Fire-fighting emergency broadcast data information acquisition system

Abstract

The invention relates to the technical field of information acquisition, in particular to a fire-fighting emergency broadcast data information acquisition system which comprises the following steps: performing sound source positioning analysis on original data to obtain sound source distribution data; performing sound propagation loss estimation according to the sound source movement track data to obtain sound propagation loss data; emergency broadcast coverage demand period reference simulation evaluation is carried out according to the sound propagation loss data to obtain period reference data; and according to a strategy gradient algorithm, an emergency broadcast putting strategy adjustment model is constructed through the environment correction data, and the emergency broadcast putting strategy adjustment model is sent to a fire-fighting command cloud platform so as to execute fire-fighting emergency broadcast data information acquisition and analysis. According to the method, the attenuation degree of the broadcast signals under different environment conditions can be analyzed through estimation of the sound propagation loss, data support can be provided for the actual broadcast propagation effect, the broadcast signals can be optimally delivered according to the environment conditions, and therefore effective transmission of the signals is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of information collection, and in particular to a fire emergency broadcast data information collection system. Background Art

[0002] Traditional systems often rely on fixed broadcast modes or static coverage areas, and are unable to accurately locate changes in the position of fire sources or other key sound sources in real time. Therefore, they are unable to adjust broadcast strategies based on dynamic changes on site (such as fire source movement, obstacle changes, etc.), which may lead to insufficient or wasted information coverage. The lack of sound source positioning and dynamic tracking capabilities may not ensure that broadcast information is delivered to the required areas in a timely manner. Traditional emergency broadcast systems generally broadcast according to fixed preset parameters and lack adaptive adjustment mechanisms based on on-site environment and needs. Faced with complex and changing on-site conditions (such as population density, fire spread speed, etc.), traditional systems are often unable to automatically adjust broadcast strategies, resulting in delayed and inflexible broadcast responses. Traditional systems usually ignore the impact of factors such as ambient temperature, humidity, and airflow on sound propagation, and ignore the propagation loss caused by environmental changes, resulting in distortion or insufficient coverage of broadcast effects under complex environmental conditions. In addition, due to the lack of an environmental correction mechanism, it is impossible to evaluate and optimize broadcast signals in real time. Summary of the Invention

[0003] In view of the above-mentioned technical problems existing in the prior art, the purpose of the present invention is to provide a fire emergency broadcast data information collection system to solve the problems of optimizing the real-time collection and analysis of sound data, dynamic environment correction, and intelligent strategy adjustment, so that the fire emergency broadcast system can effectively play a role in complex and constantly changing on-site environments.

[0004] The technical solution adopted to solve the above technical problems is: Fire emergency broadcast data information collection system, including: A sound source localization unit is used to obtain the original data of the fire scene environment sound; perform sound source localization analysis on the original data to obtain sound source distribution data; a sound source analysis unit, configured to perform sound source area analysis based on the sound source distribution data to obtain sound source distribution density data; and to mark the sound source movement trajectory based on the sound source distribution density data to obtain sound source movement trajectory data; a loss estimation unit, configured to estimate the sound propagation loss based on the sound source motion trajectory data to obtain sound propagation loss data; and perform a simulation evaluation of a periodic benchmark of emergency broadcast coverage requirements based on the sound propagation loss data to obtain periodic benchmark data; An environmental identification unit is configured to obtain real-time environmental temperature and humidity data at the fire scene; identify environmental impact factors on the sound propagation loss data based on the real-time environmental temperature and humidity data to obtain propagation impact factors; and perform coverage requirement environmental correction on the periodic baseline data based on the propagation impact factors to obtain environmental correction data; The data acquisition unit is used to construct an emergency broadcast delivery strategy adjustment model through the environmental correction data according to the policy gradient algorithm, and send the emergency broadcast delivery strategy adjustment model to the fire command cloud platform to perform fire emergency broadcast data information collection and analysis.

[0005] Preferably, performing sound source localization analysis on the original data to obtain sound source distribution data includes: Performing multi-channel signal synchronous acquisition on the raw data using a microphone array device to obtain a synchronized sound wave signal; performing consistency calibration on the synchronized sound wave signal and extracting phase features; wherein the characteristic dimension of the phase feature is determined by the number of array elements of the microphone array and the phase dimension of each array element, thereby obtaining multi-channel phase feature data; Performing feature sampling to a target dimension based on the multi-channel phase feature data to obtain a first feature; performing feature extraction based on physical layout attribute information of the microphone array to obtain a second feature of the target dimension; wherein the attribute information includes the array arrangement and the number of array elements, and the target dimension is a unified feature dimension; The first feature and the second feature are combined to obtain a target feature, which is input into a pre-trained sound source localization model; the sound source localization model is trained based on a sample audio set, and the spatial coordinates of the sound source are predicted using the target feature to obtain the sound source position; Performing time-frequency analysis on the sound source position, constructing a time-frequency-space correlation matrix of the sound source based on short-time Fourier transform, and obtaining time-frequency-space distribution data of the sound source; marking the dynamic trajectory of the sound source based on the time-frequency-space distribution data of the sound source, and obtaining trajectory data; Ambient noise analysis is performed based on the trajectory data, and background noise unrelated to the sound source trajectory is removed using an adaptive noise cancellation algorithm to obtain corrected trajectory data. Density clustering of the spatial distribution of sound sources is performed based on the corrected trajectory data to identify spatial aggregation areas of different sound sources to obtain sound source distribution data.

