Fire rescue positioning methods, systems, equipment and storage media

Abstract

This application provides a fire rescue positioning method, system, equipment, and storage medium. The method includes: acquiring sensor data, inertial navigation data, and wireless positioning data from the fire scene; fusing the sensor data, inertial navigation data, and wireless positioning data into a time series, and then inputting this into a trained neural network model for processing to obtain target positioning information including the distribution location of various targets and environmental parameters in the fire environment. By fusing heterogeneous data sources such as inertial navigation data, wireless positioning data, and environmental data from the fire scene, this application can provide accurate and continuous positioning information and environmental monitoring, significantly enhancing positioning accuracy and thus providing reliable data support for fire rescue.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a fire rescue positioning method, system, device and storage medium. Background Technology

[0002] Urban high-rise fires are characterized by high loads, rapid fire development, and a constantly changing and complex on-site environment, making efficient evacuation and rescue difficult and resulting in significant casualties and property losses.

[0003] To address the challenges of location and route planning in fire rescue operations, existing technologies offer several solutions. Some rely on pre-defined CAD drawings of the building's internal structure for personnel location and route planning. However, in a fire, the building structure may change due to collapsed walls or blocked passageways, making location and route planning based on static drawings inaccurate or unsafe. Other solutions typically use GPS for indoor positioning, but the accuracy of GPS is significantly affected by the complexity of the indoor environment during a fire.

[0004] It can be seen that existing technologies are unable to reliably achieve high-precision indoor positioning in fire rescue scenarios, thus failing to provide effective data support for fire rescue and evacuation operations. Summary of the Invention

[0005] This application provides a fire rescue positioning method, system, equipment, and storage medium. By integrating multiple heterogeneous data sources at the fire scene, it can provide accurate and continuous positioning information and environmental monitoring, significantly enhancing the accuracy of positioning and thus providing reliable data support for fire rescue.

[0006] To achieve the above objectives, firstly, this application provides a fire rescue positioning method, the method comprising:

[0007] S1. Acquire sensor data, inertial navigation data, and wireless positioning data at the fire scene;

[0008] S2. The sensor data, inertial navigation data and wireless positioning data are fused into a time series, and the time series is input into a trained neural network model to obtain target positioning information. The target positioning information includes the distribution location and environmental parameters of each target in the fire environment.

[0009] In one possible implementation, step S1 includes:

[0010] S11. Obtain at least one sensor data from the temperature sensor and gas sensor at the fire scene;

[0011] S12. For any target in the fire scene, obtain the inertial navigation data of the target from the mobile node carried by the target;

[0012] S13. Obtain the signal strength recorded by multiple fixed access nodes at the fire scene, and determine the wireless positioning data of the target based on the multiple signal strengths.

[0013] In one possible implementation, the method further includes:

[0014] The mobile node and multiple fixed access nodes at the fire scene form a communication network;

[0015] The mobile node sends inertial navigation data to the fixed access node.

[0016] In one possible implementation, step S2 includes:

[0017] S21. The sensor data, inertial navigation data and wireless positioning data are concatenated into a time series format input sequence according to the time order. Each time step in the time series contains sensor data, inertial navigation data and wireless positioning data at the same time.

[0018] S22. Input the input sequence into the trained neural network model to extract the current position features and environmental parameter features;

[0019] S23. Predict the current location features and the environmental parameter features respectively, fit the current location, and predict the future environmental parameters.

[0020] In one possible implementation, the neural network model includes an LSTM network layer, a fully connected layer, and an output layer;

[0021] The LSTM network layer is used to learn the temporal features in the input sequence; the fully connected layer is used to predict the position and / or environmental parameters corresponding to the input sequence of the model based on the temporal features.

[0022] The number of units in the output layer is determined based on the dimensions of location and / or environmental parameters.

[0023] In one possible implementation, the method further includes:

[0024] After fusing the current location features and the environmental parameter features, the input is used to predict the future environmental parameters of the current location using a fully connected layer.

[0025] In one possible implementation, the process of generating the inertial navigation data includes:

[0026] A. Obtain the acceleration and angular velocity of the target's movement from the inertial measurement element in the moving node carried by any target;

[0027] B. Based on the acceleration and angular velocity, the PDR (Pedestrian Dead Retrieval) algorithm is used to calculate the number of steps, step length, and heading angle;

[0028] C. Based on the target's current position, use the number of steps, step length, and heading angle to predict the target's position at a future time.

[0029] In one possible implementation, step B includes:

[0030] B1. Perform step count detection and calculate the number of steps;

[0031] B2. For each step, the step size is obtained by introducing the gait category influence factor into the nonlinear step size model.

[0032] B3. For each step, according to the weighted parameters corresponding to the gait category, the heading angles of the first m time steps are summed to obtain the heading angle, where m is a positive integer;

[0033] Step C includes: for the coordinates of the current position in any direction, weighting the step size of the nth step with the component of the heading angle in the corresponding direction, and then superimposing it on the coordinates of the current position in the corresponding direction to obtain the position after moving n steps from the current position, where n is a positive integer.

[0034] In one possible implementation, the method further includes:

[0035] S3. Update the fire rescue route at the fire scene in real time based on the target location information.

[0036] In one possible implementation, step S3 includes:

[0037] The environmental change trend of the fire scene is determined based on the environmental parameters, and a fire rescue route that meets the rescue requirements is generated based on the distribution location of each target and the environmental change trend.

