A chemical industry park personnel safety monitoring method and system based on dual-channel heterogeneous data transmission and multi-source fusion

Abstract

This application relates to the field of Internet of Things (IoT) technology, and in particular to a method and system for personnel safety monitoring in chemical industrial parks based on dual-channel heterogeneous data transmission and multi-source fusion. The method synchronously collects environmental parameters and inertial data through mobile sensing terminals; constructs an extended Kalman filter model to tightly couple and fuse ultra-wideband positioning coordinates with inertial dead reckoning data to correct cumulative drift in predicted values; employs a vertical distance threshold trajectory compression algorithm at the edge to eliminate redundant positioning points on linear paths, reducing transmission load; dynamically adjusts the radius of the electronic fence based on gas concentration and determines fall events by combining acceleration amplitude temporal characteristics; and utilizes a dual-channel WiFi collaborative mechanism to normally upload lightweight data via a low-power channel, only activating the high-speed channel to upload on-site images when boundary crossings or fall anomalies are detected. This application effectively solves the technical problems of low positioning accuracy, multimedia data transmission bandwidth conflicts, and delayed safety warnings in industrial sites.

Description

Technical Field

[0001] This invention belongs to the field of industrial Internet of Things and edge computing technology, specifically relating to a mobile target security monitoring method and system based on edge computing and multi-source heterogeneous fusion. Background Technology

[0002] With the rapid development of the chemical industry, chemical industrial parks, as high-risk industrial clusters, have complex production environments involving the production and storage of flammable, explosive, toxic, and hazardous gases. In the daily operation and maintenance of chemical industrial parks, inspection personnel work on the front lines, facing extremely high personal safety risks. In the event of emergencies such as gas leaks, falls, or accidental entry into dangerous areas, failure to detect and accurately rescue in a timely manner can easily lead to serious safety accidents, causing significant casualties and property damage. Therefore, constructing a highly reliable and accurate personnel safety monitoring system for chemical industrial parks is of great significance for improving the level of intelligent management in the parks and ensuring life safety.

[0003] However, existing industrial safety monitoring technologies still face many challenges in practical applications. Firstly, regarding positioning technology, chemical industrial parks are filled with metal pipelines and have complex electromagnetic environments. Traditional GPS positioning suffers from weak signals indoors or in obstructed environments. While single ultra-wideband (UWB) positioning technology offers relatively high accuracy, it is susceptible to interference and data drift in non-line-of-sight (NLOS) scenarios, resulting in uneven positioning trajectories and decreased accuracy, failing to meet centimeter-level safety monitoring requirements. Secondly, in terms of data transmission, existing monitoring terminals typically use a single channel to transmit all data. When uploading high-definition images or video streams for verification is required, it often consumes significant network bandwidth, causing congestion or delays in the transmission of critical location data and environmental sensor data, resulting in "bandwidth conflicts."

[0004] Furthermore, traditional electronic fence technology typically uses a fixed warning radius, which cannot dynamically adjust the danger zone based on the real-time concentration and diffusion of toxic gas leaks. This results in delayed safety warnings or high false alarm rates. Simultaneously, if mobile terminals directly upload all collected data without processing, it not only increases the server's processing load but also exacerbates the power consumption of terminal devices, shortening the battery life of inspection equipment. Therefore, there is an urgent need to develop a method and system for personnel safety monitoring in chemical industrial parks based on dual-channel heterogeneous data transmission and multi-source fusion. This system should improve positioning accuracy through multi-sensor fusion algorithms, reduce transmission load using edge computing strategies, and resolve concurrency conflicts between multimedia data and sensor data using a dual-channel mechanism, thereby achieving comprehensive, real-time intelligent protection for personnel in chemical industrial parks. Summary of the Invention