[0006] Preferably, performing sound source area analysis based on the sound source distribution data to obtain sound source distribution density data includes: Constructing a three-dimensional spatial grid model based on the sound source distribution data, dividing the fire scene into multi-scale adaptive grid units, and obtaining spatial grid data; Performing sound source position mapping based on the spatial grid data, matching the spatial coordinates of each sound source to a corresponding grid unit, and counting the number of sound sources in each grid unit to obtain sound source basic density data; Performing time correction on the basic density data of the sound source according to the trajectory data, counting the residence time and movement frequency of the sound source in each grid unit through a sliding time window, and calculating the time-weighted density to obtain the dynamic density data of the sound source; The sound source dynamic density data is corrected for environmental impacts, the density calculation range is adjusted using a sound propagation attenuation model associated with smoke concentration, and the density calculation time interval is corrected using a sound propagation speed model associated with temperature to obtain sound source distribution density data.

[0007] Preferably, marking the sound source movement trajectory according to the sound source distribution density data to obtain the sound source movement trajectory data includes: Constructing a tracking grid according to the sound source distribution density data to obtain tracking grid data; performing density gradient analysis on the sound source distribution density data according to the tracking grid data, calculating the sound source density difference between adjacent grid cells, identifying the possible movement direction of the sound source, and obtaining candidate data of the sound source movement direction; Counting the continuity of the sound source density in each grid unit in the candidate data of the sound source movement direction according to the sliding time window, marking the nodes of the sound source trajectory to obtain sound source trajectory node data; A trajectory prediction model is constructed based on the sound source trajectory node data. The historical spatiotemporal pattern of sound source movement is learned through a long short-term memory network, the trajectory node position of the next time window is predicted, and the actually collected sound source position data is matched and verified to obtain the sound source motion trajectory data.

[0008] Preferably, estimating the sound propagation loss based on the sound source motion trajectory data to obtain the sound propagation loss data includes: Constructing a spatial model of the sound source movement path based on the sound source motion trajectory data, and marking key environmental nodes on the trajectory path to obtain associated model data; Performing sound propagation analysis based on the correlation model data, identifying the sound propagation mode on the trajectory path, calculating the initial propagation loss coefficient under each propagation mode, and obtaining initial propagation loss coefficient data; Correcting the initial propagation loss coefficient by adjusting the loss coefficient in each propagation mode using a sound wave attenuation model associated with smoke concentration and a sound velocity correction model associated with temperature to obtain corrected propagation loss coefficient data; The corrected propagation loss coefficient data is regionally verified according to the sensor, the error between the theoretical loss and the actual loss is calculated, and the loss coefficient is corrected by an adaptive filtering algorithm to obtain the sound propagation loss data.

[0009] Preferably, performing a periodic benchmark simulation assessment of emergency broadcast coverage requirements based on the sound propagation loss data to obtain periodic benchmark data includes: Constructing a coverage requirement model according to the sound propagation loss data to obtain coverage requirement grid data; According to the sound source motion trajectory data and the coverage requirement grid data, the frequency of each grid being covered by the sound source in different time periods is counted to obtain coverage requirement distribution data; Performing coverage simulation on the coverage demand distribution data, calculating the number of high-priority grids that each device can cover in different time periods, to obtain coverage efficiency data; A coverage demand cycle simulation model is constructed based on the coverage efficiency data. The total coverage demand and total coverage supply of high-priority grids in each cycle are counted through a sliding time window, and the matching degree of coverage demand and supply is calculated to obtain cycle benchmark data.

[0010] Preferably, identifying environmental impact factors on the sound propagation loss data based on the real-time environmental temperature and humidity data to obtain the propagation impact factors includes: Real-time environmental temperature and humidity data collection is performed based on the wireless sensor network to obtain the original temperature and humidity data; The raw temperature and humidity data are cleaned and aligned, sensor noise is removed through Kalman filtering, and the temperature and humidity data collected by the drone are mapped to a ground grid coordinate system to obtain temperature and humidity data; Calculating a correlation index between temperature and humidity and propagation loss based on the temperature and humidity data and the sound propagation loss data to obtain correlation index data; An environmental impact factor model is constructed based on the correlation index data, and the temperature correlation coefficient and the humidity correlation coefficient are mapped into impact factor values ​​through a fuzzy logic reasoning system to obtain a propagation impact factor.