[0038] The newly generated fire rescue route is sent to the target device, which then uses augmented reality technology to display the fire rescue route, the distribution location of each target, and the trend of environmental changes.

[0039] Secondly, a fire rescue positioning system is provided, the system comprising: a service platform and data nodes;

[0040] The data nodes are used to: collect sensor data, wireless positioning data, and inertial navigation data at the fire scene;

[0041] The service platform is used to: acquire sensor data, inertial navigation data and wireless positioning data of the fire scene from the data nodes; fuse the sensor data, inertial navigation data and wireless positioning data into a time series; input the time series into a trained neural network model to obtain target positioning information, the target positioning information including the distribution location and environmental parameters of each target in the fire environment.

[0042] In one possible implementation, the data node includes:

[0043] A fixed access node and a portable mobile node are fixed at the fire scene; after the mobile node enters the fire scene, it establishes a communication network with the fixed access node.

[0044] The fixed access node is used to: collect sensor data and wireless positioning data at the fire scene and send them to the service platform; the mobile node is used to: collect inertial navigation data at the fire scene, send the inertial navigation data to the fixed access node, and then have the fixed access node send it to the service platform.

[0045] In one possible implementation, the system further includes additional functional modules for performing corresponding steps in the fire rescue positioning method provided in the first aspect.

[0046] Thirdly, a computing device is provided, comprising a memory and a processor, the memory storing at least one program, the at least one program being executed by the processor to implement the fire rescue positioning method as provided in the first aspect.

[0047] Fourthly, a computer-readable storage medium is provided, wherein at least one program is stored therein, the at least one program being executed by a processor to implement the fire rescue positioning method as provided in the first aspect.

[0048] Fifthly, a computer program product is provided, comprising a computer program / instructions, characterized in that, when the computer program / instructions are executed by a processor, they implement the fire rescue positioning method provided in the first aspect.

[0049] This application includes at least the following technical effects:

[0050] By fusing heterogeneous data sources such as inertial navigation data, wireless positioning data, and environmental data at the fire scene, accurate and continuous positioning information and environmental monitoring can be provided, significantly enhancing the accuracy of positioning and thus providing reliable data support for fire rescue. Attached Figure Description

[0051] Figure 1This is a schematic diagram of a fire rescue positioning system provided in an embodiment of this application;

[0052] Figure 2 This is a flowchart of a fire rescue positioning method provided in an embodiment of this application;

[0053] Figure 3 This is a schematic diagram of a PDR provided in an embodiment of this application;

[0054] Figure 4 This is a flowchart illustrating the positioning process of an inertial navigation system provided in an embodiment of this application;

[0055] Figure 5 This is a flowchart of another fire rescue positioning method provided in the embodiments of this application;

[0056] Figure 6 This is a schematic diagram of the hardware structure of a computing device provided in an embodiment of this application. Detailed Implementation

[0057] To further illustrate the various embodiments, this application provides accompanying drawings. These drawings are part of the disclosure of this application and are mainly used to illustrate the embodiments, and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementations and the advantages of this application. Components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components. In this application, the term "at least one" means one or more, and the term "multiple" means two or more; for example, multiple nodes refer to two or more nodes.

[0058] To address the positioning challenges within buildings during fire and rescue operations, existing solutions suffer from reliance on fixed structures and poor real-time data updates. Therefore, this application proposes a fire and rescue positioning method and system that significantly improves the accuracy and reliability of indoor positioning in fire and rescue scenarios.

[0059] The technical solution provided in this application will now be further described in conjunction with the accompanying drawings and specific embodiments.

[0060] Example 1

[0061] This application provides a fire rescue positioning system, which includes a service platform and data nodes.

[0062] Among them, data nodes are used to collect sensor data, wireless positioning data, and inertial navigation data at the fire scene.

[0063] In one possible implementation, the data nodes include: fixed access nodes stationed at the fire scene and portable mobile nodes. There may be one or more mobile nodes, and since the fixed access nodes need to perform triangulation based on the signal strength of multiple locations, there are at least three fixed access nodes.

[0064] The mobile node includes at least an Inertial Measurement Unit (IMU), which may include accelerometers and gyroscopes. The IMU mainly consists of three single-axis accelerometers and three single-axis gyroscopes. The accelerometers detect the acceleration signals of the object along three independent axes of the carrier coordinate system, while the gyroscopes detect the angular velocity signals of the carrier relative to the navigation coordinate system. After processing these signals, the attitude and position of the corresponding target can be calculated.

[0065] For example, a portable mobile node is a wearable monitoring device. Firefighters can wear this monitoring device and use its accelerometer and gyroscope for inertial navigation. Based on this, the mobile node can be used to detect the movement of rescue personnel or equipment.

[0066] Optionally, the mobile node also includes temperature and gas sensors to collect environmental parameters around firefighters. Gas sensors include, for example, CO2 sensors, CO sensors, smoke sensors, PM2.5 sensors, and TVOC sensors (for detecting volatile organic compounds), etc., and this application is not limited to these. Of course, different sensor combinations can be used depending on the specific fire rescue scenario. For example, in indoor fire rescue, a combination of CO2 sensors, temperature sensors, and smoke sensors can be used; in rescue operations involving hazardous gas leaks, a combination of TVOC sensors and temperature sensors can be used. Furthermore, in other rescue scenarios, humidity sensors can be used as an extension; this application is not limited to the examples mentioned above.

[0067] This application improves the system's responsiveness and positioning accuracy in complex environmental conditions by integrating multiple sensors (temperature, humidity, CO2, smoke, etc.) and positioning technology.