[0005] The present invention discloses a method for personnel safety monitoring in chemical industrial parks based on dual-channel heterogeneous data transmission and multi-source fusion, the core process of which includes the following seven key steps: Step 1: Construct mobile sensing terminals and synchronously collect data from multiple sources This invention first constructs a mobile sensing terminal based on STM32F103ZET6 as the edge computing core; during system initialization, the communication interfaces of each peripheral are configured, and a unified time base is established to solve the timestamp alignment problem of multi-sensor data; the data acquisition process performs the following operations in parallel: 1. Environmental Sensing: Drives the AM2301 temperature and humidity sensor and gas and smoke sensor via I2C or a single bus interface to acquire real-time information on the temperature, humidity, and concentration of harmful gases at the work site; 2. Motion Attitude Sensing: Data from the MPU9250 nine-axis inertial measurement unit, including three-axis acceleration, is read via the I2C interface. Triaxial angular velocity And magnetometer data. The sampling rate of this data needs to be set to meet the high-frequency requirements of human motion capture; 3. Position Awareness: The system reads raw ranging data or calculated raw coordinates between the UWB positioning tag and the base station via the UART serial port. Since the update frequency of UWB data is much lower than that of inertial sensors, the system needs to establish a data buffer queue in memory to perform frequency matching of heterogeneous data. Step 2: Perform multi-sensor fusion localization based on extended Kalman filter (EKF). To address the complementary characteristics of UWB's susceptibility to interference-induced jumps and the cumulative drift of PDR over time, this invention constructs an extended Kalman filter (EKF) model within the STM32 microcontroller for tight coupling and fusion. 1. Establish the system state space Define the system state vector It contains the target's position and velocity information on a two-dimensional plane:

[0006] in, for Optimal position estimation at time 10:00 for The velocity component at time step; 2. Forecasting Phase (Time Update) High-frequency acceleration data acquired using MPU9250 As the control input to the system, the state at the next moment is estimated a priori based on the kinematic equations; since human motion is nonlinear, the state transition matrix is ​​used... and control matrix Calculations are performed:

[0007] Meanwhile, the covariance matrix of the prediction system To quantify the uncertainty of current forecasts; Representative process noise covariance (mainly stemming from IMU measurement noise and integration error):

[0008] This stage utilizes the short-term, high-precision characteristics of inertial data to ensure the smoothness and continuity of the positioning trajectory; 3. Update Phase (Measurement Calibration) When the UWB module outputs new position coordinates When this happens, the observed value is used to correct the prediction result; First, calculate the Kalman gain. This gain determines whether the system trusts the predicted values ​​(IMU) or the observed values ​​(UWB) more. To observe the noise covariance, the value can be dynamically increased when the UWB signal quality is poor (such as in NLOS scenarios). To reduce the impact of UWB on the system; conversely, to reduce it:

[0009] By using Kalman gain correction to the state vector, the cumulative drift in inertial estimation can be eliminated:

[0010] in This is the observation matrix. The final output is the fused high-precision coordinates. It achieves stable positioning at the centimeter level in complex environments.

[0011] Step 3: Perform edge trajectory data compression To reduce the transmission load on the wireless network and save server storage space, this invention does not upload all positioning points. Instead, it uses a compression algorithm based on a vertical distance threshold at the edge of the STM32 to "clean" the trajectory.

[0012] Algorithm logic: This algorithm is based on a "sliding window" mechanism. Let the starting point of the trajectory fitting line be... The latest collection point is The intermediate trajectory point to be determined within the window is .

[0013] The system calculates intermediate points in real time. to the straight line vertical Euclidean distance The calculation formula is as follows:

[0014] Judgment strategy: Set an empirical distance threshold .

[0015] like : indicate a point The point located within the fluctuation range of the straight path (redundant point) contributes very little to the restoration of the trajectory shape, so it is discarded and only the local cache is updated.

[0016] like : indicate a point A deviation from the current straight path indicates a significant turn or path change for the personnel (a critical inflection point). At this point, the system retains the point, adds it to the compressed trajectory data packet, and triggers the data packaging and uploading operation.

[0017] This step significantly reduces the amount of transmitted data and improves channel utilization while ensuring that the geometric features of the trajectory are not lost.

[0018] Step 4: Perform dynamic risk assessment and anomaly detection This step simultaneously performs environmental risk assessment and personnel behavior monitoring at the edge, enabling proactive defense.

[0019] 1. Dynamic electronic fence based on nonlinear diffusion formula Traditional electronic fences have a fixed radius, while this invention uses real-time readings from gas sensors. Dynamically adjust the warning radius .

