node module is used to determine abnormal posture nodes based on the curvature change diagram, the action path of the firefighting action, and the current posture diagram of the firefighter. The abnormal region module is used to determine the corresponding abnormal action path based on the bending data and action position of the abnormal posture node during the monitoring process of the abnormal posture node, and to determine the abnormal region based on multiple abnormal action paths, the corresponding action environment and the firefighter's current position. The posture correction module is used to determine the firefighter's risky actions based on the abnormal area and the corresponding firefighting actions, and to determine the firefighter's posture correction measures based on the risky actions, the corresponding posture data and the bending data of the flexible bending sensor, and to trigger the firefighter to dynamically adjust the risky actions.

Images