[0068] In one possible implementation, the fixed access node stationed at the fire scene is, for example, a wireless access point (AP) that supports wireless communication protocols. Exemplarily, the fixed access node includes a WiFi module to support self-organizing networks with mobile nodes and to assist in positioning via RSSI (Resonance Signal Strength Index). Of course, other wireless communication modules can also be used, such as LoRa, Bluetooth, etc.; this application is not limited to WiFi as a fixed communication technology.

[0069] In one possible implementation, automatic networking occurs between mobile nodes and fixed access nodes. Specifically, when a mobile node is carried by rescue personnel or equipment to the fire scene, it can establish a communication network with fixed access nodes deployed at the fire scene. The fixed access nodes are used to: collect sensor data and wireless positioning data at the fire scene and send them to the service platform; the mobile nodes are used to: collect inertial navigation data at the fire scene, send the inertial navigation data to the fixed access nodes, and then the fixed access nodes send it to the service platform.

[0070] Optionally, fixed access nodes and mobile nodes can use MESH technology to form a self-organizing network. MESH is a network architecture based on wireless self-organizing network technology, consisting of multiple nodes, each of which can communicate directly with other nodes without the need for a central control node or infrastructure support. This network architecture features self-organization, decentralization, and flexibility, making it suitable for complex scenarios such as fire rescue. Of course, other networking protocols can also be used, and this application is not limited to them.

[0071] The service platform is used to: acquire sensor data, inertial navigation data and wireless positioning data from data nodes at the fire scene; fuse the sensor data, inertial navigation data and wireless positioning data into a time series; input the time series into a trained neural network model to obtain target positioning information, which includes the distribution location and environmental parameters of each target in the fire environment.

[0072] In one possible implementation, the fire rescue positioning system undergoes system setup and preparation before application. First, multiple fixed access nodes are deployed in the indoor environment. Each fixed access node is equipped with a WiFi module, and the exact location (a, b, c) of each fixed access node is configured in the system. Specifically, when deploying the fixed access nodes, at least three nodes are evenly distributed according to the triangulation requirements of signal strength ranging to ensure good signal coverage.

[0073] Furthermore, when selecting hardware, it is necessary to ensure that all node devices used for positioning support the wireless communication protocol used (such as WiFi) and can obtain signal strength.

[0074] In one possible implementation, the service platform can be based on any computing device. The computing device can be a server, a server cluster consisting of multiple physical servers, or a distributed file system, or a cloud server cluster that provides basic cloud computing services such as cloud storage, cloud services, cloud databases, cloud computing, cloud functions, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), big data, and artificial intelligence platforms. This application does not limit the scope of the application.

[0075] For example, the service platform is a cloud platform that acts as a fire command center, capable of remotely acquiring data from data nodes, analyzing it, and issuing decisions. Optionally, the cloud platform can provide a user interface that displays in real time the analyzed location information, environmental data (including data acquired from fixed access nodes and data acquired from mobile wearable monitoring devices), and real-time updated rescue routes.

[0076] Figure 1 This is a schematic diagram of a fire rescue positioning system provided in an embodiment of this application, such as... Figure 1 As shown, the system includes a service platform, portable mobile nodes, and fixed access nodes. The fixed access nodes and wearable mobile nodes can form a self-organizing Wi-Fi network based on a mesh architecture. After networking, the fixed access nodes can collect data from temperature and smoke sensors while achieving wireless positioning, and the mobile nodes can perform inertial navigation positioning. This data is then routed to the service platform via Wi-Fi, where the service platform performs target positioning and environmental analysis.

[0077] The system provided in this application takes into account the rapidly changing nature of the fire scene environment and adopts an architecture of mobile nodes, fixed nodes, and a remote service platform. Fixed and mobile nodes can temporarily form a self-organizing network on-site to achieve rapid data exchange. Data is then returned to the service platform to support rapid location and decision-making in the background. Traditional static path planning methods often fail to meet actual needs when new obstacles or hazardous areas appear on-site. This application, however, can achieve precise location based on real-time, continuously updated environmental and location data, thus supporting dynamic path planning. Based on this, rescue routes can be adjusted in real time to avoid hazardous areas, ensuring the safety of firefighters and the efficiency of rescue operations.

[0078] Example 2

[0079] This application provides a fire rescue positioning method, which can be applied to the system architecture or scenario provided in Embodiment 1 above. The technical solution of the fire rescue positioning method proposed in this application will be described in detail below. Figure 2 This is a flowchart of a fire rescue positioning method provided in an embodiment of this application, see reference. Figure 2 The technical solution provided in this application includes the following steps S1 to S2, which are executed by a computing device, that is, by the service platform in the above system architecture.

[0080] S1. Acquire sensor data, inertial navigation data, and wireless positioning data at the fire scene.

[0081] In one possible implementation, the mobile node is carried on personnel / equipment and, upon entering the fire scene, establishes a communication network with multiple fixed access nodes at the fire scene.

[0082] Taking MESH technology as an example, the process of building a communication network includes:

[0083] 1. Node Discovery and Connection: In a MESH ad hoc network, mobile nodes and fixed access nodes first need to discover each other. This is typically accomplished through wireless signal broadcasting or specific protocols. Once discovered, nodes attempt to establish a connection. This process includes steps such as authentication and key exchange to ensure the security and reliability of communication.