[0020] Set basic safety radius Safe concentration threshold and critical alarm concentration The computational logic employs piecewise functions: Safe state: if Gas concentration is normal, and the warning radius remains at the baseline value. .

[0021] Diffused state: if This indicates a gas leak. At this point, the natural logarithm model is used to simulate the gas diffusion risk, dynamically expanding the fence radius to allow the system to detect the risk of exceeding the boundary earlier.

[0022] Where is the gas diffusion influence coefficient; the higher the concentration, the greater the effect. The larger the size, the wider the warning range.

[0023] Dangerous state: If Set the warning radius directly to the maximum blockade radius. This triggers the highest level of alert.

[0024] The system calculates the fused coordinates in real time. With the center of danger The Euclidean distance, once satisfied This is considered an out-of-bounds anomaly.

[0025] 2. Fall detection based on temporal features To distinguish between falling and normal running or squatting, this invention uses MPU9250 data to calculate the triaxial composite acceleration amplitude. And perform strict "three-stage temporal feature matching": Phase 1 (Weightlessness): Weightlessness occurs at the moment of a fall, and it is detected... And the duration is Within the range.

[0026] Phase Two (Impact): The torso hits the ground, generating an impact, which is detected immediately after weightlessness. .

[0027] Phase 3 (Stationary): A brief period of immobility usually follows a fall. The time window following the detection of Phase 2 is [not specified]. Internal, angular velocity rate of change .

[0028] Only when the above three stages occur consecutively in strict chronological order is it determined to be a fall anomaly, effectively filtering out false alarms.

[0029] Step 5: Perform dual-channel heterogeneous data transmission This invention designs a unique physical dual-channel architecture to solve the bandwidth conflict problem.

[0030] Channel 1 (Low-Power Normal Channel): Utilizes the ESP8266 WiFi module. This module maintains a constant connection and is responsible for uploading the environmental parameters (temperature, humidity, gas) collected in Step 1 and the fused location coordinates calculated in Step 2 at a fixed frequency (e.g., 1Hz). Since these data are all lightweight text-based data, they consume minimal bandwidth, ensuring real-time monitoring and low power consumption.

[0031] Channel 2 (High-speed event-driven channel): Uses the ESP32-CAM image acquisition and communication module. In normal monitoring mode, the STM32 controls the ESP32-CAM's EN (enable) pin to be low, putting it in a power-off or deep sleep state, consuming no power and not occupying network channels.

[0032] Step Six: Perform Event-Driven Image Acquisition When the logical judgment in step four results in an electronic fence boundary violation or a fall, the system enters alarm mode: 1. Trigger Wake-up: The STM32 main control unit, acting as the decision center, immediately sets the GPIO connected to the ESP32-CAM EN pin to a high level, thus waking up the second channel in hardware.

[0033] 2. Image capture and upload: After the ESP32-CAM starts, it immediately controls the OV2640 camera to capture images of the scene and uploads them to the server asynchronously via HTTP protocol (different from the MQTT or TCP long connection of channel one).

[0034] 3. Channel isolation: Alarm data (uploaded via Channel 1 in seconds) and streaming media data (uploaded via Channel 2) are completely isolated in terms of physical hardware and network protocols, ensuring that even if network fluctuations are caused by image transmission, the command center can still receive the latest personnel location information through Channel 1, achieving highly reliable transmission of alarm data.

[0035] Step 7: Backend Data Analysis and Visualization The server receives data packets from both channels. The environmental / location data from channel one is aligned and associated with the image data from channel two based on a unified timestamp.

[0036] After the data is stored in the MySQL database, the web-based visualization platform performs the following operations: Trajectory reconstruction: Decompress the compressed trajectory points and connect them smoothly to draw the historical path of the personnel.

[0037] 1. Dynamic Fence Rendering: Dynamically draws a circle on the GIS map with the hazard source as the center, based on real-time gas concentration. The warning circle visually displays the spread of the danger zone.

[0038] 2. Pop-up window: Once an abnormal flag is received, an alarm window will pop up immediately, and the captured image at the scene will be displayed over the screen to allow managers to make quick decisions.