[0011] Preferably, performing coverage requirement environment correction on the periodic baseline data according to the propagation impact factor to obtain environment correction data includes: An environmental impact map is constructed based on the propagation impact factor data, and the fire scene is divided into multi-level dynamic grids. Each grid is labeled with a temperature and humidity impact level to obtain impact grid data; wherein the grids include high-impact grids, medium-impact grids, and low-impact grids; Calculating the coverage demand correction coefficient of each grid under the influence of temperature and humidity based on the periodic benchmark data and the affected grid data to obtain coverage supply data; The coverage supply data is corrected twice, and the equipment coverage supply is adjusted through a line of sight obstruction model associated with smoke concentration and a sound propagation deviation model associated with wind speed to obtain environmental correction data.

[0012] Preferably, constructing an emergency broadcast delivery strategy adjustment model using the environmental correction data according to a policy gradient algorithm includes: constructing a state space according to the environmental correction data to obtain a state space feature; Defining an action space of the emergency broadcast delivery strategy to obtain action space parameters; wherein the action space parameters include a delivery period, a delivery power, and a delivery direction, and each action combination corresponds to a strategy parameter vector; A multi-objective reward function is designed, combining coverage requirement satisfaction and environmental adaptability to obtain the reward function calculation rules.

[0013] Preferably, constructing an emergency broadcast delivery strategy adjustment model using the environmental correction data according to a policy gradient algorithm further includes: Constructing a policy gradient learning model based on the state space characteristics, the action space parameters, and the reward function calculation rule, initializing the policy network, and simulating coverage effects under different strategies using the Monte Carlo method to obtain initial policy network parameters; The initial policy network parameters are modified, the state space features are updated through an incremental learning mechanism, and the policy adaptability is maintained through exploration-exploitation balance to obtain a policy adjustment model.

[0014] The beneficial effects of the present invention are as follows: (1) Through sound source localization analysis and sound source area analysis, the present invention can provide a detailed understanding of the distribution and density of on-site sound sources, which helps to locate the core area of ​​a fire or other emergency situation and determine the priority areas for broadcast delivery, thereby avoiding waste of broadcast information. Moreover, by marking the movement trajectory of the sound source, the system can track the changes of the sound source in real time and understand the position changes of the fire source or other key sound sources. This is crucial for adjusting the broadcast strategy in real time, ensuring that the broadcast information can cover all required areas in real time as the fire scene changes. (2) The present invention helps analyze the attenuation degree of broadcast signals under different environmental conditions by estimating the sound propagation loss. The system can provide data support for the actual propagation effect of the broadcast, so that the broadcast signal can be optimized according to the environmental conditions, thereby ensuring the effective transmission of the signal; and through the real-time acquisition of environmental temperature and humidity data, the sound propagation loss is corrected for the environment. Different environmental conditions (such as humidity, temperature, airflow, etc.) will affect the sound propagation effect. The system quantifies the impact of these environmental factors and conducts a post-correction evaluation to further improve the actual effect and accuracy of the broadcast; (3) Through the policy gradient algorithm, the system can construct the emergency broadcast delivery strategy based on the environmental correction data and make dynamic adjustments, so that the emergency broadcast system has the ability of self-learning and optimization. In the complex and changing on-site environment, it can automatically adjust the broadcast strategy and improve the real-time response and efficiency of emergency broadcast. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of the system architecture of the overall system in an embodiment of the present invention; Figure 2 This is a schematic diagram of data flow in an embodiment of the present invention.

[0016] Reference numerals: 1. Sound source localization unit; 2. Sound source analysis unit; 3. Loss estimation unit; 4. Environment recognition unit; 5. Data acquisition unit. DETAILED DESCRIPTION

[0017] Example 1, as Figure 1 - Figure 2 As shown, the fire emergency broadcast data information collection system proposed by the present invention includes: The sound source localization unit 1 is used to obtain the original data of the fire scene environmental sound; perform sound source localization analysis on the original data to obtain sound source distribution data; The sound source analysis unit 2 is used to perform sound source area analysis based on the sound source distribution data to obtain sound source distribution density data; and to mark the sound source movement trajectory based on the sound source distribution density data to obtain sound source movement trajectory data; The loss estimation unit 3 is configured to estimate the sound propagation loss based on the sound source motion trajectory data to obtain sound propagation loss data; and to perform a simulation evaluation of the emergency broadcast coverage requirement period benchmark based on the sound propagation loss data to obtain period benchmark data; Environmental identification unit 4 is used to obtain real-time environmental temperature and humidity data at the fire scene; identify environmental impact factors on the sound propagation loss data based on the real-time environmental temperature and humidity data to obtain propagation impact factors; and perform coverage requirement environmental correction on the periodic baseline data based on the propagation impact factors to obtain environmental correction data; The data acquisition unit 5 is used to construct an emergency broadcast delivery strategy adjustment model through environmental correction data according to the policy gradient algorithm, and send the emergency broadcast delivery strategy adjustment model to the fire command cloud platform to perform fire emergency broadcast data information collection and analysis.

[0018] In the present invention, raw data collection requires collecting environmental sound data at the fire scene; this data includes various environmental noises at the scene, which may include noises such as fire, equipment operation, and personnel activities; once the strategy adjustment model is completed, it will be sent to the fire command cloud platform for actual execution; the cloud platform can integrate various real-time data and schedule and execute emergency broadcasts based on the strategy adjustment model; the cloud platform will collect and analyze emergency broadcast data based on real-time data and the strategy adjustment model to ensure that in emergency situations such as fire, the broadcast system can provide information in a timely and effective manner.