[0084] 2. Network Topology Construction: After connections are established between nodes, network topology construction begins. Nodes automatically select the optimal communication path and routing strategy based on the network environment and communication needs. This self-organizing nature enables MESH self-organizing networks to adapt to various complex network environments.

[0085] 3. Data Transmission and Routing: After the network topology is established, mobile nodes and fixed access nodes begin exchanging data. For example, data is transmitted between nodes using multi-hop communication, with each node (mobile and fixed access nodes) acting as both a router and a terminal device. When a node needs to send data, it forwards the data packet to its neighboring nodes, which then forward it until the packet reaches the destination node. This routing strategy ensures that data is transmitted along the optimal path within the network, improving communication efficiency and reliability.

[0086] In one possible implementation, step S1 includes:

[0087] S11. Obtain at least one sensor data from the temperature sensor and gas sensor at the fire scene.

[0088] The sensor data is determined based on the type of sensor used. For example, for gas sensors such as CO2 sensors and smoke sensors, the sensor data will at least include the ambient smoke concentration and CO2 concentration.

[0089] Specifically, the temperature t, smoke concentration l, and carbon dioxide concentration q at the current location can be used to obtain the flue gas state matrix (t, l, q).

[0090] In this embodiment, data from environmental sensors (such as temperature, smoke, and gas sensors) will be fully utilized to enhance the accuracy of target localization.

[0091] S12. For any target at the fire scene, obtain the target's inertial navigation data from the mobile node carried by the target.

[0092] In one possible implementation, the mobile node sends inertial navigation data to the fixed access node, which then uploads it to the service platform.

[0093] Inertial navigation data is obtained by processing data such as triaxial acceleration and angular velocity collected by inertial measurement elements to obtain data such as position, velocity, and attitude.

[0094] In one possible implementation, the process of generating inertial navigation data includes:

[0095] A. Obtain the acceleration and angular velocity of the target's movement from the inertial measurement element in the moving node carried by any target.

[0096] B. Based on acceleration and angular velocity, the number of steps, step length, and heading angle are calculated using PDR (Pedestrian Dead Retrieval).

[0097] Pedestrian Dead Reckoning (PDR) is widely used in indoor pedestrian positioning. The PDR algorithm primarily analyzes triaxial acceleration and angular velocity data collected from inertial sensors to determine the number of steps a pedestrian takes, calculate step length and heading angle, and then predict the next position based on the known current position. This algorithm relies on recursive estimation of position over continuous time.

[0098] Figure 3 This is a schematic diagram of a PDR provided in an embodiment of this application. Figure 3 In the diagram, the horizontal axis x represents the target's lateral distance, and the vertical axis y represents the target's longitudinal distance; (x i y i Let ) be the target's position at time i, ψi be the target's heading angle at time i, and L be the target's heading angle at time i. i It is the step size of the i-th step; (x i+1 y i+1 ) is the position of the target at time i, ψ i+1 L is the heading angle of the target at time i+1. i+1 It is the step size of the (i+1)th step;; (x i+2 y i+2 ) is the position of the target after the (i+1)th step.

[0099] based on Figure 3 This application provides a formula for PDR to locate a target position, see formula (1).

[0100]

[0101] In formula (1), (x i ,y iL represents the initial position of the pedestrian target. k Let ψ be the step size at the k-th step. k Let x be the heading angle at step k, (x) j ,y j ) represents the position of the target after moving k steps.

[0102] From the above Figure 3 As can be seen from formula (1), the key to the PDR positioning algorithm is to determine the pedestrian's step length and heading angle. The accuracy of the step length and heading angle calculation will greatly affect the positioning accuracy.

[0103] In one possible implementation, the positioning process using the PDR algorithm mainly includes: step count detection, step length calculation, and heading angle calculation. The implementation flow is as follows: Figure 4 As shown, the accelerometer in the inertial measurement unit (IMU) enables step count detection and step length estimation; based on the gyroscope and magnetometer in the IMU, the heading angle can be determined, and the direction can be estimated; by combining the step length estimation result and the direction estimation result, pedestrian dead reckoning can be achieved. Figure 4 This is a positioning flowchart of an inertial navigation system (INS) provided in an embodiment of this application.

[0104] Based on the PDR principle, this application embodiment further introduces gait type-related factors to optimize the calculation accuracy of heading angle and stride length. In this example, step B includes:

[0105] B1. Perform step count detection and calculate the number of steps;

[0106] Specifically, the step count is performed according to steps one through five below.

[0107] Step 1: Determine whether the current acceleration magnitude is less than the set lower threshold Tmin. If yes, proceed to Step 2. Otherwise, determine if there is a minimum value. If yes, proceed to Step 3. Otherwise, continue the loop.

[0108] Step 2: Determine if the current acceleration modulus is less than the trough value Min. If so, record the current time as the trough time t1, record the trough value, and set FlagMin to 1. Then jump back to Step 1;

[0109] Step 3: Determine whether the current acceleration magnitude is greater than the set threshold upper limit Tmax. If yes, proceed to step 4; otherwise, determine whether there is already a maximum value. If yes, proceed to step 5; otherwise, return to step 1.

[0110] Step 4: Determine if the current acceleration modulus is greater than the peak value Max. If so, record the current time as the peak time t2, record the peak value Max, and set FlagMax to 1. Then jump back to Step 1. Otherwise, determine if the difference between the current time and t1 is greater than T3. If so, set FlagMin to zero. Then execute Step 1.