[0039] Compared with existing monitoring technologies for chemical industrial parks, this invention has the following significant advantages: 1. Solved the problem of positioning inaccuracy in complex industrial environments. Existing technologies mostly rely on single UWB positioning, which can result in errors of several meters under the metal obstruction of chemical industrial parks. This invention constructs an Extended Kalman Filter (EKF) model, utilizing the short-term high-precision characteristics of inertial sensors to fill the blind spots caused by UWB signal fluctuations or loss, and using the absolute position characteristics of UWB to correct inertial drift. Experiments demonstrate that this tightly coupled fusion mechanism can effectively suppress non-line-of-sight (NLOS) errors, outputting smooth, continuous, and high-precision coordinates.

[0040] 2. Resolved the bandwidth and power consumption conflict between multimedia transmission and real-time monitoring. Existing systems that continuously transmit camera data lead to network congestion and rapid power consumption. This invention innovatively proposes a dual-channel heterogeneous transmission mechanism. It utilizes the ESP8266 to transmit low-bandwidth telemetry data, while the ESP32-CAM transmits high-bandwidth image data only when an anomaly is triggered. This "normal-time hibernation, wartime wake-up" strategy ensures the acquisition of on-site visual evidence in critical moments, avoids multimedia data preempting channel access for critical location data, and significantly reduces the average power consumption of the terminal, extending the inspection endurance.

[0041] 3. Improved the efficiency of data transmission and storage. To address the issue of massive amounts of location data generated by mobile terminals, this invention implements a trajectory compression algorithm based on vertical distance at the edge. By eliminating redundant points on linear paths and retaining only key inflection points, the amount of location data transmitted can be reduced by approximately 70%-80% without affecting the accuracy of trajectory playback, significantly alleviating the load on wireless networks and the storage pressure on server databases.

[0042] 4. It achieves dynamic and accurate intelligent safety early warning. Unlike traditional static electronic fences and simple threshold alarms, this invention introduces a dynamic risk assessment model. Based on a nonlinear diffusion formula, the system can expand the warning radius in real time according to the concentration of toxic gases, allowing personnel downwind or near the leak point to receive alerts earlier. Simultaneously, by incorporating a fall detection algorithm based on the three-stage characteristics of "weightlessness-impact-stationary," false alarms caused by normal operational actions are effectively filtered out, significantly improving the system's safety assurance capabilities and intelligence level. Attached Figure Description

[0043] Figure 1 This is a flowchart of the personnel safety monitoring system for chemical industrial parks based on dual-channel heterogeneous data transmission and multi-source fusion proposed in this invention. Figure 2 This is a schematic diagram of the model structure of the dynamic electronic fence proposed in this invention. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0045] This invention provides a personnel safety monitoring system and method for chemical industrial parks based on dual-channel heterogeneous data transmission and multi-source fusion. The system includes a mobile sensing terminal, an edge computing unit, a dual-channel communication module, and a cloud monitoring platform. The mobile sensing terminal includes a main control module and a sensor acquisition group; the edge computing unit is embedded in the main control module; and the dual-channel communication module connects the main control module and the cloud monitoring platform.

[0046] In one optional implementation of this embodiment, such as Figure 1 As shown, Figure 1 The flowchart of the dual-channel heterogeneous data transmission and multi-source fusion system in an embodiment of the present invention is shown. The system uses an STM32 main control board as its core, connecting a temperature and humidity sensor, a hazardous gas sensor, a 9-axis attitude sensor, a UWB positioning tag, and an OV2640 camera. Through edge computing steps such as data time alignment and preprocessing, EKF multi-sensor fusion positioning, dynamic electronic fence calculation, and fall feature extraction, the data is finally uploaded to a cloud server for storage and web visualization via a dual-channel heterogeneous transmission layer (ESP8266 and ESP32-CAM).

[0047] In an optional implementation of this embodiment, the main control module integrates an STM32 microcontroller. Specifically, the main control module uses an STM32F103ZET6 microcontroller, which has rich peripheral interfaces (I2C, SPI, UART, GPIO) for driving multi-source sensors and executing complex edge computing algorithms. During system initialization, the main control module configures each communication interface and establishes a unified time base to solve the timestamp alignment problem of multi-sensor data.