[0019] In an optional embodiment, performing sound source localization analysis on the original data to obtain sound source distribution data includes: The original data is synchronously collected using a microphone array device to obtain a synchronous sound wave signal; the synchronous sound wave signal is calibrated for consistency and then a phase feature is extracted; wherein the characteristic dimension of the phase feature is determined by the number of array elements of the microphone array and the phase dimension of each array element, thereby obtaining multi-channel phase feature data; Performing feature sampling to a target dimension based on the multi-channel phase feature data to obtain a first feature; performing feature extraction based on the physical layout attribute information of the microphone array to obtain a second feature of the target dimension; wherein the attribute information includes the array arrangement and the number of array elements, and the target dimension is a unified feature dimension; The first feature and the second feature are combined to obtain a target feature, which is then input into a pre-trained sound source localization model. The sound source localization model is trained based on a sample audio set and uses the target feature to predict the spatial coordinates of the sound source to obtain the sound source location. Perform time-frequency analysis on the sound source position, construct the time-frequency-space correlation matrix of the sound source based on short-time Fourier transform, and obtain the time-frequency-space distribution data of the sound source; mark the dynamic trajectory of the sound source based on the time-frequency-space distribution data of the sound source to obtain trajectory data; Ambient noise analysis is performed based on the trajectory data, and background noise unrelated to the sound source trajectory is removed using an adaptive noise cancellation algorithm to obtain corrected trajectory data. Density clustering of the spatial distribution of sound sources is performed based on the corrected trajectory data to identify spatial aggregation areas of different sound sources to obtain sound source distribution data.

[0020] It should be noted that multiple microphone arrays collect sound wave signals at the same time to ensure that these signals are synchronized in time; through these synchronized data, the sound propagation process can be analyzed more accurately; the collected synchronized sound wave signals are calibrated, and then the phase features are extracted; the phase features reflect the phase differences of the signals received by each microphone, thereby helping to locate the position of the sound source; the dimension of the feature is related to the number of microphone arrays and the phase characteristics of each microphone; for example: assuming that the array has 4 microphones, the sound signals recorded by each microphone are different in phase, and the phase differences between the collected signals will be used for further positioning calculations; the multi-channel phase feature data extracted from the microphone array will be sampled and scaled to the target dimension for subsequent processing; the physical layout of the microphone array will affect the signal collection effect; using this information, features based on the physical layout can be extracted; these features are related to the position and structure of the array, which helps to improve the accuracy of positioning; the sampled data based on phase features and physical layout features are fused to form a unified target feature; these target features are used as input to the pre-trained In the trained sound source localization model, the model is trained based on existing audio sample data and can predict the spatial location of the sound source based on the target features. For example, assuming that the model is trained to know the relationship between the phase information collected from microphone arrays at different positions and the sound source position, the model can predict the location of the sound source through calculation, such as that the sound source is located on the left side of the room. Time-frequency analysis methods such as short-time Fourier transform (STFT) are used to construct the time-frequency-space correlation matrix of the sound source. This matrix reflects the distribution of the sound source at different time, frequency, and spatial positions. The dynamic trajectory of the sound source is marked and its changing process is tracked through the time-frequency-space distribution data. The environmental noise is analyzed through an adaptive noise cancellation algorithm to remove background noise unrelated to the sound source trajectory, thereby obtaining more accurate sound source trajectory data. By removing noise, clearer corrected trajectory data is obtained. The corrected trajectory data is used to cluster the spatial distribution density of the sound source. This step is used to identify the clustered areas of sound sources in space, thereby helping to analyze the distribution characteristics of the sound sources. Finally, the sound source distribution data is obtained, indicating the distribution of the sound source in space.

[0021] In an optional embodiment, performing sound source area analysis based on the sound source distribution data to obtain sound source distribution density data includes: A three-dimensional spatial grid model is constructed based on the sound source distribution data, and the fire scene is divided into multi-scale adaptive grid units to obtain spatial grid data; Map the sound source positions based on the spatial grid data, match the spatial coordinates of each sound source to the corresponding grid unit, and count the number of sound sources in each grid unit to obtain the basic density data of the sound source; The basic density data of the sound source is time-corrected according to the trajectory data. The residence time and movement frequency of the sound source in each grid unit are counted through a sliding time window, and the time-weighted density is calculated to obtain the dynamic density data of the sound source. The sound source dynamic density data is corrected for environmental impacts. The density calculation range is adjusted using the sound propagation attenuation model associated with smoke concentration. The density calculation time interval is corrected using the sound propagation velocity model associated with temperature to obtain the sound source distribution density data.