[0111] Step 5: Determine if the time difference between the peak and trough falls within the set time range. If yes, increment the step count, clear intermediate parameters such as peak and trough values, set the flag to zero, and return to the beginning of Step 1 for the next round of step counting. Otherwise, set FlagMax to zero and execute Step 1.

[0112] B2. For each step, introduce gait category parameters into the nonlinear step size model to estimate the corresponding step size.

[0113] In this embodiment, after step detection is completed, the step length for each step of the target is calculated. The accuracy of the step length calculation directly affects the positioning accuracy, and the positioning error accumulates continuously as the moving distance increases.

[0114] For example, step size estimation can be performed using any of the constant method, linear model method, nonlinear model method, and machine learning method.

[0115] Taking the nonlinear model method as an example, the nonlinear step size model used for step size estimation is shown in formula (2).

[0116]

[0117] In formula (2), L represents the step size, a max For the maximum acceleration, a min K is the minimum acceleration, and K is a variable constant.

[0118] This application proposes a technical solution that incorporates gait category into the step length estimation algorithm. After introducing the gait category parameter into the nonlinear step length model, the nonlinear step length model is shown in formula (3).

[0119]

[0120] In formula (3), K i and a i These are gait category parameters related to the pedestrian's gait category, with each gait corresponding to an independent K. i a i Where f is the current step frequency. Based on this, the step length estimation model will adaptively change according to the different gait of pedestrians, thus better reflecting the actual walking scenario of pedestrians and effectively reducing the error caused by step length estimation.

[0121] B3. For each step, according to the weighted parameters corresponding to the gait category, the heading angles of the first m time steps are weighted and summed to obtain the heading angle, where m is a positive integer.

[0122] Considering that the heading angle of a target is related to its gait while it is moving—for example, a decrease in speed or a change in gait type from climbing stairs to walking might be caused by turning (a change in heading angle)—this application proposes a heading angle determination method that takes gait type into account.

[0123] For example, the improved heading angle determination method in step B3 can be represented by formula (4).

[0124] ψ=a i ψ i +a i-1 ψ i-1 +a i-2 ψ i-2 (4)

[0125] In formula (4), i represents the current time, and ψ i Let ψ be the heading angle at time i. i-1 Let ψ be the heading angle at the previous moment. i- 2 represents the heading angle at the first two moments; a i a i-1 a i-2 The weighting coefficients, which are affected by gait type, can be preset.

[0126] Based on this, the technical solution of this application can adapt well to complex indoor scenarios such as high-rise buildings, improve the accuracy of heading angle, and thus improve the accuracy of indoor positioning.

[0127] C. Based on the target's current position, use the number of steps, step length, and heading angle to predict the target's position at future moments.

[0128] Specifically, for the coordinates of the current position in any direction, the step size of the nth step is weighted by the component of the heading angle in the corresponding direction, and then superimposed onto the coordinates of the current position in the corresponding direction to obtain the position after moving n steps from the current position.

[0129] The above process can be used for two-dimensional planar positioning, for example, for coordinates x in the x-direction. i Using the heading angle ψ at step k k The component sinψ in the x-direction k For the step size L of the k-th step k We perform a weighted summation to obtain the displacement L in the x-direction resulting from the k-th step of the action. k sinψ k Similarly, for the coordinate y in the y-direction... iUsing the heading angle ψ at step k k The component cosψ in the y direction k For the step size L of the k-th step k We perform a weighted summation to obtain the displacement L in the y-direction resulting from the k-th step of the action. k cosψ k .

[0130] In some implementation scenarios, such as during a high-rise fire, elevators should generally not be used for movement. This means that in addition to displacement in the two-dimensional x and y directions, there is also displacement in the height direction. Therefore, the technical solution provided in steps A to C of this application can achieve gait recognition when going up and down stairs, thereby realizing three-dimensional positioning indoors.

[0131] Based on the above calculation principle of the two-dimensional plane, the vertical direction z can be further located. If the coordinates at time i (x... i y i , z i Starting from the initial position, the current position (x) of the target can be obtained by superimposing the displacement and height increase from the i-th to the k-th steps. j y j , z j ).

[0132] In one possible implementation, the indoor three-dimensional positioning achieved in step C can be achieved by formula (5).

[0133]

[0134] In formula (5), (x i ,y i ,z i L represents the three-dimensional position of the target at time i, which is also the initial position of the target during the localization process; k H is the step size for the k-th step. k Let ψ be the vertical displacement at step k. k Let x be the heading angle at step k, (x) j ,y j ,z j H represents the position of the target after moving k steps. Optionally, H... k It is related to the climbing height corresponding to the step frequency, for example, it could be the height of one step.

[0135] Through the above process, the inertial navigation system obtains the first target position (x1, y1, z1) after collecting data based on the inertial measurement element.

[0136] S13. Obtain the signal strength recorded by multiple fixed access nodes at the fire scene, and determine the wireless positioning data of the target based on the multiple signal strengths.

[0137] The wireless positioning data is calculated based on the signal strength detected by multiple fixed access nodes for the same target. The wireless positioning data includes the location of a second target determined by wireless devices.

[0138] In this embodiment, signal strength data is first collected by fixed access nodes deployed at the fire scene. When rescue personnel enter the fire scene using wearable mobile nodes, the built-in WiFi module forms a self-organizing network with the fixed access nodes on site. While collecting and transmitting environmental parameters such as temperature and smoke, the signal strength received from each visible AP is measured and recorded.

[0139] After receiving the signal strength data, distance estimation is performed, and finally wireless positioning data is obtained.