[0048] In an optional implementation of this embodiment, the sensor acquisition group includes an environmental perception sensor, an inertial measurement unit, and a positioning tag, which are connected to the main control module via a hardware bus.

[0049] Specifically, the environmental sensing sensors include an AM2301 temperature and humidity sensor and a gas and smoke sensor, which are connected to the main control module via an I2C or single-bus interface to acquire the temperature, humidity and concentration of harmful gases at the work site in real time.

[0050] Specifically, the inertial measurement unit uses an MPU9250 nine-axis sensor, which is connected to the main control module through an I2C / SPI interface. It is used to collect three-axis acceleration, three-axis angular velocity and magnetometer data at a high frequency of over 100Hz to capture human motion posture.

[0051] Specifically, the positioning tag uses a UWB positioning module, which is connected to the main control module via a UART serial port to read the raw TOF ranging data or calculate the coordinates between the tag and the base station.

[0052] In one optional implementation of this embodiment, the edge computing unit runs inside the main control module and is used to perform multi-sensor fusion positioning based on extended Kalman filtering (EKF).

[0053] Specifically, to address the issues of UWB non-line-of-sight interference and cumulative drift in inertial navigation caused by metal obstruction in the chemical industrial park, the main control module constructs an EKF model. The model uses the acceleration collected by the MPU9250 as the control input for state prediction and dead reckoning, and uses the coordinates obtained from UWB as the observation values ​​to correct the reckoning results. By calculating the Kalman gain, the system can dynamically adjust the weights of the predicted and observed values, effectively suppressing non-line-of-sight (NLOS) errors and outputting smooth, continuous, and high-precision coordinates.

[0054] Furthermore, the edge computing unit also performs a trajectory data compression algorithm. Specifically, it sets a vertical distance threshold. The vertical distance from the current location point to the fitted line of the historical trajectory is calculated using a sliding window. .like If it is identified as a redundant point, it is discarded; This is identified as a critical inflection point and retained. This method can significantly reduce the load on subsequent data transmission.

[0055] In an optional implementation of this embodiment, the edge computing unit is further used to perform dynamic risk assessment and anomaly determination.

[0056] Specifically, such as Figure 2 As shown, Figure 2 A schematic diagram of a dynamic electronic fence model structure according to an embodiment of the present invention is shown. The dynamic risk assessment changes the traditional fixed-radius fence model, based on real-time monitored gas concentration. Dynamically adjust the warning radius Set the basic safety radius. Safe concentration threshold and critical alarm concentration .when At that time, the warning radius remains ;when At that time, using the nonlinear diffusion formula Dynamically expand the fence area, among which The gas diffusion influence coefficient; when At that time, the radius is set to the maximum blocking radius. The main control module calculates the distance between the fused coordinates and the center of the hazard source in real time. Once the distance falls within the dynamic radius range, it is determined to be an out-of-bounds anomaly.

[0057] At the same time, the main control module calculates the acceleration amplitude. If the three-stage time sequence of "weightlessness-impact-rest" is detected, it is determined to be an abnormal fall.

[0058] In an optional implementation of this embodiment, the dual-channel communication module includes a first WiFi module and a second WiFi module, which are respectively connected to the main control module.

[0059] Specifically, the first WiFi module uses an ESP8266 module and connects to the UART interface of the main control module. This module serves as a "normalized channel," remaining online at all times and responsible for uploading lightweight data such as compressed location coordinates and environmental parameters at a fixed frequency.

[0060] Specifically, the second WiFi module uses an ESP32-CAM image acquisition and communication module, which is connected to the GPIO pins and UART interface of the main control module. This module acts as an "event-driven channel," and its enable pin (EN) is controlled by the main control module.

[0061] In normal monitoring mode, the main control module sets the EN pin to a low level, putting the ESP32-CAM into a power-off or deep sleep state, thus consuming no power or bandwidth.

[0062] In abnormal trigger mode (i.e., when boundary crossing or a fall is detected), the main control module immediately sets the EN pin to a high level to wake up the ESP32-CAM, controls the camera to capture images of the scene, and asynchronously uploads them to the server via the HTTP protocol. This design achieves physical channel isolation between alarm image data and location data, avoiding bandwidth conflicts.