[0022] It should be noted that a three-dimensional spatial grid model is created based on the distribution data of the sound source; this means dividing the entire environment (such as a fire scene) into multiple grid cells, each cell representing an area of ​​space; this can help correspond the distribution of sound sources to specific spatial positions; the scale of the grid can be adaptively adjusted according to the different characteristics of the environment; this adaptive division can effectively improve the accuracy of the analysis; by matching the spatial coordinates of each sound source with the cells in the grid model, the specific location of the sound source can be mapped to the corresponding grid cell; the number of sound sources in each grid cell will be counted; this provides basic data for further analysis of the distribution and density of sound sources; by analyzing the trajectory data of the sound source (that is, the movement path of the sound source), the "residence time" and "movement frequency" of the sound source in each grid cell can be counted; the residence time refers to the time the sound source stays in a specific grid cell, and the movement frequency refers to the frequency of the sound source passing through the area; based on the residence time and movement frequency, the time-weighted density of each grid cell is calculated, that is After considering the time factor, the density of the sound source in space changes; this process can make the dynamic characteristics of the sound source be more accurately reflected; in actual firefighting scenarios, environmental factors (such as smoke concentration, temperature, etc.) will affect the propagation of sound waves; therefore, the sound source density needs to be corrected according to the smoke concentration and temperature; smoke concentration affects the propagation of sound waves. The thicker the smoke, the more serious the attenuation of sound propagation; therefore, based on the smoke concentration, the calculation range of the sound source distribution can be adjusted to eliminate the errors caused by environmental factors; temperature affects the propagation speed of sound waves; in a high temperature environment, the sound propagation speed is faster; in a low temperature environment, the sound propagation speed is slower; by considering the change in temperature, the time interval in the sound source distribution data can be corrected to make the density calculation more accurate; through the above steps, the final result is the sound source distribution density data after environmental correction, that is, the density of the sound source in each area, taking into account multiple factors such as time and environmental influences; these data can be used for real-time monitoring of the sound source distribution at the fire scene, assessing the danger of the environment, formulating emergency response strategies, etc.

[0023] In an optional embodiment, the sound source movement trajectory is marked according to the sound source distribution density data to obtain the sound source movement trajectory data, including: Construct a tracking grid according to the sound source distribution density data to obtain tracking grid data; Perform density gradient analysis on the sound source distribution density data based on the tracking grid data, calculate the sound source density difference between adjacent grid cells, identify the possible movement direction of the sound source, and obtain candidate data of the sound source movement direction; The continuity of the sound source density in each grid unit in the candidate data of the sound source movement direction is counted according to the sliding time window, and the nodes of the sound source trajectory are marked to obtain the sound source trajectory node data; A trajectory prediction model is constructed based on the sound source trajectory node data. The historical spatiotemporal pattern of sound source movement is learned through the long short-term memory network, and the trajectory node position of the next time window is predicted. The actual collected sound source position data is matched and verified to obtain the sound source motion trajectory data.

[0024] It should be noted that, based on the previous three-dimensional space grid model, a grid is constructed to track the movement of the sound source. The purpose is to help analyze the dynamic movement of the sound source in space and capture the changing trend of the sound source through the density data of the grid unit. This grid data is dynamic. As time changes, it can reflect the distribution and density changes of the sound source in space, thereby providing a basis for subsequent analysis. By analyzing the difference in sound source density between adjacent grid units (i.e., "density gradient"), the changing trend of the sound source can be found. By calculating the density difference between adjacent grid units, the moving direction of the sound source can be identified. Places with large density differences may indicate the starting point or end point of the sound source, indicating the moving path of the sound source. Through these density differences, the possible moving direction of the sound source can be preliminarily inferred to form "candidate data", that is, the potential direction of the sound source movement. The sliding time window method is used to observe the density changes of the sound source at multiple time points. This method can help analyze the temporal movement pattern of the sound source and count the continuity of the sound source density in different grid units. By analyzing the sound source density in each time window, it is possible to identify which grids The density of sound sources within a unit changes relatively smoothly or regularly, and which areas have more drastic density changes can help determine which areas are the key nodes of the sound source trajectory; based on the continuity of density, possible "trajectory nodes", that is, important movement nodes of the sound source in space, are marked; these nodes represent the important positions of the sound source at certain moments; the long short-term memory network (LSTM) in deep learning is used to construct a trajectory prediction model for the sound source movement; LSTM can process time series data and remember historical spatiotemporal patterns to predict future movement trajectories; by training on historical sound source trajectory data, the LSTM network can learn the movement patterns of the sound source in space and time (such as speed, direction, etc.); these historical patterns provide a strong basis for subsequent predictions; based on LSTM learning, the model can predict the possible position of the sound source in the next time window, that is, predict the future trajectory nodes of the sound source; based on the model prediction, the actual sound source position data collected can be used to verify the accuracy of the prediction; by comparing the difference between the prediction results and the actual observation data, the effectiveness of the trajectory prediction model can be verified.