[0140] Taking RSSI as a signal strength metric as an example, a predefined signal attenuation model (e.g., a Log-distance path loss model) is used to convert RSSI values ​​into estimated distances. The parameters of this signal attenuation model can be adjusted according to the specific field environment to match actual signal attenuation behavior.

[0141] After determining the estimated distances of each fixed access point (AP), triangulation is performed to determine the secondary target location (i.e., the wireless positioning location) of the target in the field.

[0142] Specifically, the target's location is determined using the estimated RSSI distances of at least three access points (APs). In some embodiments, the location is determined using the estimated RSSI distances of five nodes. The selection of these five nodes ensures that their locations cover the entire field, and the three-dimensional coordinates (x, y, y) of each AP are accurately recorded. i ,y i ,z i Specifically, the solution equation for the target location is shown in formula (6).

[0143]

[0144] In formula (6), It is the estimated distance from the i-th AP device to the target, (x i ,y i ,z iLet (x, y, z) be the position of the i-th AP, and (x, y, z) be the position of the second target to be solved. Using the numerical method of least squares to solve formula (6), the most probable position (x, y, z) of the target (moving node) can be obtained. p ,y p ,z p ).

[0145] In this embodiment, fixed access node positioning and inertial navigation positioning are fused for positioning. Deep learning is used to improve the problem of inaccurate GPS positioning in indoor environments. In fire environments, fixed access nodes and wearable temperature and smoke sensors are quickly networked. The back-end fire command center can quickly generate rescue decision plans based on detailed environmental conditions and location of the fire environment.

[0146] S2. The sensor data, inertial navigation data and wireless positioning data are fused into a time series, and the time series is input into the trained neural network model to obtain the target positioning information.

[0147] The target location information includes the distribution location and environmental parameters of each target in the fire protection environment.

[0148] In one possible implementation, step S2 includes:

[0149] S21. The sensor data, inertial navigation data, and wireless positioning data are concatenated into a time series input sequence in chronological order. Each time step in the time series contains sensor data, inertial navigation data, and wireless positioning data at the same moment.

[0150] In the field of time series forecasting, a time step refers to the number of consecutive observations used for forecasting, which can be understood as the length of a set of consecutive sample points used for forecasting.

[0151] For example, the input to the neural network model includes the first target position (x1, y1, z1) from the inertial navigation system and the second target position (x1, y1, z1) obtained by RSSI positioning. p ,y p ,z p The input data may include sensor data (t, l, q). Optionally, the input data may include measurement data from inertial measurement elements, such as velocity and acceleration data, for predicting the future position of the target.

[0152] S22. Input the input sequence into the trained neural network model to extract the current position features and environmental parameter features.

[0153] In one possible implementation, the neural network model includes a Long Short-Term Memory (LSTM) recursive network layer, a fully connected layer, and an output layer.

[0154] The LSTM network layer is used to learn the temporal features in the input sequence; the fully connected layer is used to predict the location and / or environmental parameters corresponding to the input sequence of the model based on the temporal features.

[0155] S23. Predict the current location features and environmental parameter features respectively, fit the current location, and predict the future environmental parameters.

[0156] The number of units in the output layer is determined based on the dimensions of the position and / or environmental parameters. For example, if the position has three dimensions (x, y, z), then the number of units in the output layer is 3.

[0157] In one possible implementation, future environmental parameters and the current location are further integrated to predict the future environmental conditions of the current location. Specifically, this involves fusing the current location features and environmental parameter features, then using a fully connected layer for prediction to obtain the future environmental parameters of the current location.

[0158] The following are some specific examples to illustrate the process of steps S21 to S23 above.

[0159] (1) Location information processing part: Input the input sequence into the trained neural network model, extract the current position features, and fit the current position.

[0160] 1) Model Input: Input time series data (x p ,y p ,z p (x1, y1, z1);

[0161] 2) Feature extraction using LSTM layers:

[0162] The first layer of LSTM captures complex time dependencies and outputs sequences.

[0163] The second LSTM layer further extracts information and outputs the features of the last time step.

[0164] 3) Refinement features of fully connected layers:

[0165] The extracted features are further processed through a fully connected layer to output the current position.

[0166] For example, the specific network layer design is as follows:

[0167] First LSTM layer:

[0168] Input: Location information from the time series (including inertial navigation data and wireless positioning data).

[0169] Number of units: 128

[0170] Return sequence: True (outputs the entire sequence to be passed to the next level).

[0171] Second LSTM layer:

[0172] Input: The output sequence of the first LSTM layer.

[0173] Number of units: 64

[0174] Return sequence: False (only outputs the last value of the sequence).

[0175] Fully connected layer:

[0176] Input: The output of the second LSTM layer.

[0177] Number of units: 64

[0178] Activation function: ReLU

[0179] Output: Current position.

[0180] (2) Environmental data processing part: Input the input sequence into the trained neural network model, extract the environmental parameter features, and predict the future environmental parameters.

[0181] 1) Input time series data (t, l, q).

[0182] 2) Feature extraction using LSTM layers:

[0183] The first layer of LSTM captures the temporal dependencies of environmental data and outputs a sequence.

[0184] The second LSTM layer further processes the data and outputs the features of the last time step.

[0185] 3) Fully connected layer generation prediction:

[0186] The extracted features are passed through a fully connected layer to output predictions of future temperature, smoke concentration, and carbon dioxide concentration.

[0187] The network layer design is as follows:

[0188] First LSTM layer:

[0189] Input: Time series environmental data (t, l, q).