[0063] In an optional implementation of this embodiment, the system further includes a cloud monitoring platform, which includes a server data parsing module and a web visualization terminal.

[0064] Specifically, the server receives data packets uploaded from both channels, aligns and associates the environmental location data of channel one with the alarm image of channel two based on a unified timestamp, and stores it in a MySQL database.

[0065] Specifically, web-based visualization terminals are used for data display. For example... Figure 2The dynamic fence logic shown is rendered on a web-based map, drawing a warning zone (yellow / orange dashed line area) centered on the hazard source and varying with gas concentration. When an anomaly flag is received, an alarm window automatically pops up on the web interface, overlaying a captured image to assist management personnel in making quick decisions.

[0066] Working principle: After the system is powered on, the STM32 main control module controls the sensor group to synchronously collect environmental, attitude, and UWB raw data; internally, the main control module fuses UWB and IMU data using the EKF algorithm to obtain high-precision coordinates and performs trajectory compression; simultaneously, the main control module dynamically calculates the radius of the electronic fence based on the gas concentration (e.g., ...). Figure 2 ), and combine acceleration characteristics to determine whether an abnormality such as crossing the boundary or falling has occurred; During the data transmission phase, the ESP8266 serves as the first channel for routinely uploading location and environmental data; once an anomaly is detected, the STM32 wakes up the ESP32-CAM as the second channel to capture and upload images of the scene. After receiving the data, the cloud platform draws personnel trajectories and renders the dynamic electronic fence range in real time on the web interface, and pops up alarms when anomalies occur (such as...). Figure 1 (See the flowchart).

[0067] In summary, this invention proposes a personnel safety monitoring system for chemical industrial parks based on dual-channel heterogeneous data transmission and multi-source fusion. Using STM32 as the core, it integrates UWB and inertial data to improve positioning accuracy, dynamically adjusts the electronic fence range using gas concentration to achieve proactive early warning, and solves the network congestion and power consumption problems caused by multimedia data transmission through a physically isolated dual WiFi channel mechanism, significantly improving the reliability and intelligence level of safety monitoring in chemical industrial parks.

[0068] The foregoing has provided a detailed description of a method and system for personnel safety monitoring in chemical industrial parks based on dual-channel heterogeneous data transmission and multi-source fusion, as provided in the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method and system for personnel safety monitoring in chemical industrial parks based on dual-channel heterogeneous data transmission and multi-source fusion, characterized in that, The method includes the following steps: Step 1: Construct a mobile sensing terminal based on an STM32 microcontroller and initialize the system peripherals: The STM32 reads data from the AM2301 temperature and humidity sensor, gas and smoke sensor, and MPU9250 nine-axis inertial sensor via the I2C interface, and reads the raw ranging data between the UWB positioning tag and the base station via the serial port. Step 2: Perform multi-sensor fusion localization: An extended Kalman filter (EKF) model is built within the STM32 microcontroller. Acceleration and angular velocity calculated by the MPU9250 are used as state prediction values ​​for dead reckoning. Coordinates calculated by UWB are used as observation values. The Kalman gain is calculated, and the optimal estimated coordinates are iteratively output. The accumulated error is corrected, and the fused high-precision target position coordinates are output. Step 3: Perform edge trajectory data compression; set a distance threshold ε, and use a sliding window algorithm to calculate the vertical distance d⊥ from the current positioning point to the historical trajectory fitting line; if If the current point is determined to be a redundant point and discarded; If the current point is determined to be a key point of the trajectory and is retained, a compressed trajectory data packet is generated; Step 4: Constructing a Dynamic Electronic Fence and Anomaly Detection: Set the basic electronic fence range and monitor gas and smoke concentration data in real time. The warning radius of the electronic fence is dynamically calculated using a nonlinear diffusion formula. Simultaneously calculate the triaxial composite acceleration amplitude of the MPU9250. When the positioning coordinates fall into Scope or When the impact threshold is exceeded, the abnormal state flag is triggered; Step 5: Perform dual-channel heterogeneous data transmission: Configure the first WiFi module (ESP8266) and the second WiFi module (ESP32-CAM) to work together; upload the environmental data from Step 1 and the fused location coordinates from Step 2 to the server via the first WiFi module at a fixed frequency; Step Six: Perform event-driven image acquisition: When Step Three determines that an electronic fence boundary violation or fall has occurred, the STM32 main control unit immediately wakes up the second WiFi module, which is in sleep mode, controls the camera to capture images of the scene, and asynchronously uploads them to the server via HTTP protocol, thus achieving physical channel isolation between alarm data and streaming media data. Step 7: Backend Data Parsing and Visualization: The server receives dual-channel data, aligns and associates the image data with the environmental location data based on a unified timestamp, stores it in a MySQL database, and overlays and displays the dynamic electronic fence range and alarm images on the web interface.