[0025] In an optional embodiment, estimating the sound propagation loss based on the sound source motion trajectory data to obtain the sound propagation loss data includes: Construct a spatial model of the sound source movement path based on the sound source motion trajectory data, and mark the key environmental nodes on the trajectory path to obtain associated model data; Perform sound propagation analysis based on the correlation model data, identify the sound propagation mode on the trajectory path, calculate the initial propagation loss coefficient under each propagation mode, and obtain the initial propagation loss coefficient data; Correct the initial propagation loss coefficient by adjusting the loss coefficient under each propagation mode through the sound wave attenuation model associated with smoke concentration and the sound speed correction model associated with temperature to obtain the corrected propagation loss coefficient data; The corrected propagation loss coefficient data is regionally verified based on the sensor, the error between the theoretical loss and the actual loss is calculated, and the loss coefficient is corrected through an adaptive filtering algorithm to obtain the sound propagation loss data.

[0026] It should be noted that the sound source motion trajectory data described above (including the sound source position and movement direction) is used to construct a spatial model. On this trajectory path, key environmental nodes are marked. These nodes represent the key positions where the environment changes during the movement of the sound source; through these nodes and trajectory data, correlation model data related to environmental factors (such as obstacles, climatic conditions, etc.) are constructed; based on the above-constructed correlation model data, sound propagation analysis is performed; here, the sound propagation mode usually includes free field propagation (in the absence of obstacles) and propagation affected by obstacles (such as reflection and refraction of walls, buildings, etc.); these propagation modes will affect the propagation speed and path of sound waves; for each propagation mode, the corresponding initial propagation loss coefficient is calculated; this coefficient represents the degree to which energy is attenuated due to factors such as distance, obstacles, and air medium during the propagation of sound waves; the initial propagation loss coefficient is usually calculated based on factors such as the distance between the sound source and the receiving point, the influence of obstacles, and air humidity; based on these calculations, the initial propagation loss under each propagation mode is obtained. The loss coefficient data lays the foundation for further correction of the loss coefficient; temperature also affects the speed of sound propagation; the higher the temperature, the faster the sound propagation speed and the different the attenuation effect; therefore, the temperature-related sound speed correction model is used to adjust the propagation loss coefficient to make it consistent with the actual temperature conditions; by taking the smoke concentration and temperature factors into consideration, the corrected propagation loss coefficient data is obtained; these data are more in line with the characteristics of sound source propagation under actual environmental conditions; the sound wave propagation in the actual environment is measured by sensors to verify the corrected propagation loss coefficient data; the sensor will collect actual propagation loss data and compare it with the theoretically calculated loss coefficient; the actual measured data is compared with the theoretically calculated loss coefficient to calculate the error between them; this error will reflect the deviation of the model and can also be used as a basis for correction; the adaptive filtering algorithm is used to automatically adjust the propagation loss coefficient according to the size of the error; the adaptive filtering algorithm will dynamically adjust the model parameters so that the calculated result of the loss coefficient is closer to the actual situation; through continuous correction, more accurate sound propagation loss data can be obtained.

[0027] In an optional embodiment, a periodic benchmark simulation assessment of emergency broadcast coverage requirements is performed based on the sound propagation loss data to obtain periodic benchmark data, including: Construct a coverage requirement model based on sound propagation loss data to obtain coverage requirement grid data; According to the sound source motion trajectory data and coverage demand grid data, the frequency of each grid being covered by the sound source in different time periods is counted to obtain coverage demand distribution data; Perform coverage simulation on the coverage demand distribution data and calculate the number of high-priority grids that each device can cover in different time periods to obtain coverage efficiency data; A coverage demand cycle simulation model is constructed based on coverage efficiency data. The total coverage demand and total coverage supply of high-priority grids in each cycle are counted through a sliding time window, and the matching degree of coverage demand and supply is calculated to obtain cycle benchmark data.

[0028] It should be noted that a coverage demand model is constructed based on the sound propagation loss data, with the aim of calculating the coverage demand of each area based on the attenuation characteristics of sound source propagation; coverage demand refers to the sound propagation intensity or degree of sound coverage required in a certain area to meet its usage needs; the coverage demand data of each grid is obtained through the coverage demand model; the coverage demand value of each grid reflects the intensity demand of the area for sound wave coverage; calculate how many times the sound source covers a certain grid at different time points; this statistical result is called coverage demand distribution data, which reflects which areas have higher sound coverage needs and which areas may be less affected by the sound source in a certain period of time; in the coverage range simulation, the coverage range of the device is simulated based on the known coverage demand distribution data; the coverage range of the device may be affected by multiple factors; according to the coverage demand distribution of the grid, some "high priority grids" can be identified, which are areas with relatively high sound coverage needs and need to be met first; by simulating the coverage range, it can be calculated how many high priority grids each device can cover in different time periods, that is, how many high priority grids the device can cover Effectively cover these areas with higher demand; the number of high-priority grids that the device can cover in different time periods is the coverage efficiency data, which reflects the effectiveness and efficiency of the device in meeting coverage needs; build a periodic model to simulate the coverage of the device in multiple periods (such as one day, one week, etc.); within each period, the distribution of sound sources and the coverage capability of the device may change, so dynamic simulation is required; using sliding time window technology, the coverage demand data is segmented and analyzed by time period; the sliding window will continue to move forward within each period to calculate the coverage demand and equipment supply within that period; within each sliding time window, the total coverage demand of high-priority grids (that is, the sound wave coverage intensity required in these areas) and the total coverage supply of equipment (that is, the sound wave coverage capability provided by the equipment within that period) are counted; by calculating the matching degree between coverage demand and equipment supply, the effectiveness of the equipment in meeting demand is evaluated; the matching degree reflects the degree of balance between demand and supply; the resulting periodic benchmark data reflects the overall performance of the device within the period, including key indicators such as the matching degree between demand and supply.