[0190] Number of units: 128

[0191] Return sequence: True

[0192] Second LSTM layer:

[0193] Input: The output sequence of the first LSTM layer.

[0194] Number of units: 64

[0195] Return sequence: False

[0196] Fully connected layer:

[0197] Input: The output of the second LSTM layer.

[0198] Number of units: 64

[0199] Activation function: ReLU

[0200] Output: Predictions of future environmental conditions.

[0201] (3) Fusion prediction part: The results of the two parts are fused and processed by a fully connected layer to output the environmental state prediction of the future location.

[0202] Fusion layer: Connects the processing results of location information and environmental data.

[0203] Output layer: 3 units, linear activation, outputting predicted temperature t, smoke concentration l, and carbon dioxide concentration q.

[0204] For example, the specific network design is as follows:

[0205] 1) Input: Output features from the localization processing section and the environment prediction section.

[0206] Operation: Concatenate the two feature vectors.

[0207] 2) Processing flow

[0208] Receives location and environmental feature outputs;

[0209] The splicing and fusion operation involves splicing the output feature vectors of the localization part and the environment prediction part along the feature dimension to form a comprehensive feature vector.

[0210] 3) Output layer

[0211] Input: The fused composite feature vector.

[0212] Number of units: 6 (x,y,z,t) p ,y p ,z p This corresponds to location information and predicted temperature, smoke concentration, and carbon dioxide concentration.

[0213] Activation function: Linear (suitable for regression problems, directly outputs predicted values).

[0214] In one possible implementation, the training process of the above-described neural network model includes:

[0215] Data annotation: When processing training samples, the true location coordinates (x, y, z) of the data are labeled. r ,y r ,z r () is used as a label, and correspondingly, the actual future environmental parameters are used as labels.

[0216] Data preprocessing includes data cleaning (removing outliers and filling in missing values) and standardizing or normalizing the data to make it suitable for neural network processing.

[0217] Designing an LSTM model: Design a network containing multiple LSTM layers:

[0218] Model training: The Adam optimizer is used, employing mean squared error (MSE) as the loss function to optimize the accuracy of location and environment predictions. Optionally, the learning rate is set to 0.001. The batch size is set to 32 or 64, adjusted according to available computing resources and the size of the dataset.

[0219] Optionally, during the model fitting process, preprocessed data can be input into the model for training, and a validation set can be used to monitor the performance during the training process to prevent overfitting.

[0220] Optionally, model evaluation can be performed. The model's performance is evaluated on a test set to ensure it has good generalization ability.

[0221] The technical solution provided in this application possesses enhanced real-time data processing capabilities: by using deep learning techniques (such as LSTM), the system can learn from time-series data and predict future states, achieving efficient real-time data processing. This approach improves the system's ability to handle complex datasets, ensuring rapid response and decision-making in emergency situations, such as quickly navigating firefighter evacuation or intervention routes, thereby improving rescue efficiency and safety.

[0222] Furthermore, by processing data collected from environmental sensors and fusion positioning systems (including inertial navigation and WiFi self-organizing networks) in real time, the system can adjust rescue strategies and route planning in an instant, reducing reliance on outdated information.

[0223] In one possible implementation, after the target location information is determined, the service platform makes a further rescue decision, and the decision-making process includes the following step S3. Figure 5 This is a flowchart of another fire rescue positioning method provided in the embodiments of this application, such as... Figure 5 As shown, the method includes the steps S1 to S2 described above and the step S3 below.

[0224] S3. Update the fire rescue route at the fire scene in real time based on the target location information.

[0225] Specifically, step S3 includes:

[0226] S31. Determine the environmental change trend at the fire scene based on environmental parameters, and generate a fire rescue route that meets the rescue requirements based on the distribution location and environmental change trend of each target.

[0227] S32. The newly generated fire rescue path is sent to the target equipment. The target equipment uses augmented reality technology to display the fire rescue path, the distribution location of each target, and the trend of environmental changes.

[0228] The technical solution provided in this application has dynamic environmental adaptability: by combining real-time environmental information detection and fusion positioning technology, it can dynamically adapt to environmental changes caused by fire, such as updating path planning by real-time monitoring of changes in walls and passages. This enables the positioning system to not only locate based on static environment, but also adapt to dynamic changes (such as smoke and temperature changes at the fire scene), enhancing the system's performance and availability under various environmental conditions.

[0229] Furthermore, compared to existing indoor positioning technologies (such as single RSSI positioning or INS-only positioning) or GPS, which are often limited by environmental factors such as multipath effect and signal interference, resulting in insufficient positioning accuracy and reliability, the technical solution of this application integrates data from RSSI and inertial navigation to compensate for the shortcomings of each technology. For example, by utilizing the continuous tracking capability of INS and the stability of RSSI, the impact of error accumulation and signal interference can be effectively reduced, thereby improving the overall positioning accuracy and reliability.

[0230] This application also provides a computing device that can be used to perform the above-described fire rescue positioning method. Figure 6 This is a schematic diagram of the hardware structure of a computing device provided in an embodiment of this application, such as... Figure 6 As shown, the computing device includes a processor 601, a memory 602, a bus 603, and a computer program stored in the memory 602 and executable on the processor 601. The processor 601 includes one or more processing cores. The memory 602 is connected to the processor 601 via the bus 603. The memory 602 is used to store program instructions. When the processor executes the computer program, it implements all or part of the steps in the above-described method embodiments provided in this application.