2. The mobile target security monitoring method based on multi-source heterogeneous data fusion according to claim 1, characterized in that, In step two, the specific state update of the Extended Kalman Filter (EKF) model is performed according to the following formula: Establish the system state vector. in For position coordinates, For velocity components; Prediction stage: Acceleration data acquired using MPU9250 a k As a control input, it is transmitted through the state transition matrix. and control matrix Update the time: Update phase: Coordinates calculated using UWB Z k Calculate the Kalman gain as an observation. And correct the state estimate: in, Let covariance matrix be the variance matrix. For process noise covariance, To observe the noise covariance, Observation matrix; output optimal estimated coordinates through iterative calculation. .

3. The mobile target security monitoring method based on multi-source heterogeneous data fusion according to claim 1, characterized in that, Calculate vertical distance The specific algorithm is as follows: Let the coordinates of the starting point of the trajectory fitting line be... The latest data collection point coordinates are: The coordinates of the intermediate trajectory point to be determined are: ; calculate using the point-to-line distance formula when At that time, the judgment Within the linear path redundancy range, only update the local cache and do not upload; when At that time, the judgment For inflection points on the trajectory, add them to the compressed trajectory data packet and trigger the upload operation.

4. The mobile target security monitoring method based on multi-source heterogeneous data fusion according to claim 1, characterized in that, In step four, the logic for dynamically calculating the electronic fence warning radius using the nonlinear diffusion formula is as follows: Set the basic safety radius. R 0 Safe concentration threshold and critical alarm concentration ; Obtain real-time gas sensor concentration values Calculated using the following piecewise function : like ,but ;like ,but ; ,but ; in Let ln be the gas diffusion effect coefficient, and ln be the natural logarithm. The maximum blocking radius; the fused positioning coordinates output in step two. Coordinates of the center of the hazard source Perform Euclidean distance calculations, and once the positioning coordinates are fused... satisfy This is considered an out-of-bounds anomaly.

5. The mobile target security monitoring method based on multi-source heterogeneous data fusion according to claim 1, characterized in that, In step four, the specific steps for calculating the triaxial composite acceleration amplitude and determining the fall anomaly include: calculating the composite acceleration amplitude. Perform a three-stage temporal feature matching and determination: Stage 1 (Weight Loss): Detection And duration Phase Two (Impact): Immediately following Phase One, an impact was detected. Phase Three (Stillness): Within the preset time window following Phase Two. Internal, MPU9250 angular velocity The rate of change satisfies The fall exception flag is triggered only when the above three stages occur sequentially in order, and is sent to the STM32 master control unit as the highest priority interrupt signal; among which... The impact acceleration threshold, The threshold for weightlessness acceleration. This is the threshold value for the rate of change of stationary angular velocity.

6. A mobile target security monitoring system based on multi-source heterogeneous data fusion, applying the method as described in any one of claims 1 to 5, characterized in that, include: The sensing layer includes an STM32F103ZET6 main control board, an AM2301 temperature and humidity sensor, a gas and smoke sensor, an MPU9250 inertial measurement unit, and a UWB positioning tag connected via a serial port, all connected via an I2C / SPI bus. Transport layer: Includes the ESP8266 communication module connected to STM32 via UART interface, and the ESP32-CAM image acquisition module connected to STM32 via GPIO and UART interfaces; Power management unit: STM32 controls the EN enable pin of ESP32-CAM through GPIO pin; in normal monitoring mode, setting EN to low level puts ESP32-CAM into a power-off or deep sleep state; in abnormal trigger mode, setting EN to high level wakes up ESP32-CAM to perform image acquisition and transmission.

Images