[0029] In an optional embodiment, identifying environmental impact factors of sound propagation loss data based on real-time environmental temperature and humidity data to obtain propagation impact factors includes: Real-time environmental temperature and humidity data collection is performed based on the wireless sensor network to obtain the original temperature and humidity data; The raw temperature and humidity data are cleaned and aligned, sensor noise is removed through Kalman filtering, and the temperature and humidity data collected by the drone are mapped to the ground grid coordinate system to obtain temperature and humidity data; Calculating a correlation index between temperature and humidity and propagation loss based on the temperature and humidity data and the sound propagation loss data to obtain correlation index data; An environmental impact factor model is constructed based on the correlation index data, and the temperature correlation coefficient and humidity correlation coefficient are mapped to impact factor values ​​through the fuzzy logic reasoning system to obtain the propagation impact factor.

[0030] It should be noted that a wireless sensor network (WSN) is a network composed of multiple sensor nodes that collect and transmit data through wireless communication. In this model, sensors in the network collect ambient temperature and humidity data in real time. The data collected by these sensors is the raw temperature and humidity data, which records the changes in ambient temperature and humidity at different locations and times. After the raw temperature and humidity data are collected, there may be some errors or invalid data, and data cleaning is required to remove these outliers. In addition, data from different sensors may not be completely aligned in time or space, and data alignment is required to ensure that all data are compared at the same time and location. In this model, Kalman filtering technology is used to remove noise from sensor data to reduce the impact of sensor errors on temperature and humidity data. When using temperature and humidity data collected by drones, they need to be converted from the coordinate system on the flight trajectory to the ground grid coordinate system. This step is to map the data to the actual geographical area so that the data can be compared and integrated with other data on the ground. Finally, accurate, denoised data is obtained. Temperature and humidity data; based on the temperature and humidity data and sound propagation loss data, a correlation model is established to calculate the correlation index between temperature and humidity and sound propagation loss; this index can be a quantitative measurement that reflects the impact of temperature and humidity changes on sound propagation loss; the correlation index data helps to understand the impact of environmental factors on propagation loss; based on the correlation index data calculated previously, an environmental impact factor model is constructed to evaluate the specific impact of environmental factors such as temperature and humidity on sound propagation loss, and express it using a mathematical model; through this model, the propagation loss under different environmental conditions can be predicted, and a basis for subsequent optimization of propagation strategies can be provided; fuzzy logic is a mathematical method for dealing with uncertainty and ambiguity; through the fuzzy logic reasoning system, the temperature and humidity can be converted into a propagation impact factor based on their correlation coefficient; the fuzzy logic system can quantify the impact of temperature and humidity into impact factor values, which reflect the specific impact of temperature and humidity changes on sound propagation loss; the final propagation impact factor helps to further optimize wireless communication, monitoring and control strategies.

[0031] In an optional embodiment, performing coverage requirement environment correction on the periodic baseline data according to the propagation impact factor to obtain environment correction data includes: An environmental impact map is constructed based on the propagation impact factor data. The fire scene is divided into multi-level dynamic grids, and each grid is labeled with the temperature and humidity impact level to obtain impact grid data. The grids include high-impact grids, medium-impact grids, and low-impact grids. Calculate the coverage demand correction coefficient of each grid under the influence of temperature and humidity based on the periodic benchmark data and the influencing grid data to obtain the coverage supply data; The coverage supply data is corrected twice, and the equipment coverage supply is adjusted through the line of sight obstruction model associated with smoke concentration and the sound propagation deviation model associated with wind speed to obtain environmental correction data.

[0032] It should be noted that, based on the above-mentioned influencing factor data, an environmental impact map is constructed, which is a visual image showing the impact of environmental factors such as temperature, humidity, smoke, etc. in different areas of the fire scene; the grids are marked as different temperature and humidity impact levels: high impact grid: temperature, humidity or other environmental factors have a greater impact on sound propagation, sight, etc., which may affect the performance of fire-fighting equipment or the operation of firefighters; medium impact grid: temperature, humidity or other environmental factors have a moderate impact; low impact grid: temperature, humidity or other environmental factors have a small impact on fire-fighting operations; based on the periodic baseline data and the influencing grid data, the actual demand adjustment for fire-fighting equipment (such as fire extinguishers, sprinkler systems, etc.) under different temperature and humidity conditions is calculated for each grid; the correction factor is an adjustment factor used to correct the equipment coverage demand according to environmental changes; the equipment coverage demand data obtained after adjustment by the correction factor represents the actual supply demand for fire-fighting equipment under different environmental conditions; after the first correction, it can still be There may be some unconsidered environmental factors that affect the coverage effect of the equipment; therefore, secondary corrections are made to further adjust the equipment coverage data to ensure that the equipment coverage can adapt to more complex environmental changes; at the fire scene, smoke concentration is an important environmental factor, which will directly affect the firefighters' line of sight and operational safety; through the line of sight obstruction model, it can be calculated which areas of sight are blocked under different smoke concentrations, affecting the firefighters' ability to judge and act; wind speed will affect the path and distance of sound propagation, especially at the fire scene, when the wind speed is high, it may cause sound propagation deviation; through the sound propagation deviation model, the actual coverage range and effect of sound propagation under different wind speed conditions can be calculated; all the equipment coverage data obtained after these secondary corrections are environmental correction data; these data can accurately reflect the adjustments required for firefighting equipment and operations under different environmental factors such as temperature and humidity, smoke concentration, and wind speed, so as to ensure the optimal equipment configuration and action efficiency.