[0231] Furthermore, as an executable solution, the aforementioned electronic device can be a computer unit, which may be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer unit may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above-described computer unit structure is merely an example and does not constitute a limitation on the computer unit; it may include more or fewer components, or combine certain components, or different components. For example, the computer unit may also include input / output devices, network access devices, buses, etc., which are not limited in this application.

[0232] Furthermore, as an executable solution, the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The processor is the control center of the computer unit, connecting various parts of the entire computer unit via various interfaces and lines.

[0233] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the computer unit by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a function; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0234] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements all or part of the steps of the fire rescue positioning method described in the embodiments of this application.

[0235] This application also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement all or part of the steps of the fire rescue positioning method described in the embodiments of this application.

[0236] If the modules / units integrated by the aforementioned computing units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), and a software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction.

[0237] Although this application has been specifically shown and described in conjunction with preferred embodiments, those skilled in the art should understand that various changes in form and detail may be made to this application without departing from the spirit and scope of this application as defined by the appended claims, and all such changes shall be within the scope of protection of this application.

Claims

1. A fire rescue positioning method, characterized in that, The method includes: S1. Acquire sensor data, inertial navigation data, and wireless positioning data at the fire scene; S2. The sensor data, inertial navigation data, and wireless positioning data are fused into a time series, and the time series is input into a trained neural network model to obtain target positioning information. The target positioning information includes the distribution location and environmental parameters of each target in the fire scene; wherein, step S1 includes: S11. Obtain at least one sensor data from the temperature sensor and gas sensor at the fire scene; S12. For any target in the fire scene, obtain the inertial navigation data of the target from the mobile node carried by the target; S13. Obtain the signal strength recorded by multiple fixed access nodes at the fire scene, and determine the wireless positioning data of the target based on the multiple signal strengths; The mobile node and multiple fixed access nodes at the fire scene form a communication network; The mobile node sends inertial navigation data to the fixed access node; Step S2 includes: S21. The sensor data, inertial navigation data and wireless positioning data are concatenated into a time series format input sequence according to the time order. Each time step in the time series contains sensor data, inertial navigation data and wireless positioning data at the same time. S22. Input the input sequence into the trained neural network model to extract the current position features and environmental parameter features; S23. Predict the current location features and the environmental parameter features respectively, fit the current location, and predict the future environmental parameters; The neural network model includes an LSTM network layer, a fully connected layer, and an output layer; The LSTM network layer is used to learn temporal features in the input sequence; the fully connected layer is used to predict the position and / or environmental parameters corresponding to the input sequence of the model based on the temporal features; wherein, the number of units in the output layer is determined according to the dimension of the position and / or environmental parameters.

2. The fire rescue positioning method according to claim 1, characterized in that, The process of generating the inertial navigation data includes: A. Obtain the acceleration and angular velocity of the target's movement from the inertial measurement element in the moving node carried by any target; B. Based on the acceleration and angular velocity, the PDR (Pedestrian Dead Retrieval) algorithm is used to calculate the number of steps, step length, and heading angle; C. Based on the target's current position, use the number of steps, step length, and heading angle to predict the target's position at a future time.

3. The fire rescue positioning method according to claim 2, characterized in that, Step B includes: B1. Perform step count detection and calculate the number of steps; B2. For each step, the step size is obtained by introducing the gait category influence factor into the nonlinear step size model. B3. For each step, according to the weighted parameters corresponding to the gait category, the heading angles of the first m time steps are weighted and summed to obtain the heading angle, where m is a positive integer.

4. A fire rescue positioning system, characterized in that, The system includes: a service platform and data nodes; The data nodes are used to: collect sensor data, wireless positioning data, and inertial navigation data at the fire scene; The service platform is used to: acquire sensor data, inertial navigation data, and wireless positioning data from the data nodes at the fire scene; fuse the sensor data, inertial navigation data, and wireless positioning data into a time series; input the time series into a trained neural network model to obtain target positioning information, the target positioning information including the distribution location and environmental parameters of each target at the fire scene; wherein, the data nodes are also used to: S11. Obtain at least one sensor data from the temperature sensor and gas sensor at the fire scene; S12. For any target in the fire scene, obtain the inertial navigation data of the target from the mobile node carried by the target; S13. Obtain the signal strength recorded by multiple fixed access nodes at the fire scene, and determine the wireless positioning data of the target based on the multiple signal strengths; The mobile node and multiple fixed access nodes at the fire scene form a communication network; The mobile node sends inertial navigation data to the fixed access node; The service platform is also used for: S21. The sensor data, inertial navigation data and wireless positioning data are concatenated into a time series format input sequence according to the time order. Each time step in the time series contains sensor data, inertial navigation data and wireless positioning data at the same time. S22. Input the input sequence into the trained neural network model to extract the current position features and environmental parameter features; S23. Predict the current location features and the environmental parameter features respectively, fit the current location, and predict the future environmental parameters; The neural network model includes an LSTM network layer, a fully connected layer, and an output layer; The LSTM network layer is used to learn temporal features in the input sequence; the fully connected layer is used to predict the position and / or environmental parameters corresponding to the input sequence of the model based on the temporal features; wherein, the number of units in the output layer is determined according to the dimension of the position and / or environmental parameters.

5. A computing device, characterized in that, It includes a memory and a processor, the memory storing at least one program, the at least one program being executed by the processor to implement the fire rescue positioning method as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The storage medium stores at least one program, which is executed by a processor to implement the fire rescue positioning method as described in any one of claims 1 to 3.

Images