[0033] In an optional embodiment, an emergency broadcast delivery strategy adjustment model is constructed using environmental correction data according to a policy gradient algorithm, including: Constructing a state space based on the environmental correction data to obtain state space features; Define the action space of the emergency broadcast delivery strategy to obtain action space parameters. The action space parameters include delivery period, delivery power, and delivery direction. Each action combination corresponds to a strategy parameter vector. A multi-objective reward function is designed, combining coverage requirement satisfaction and environmental adaptability to obtain reward function calculation rules; rewards include positive rewards and negative penalties.

[0034] It should be noted that in the optimization problem, the state space refers to the set of all possible states, representing the state of the system at a certain moment; in this context, environmental correction data (such as temperature and humidity, smoke concentration, wind speed and other factors) provide the system with key information that affects the decision-making of emergency broadcasting; the state space converts these data into a high-dimensional space, and each dimension may represent a different level of an environmental factor (such as temperature and humidity levels, smoke concentration, wind speed, etc.); the state space feature is the attribute or variable of the state space, which reflects the characteristics of the environment at a certain moment; for example, the state feature can be the temperature, humidity, smoke concentration and other values ​​of a grid, which have a direct impact on the effectiveness of the broadcasting strategy; by analyzing These state characteristics can better understand the impact of environmental changes on emergency broadcast delivery; in reinforcement learning or optimization models, the action space is the set of all possible actions, representing the different decisions that the system can take; these parameters define the specific strategy of emergency broadcast delivery, which usually includes the following items: delivery cycle refers to the time interval between the delivery of the broadcast signal; for example, the broadcast signal may need to be repeated at a certain time interval, and this time interval is the delivery cycle; delivery power refers to the strength of the signal sent by the broadcast equipment; under different environmental conditions, the broadcast power may need to be adjusted to ensure that the signal can cover the entire area that needs to be broadcast; delivery direction refers to the directionality of the broadcast signal; the broadcast equipment may need to target different The direction of the emergency broadcast delivery is adjusted by the region or grid to cover a wider area at the fire or disaster scene; the strategy parameter vector represents the specific settings of a delivery strategy, which can be continuously adjusted through the optimization algorithm to find the optimal emergency broadcast delivery strategy; in reinforcement learning, the reward function is used to measure the quality of a strategy; designing a suitable reward function is very critical because it directly affects the optimization direction of the learning process; a multi-objective reward function is designed here, which combines two important goals: coverage requirement satisfaction represents the coverage range and effect of the emergency broadcast signal; if the broadcast signal can cover all required areas, the satisfaction is higher and the reward will be correspondingly higher; environmental adaptability represents the emergency broadcast delivery The strategy can effectively adapt and adjust under different environmental conditions (such as high temperature, smoke concentration, wind speed, etc.); if the environmental conditions change, the broadcast delivery strategy can be dynamically adjusted to improve environmental adaptability; specifically, the system will evaluate the strategy based on actual conditions to ensure that the broadcast signal covers the largest range and can adapt to different environmental conditions; when the broadcast delivery strategy can effectively meet the coverage requirements and can adjust under changing environmental conditions, it will give positive rewards to help achieve the goal; when the broadcast delivery strategy fails to meet the coverage requirements or cannot adapt to environmental changes, it will give negative penalties; this means that the strategy is not effective in this state and needs to be adjusted or improved.

[0035] In an optional embodiment, the emergency broadcast delivery strategy adjustment model is constructed using environmental correction data according to a policy gradient algorithm, further comprising: A policy gradient learning model is constructed based on the state space characteristics, action space parameters, and reward function calculation rules. The policy network is initialized and the coverage effects under different policies are simulated using the Monte Carlo method to obtain the initial policy network parameters. The initial policy network parameters are modified, the state space features are updated through the incremental learning mechanism, and the policy adaptability is maintained through the exploration-exploitation balance to obtain a policy adjustment model.

[0036] It should be noted that the core idea of ​​the policy gradient learning method is to maximize the expected reward of taking a certain action from the current state by fine-tuning the policy parameters (i.e., the weights in the policy network); in this context, we need to design a model to generate an emergency broadcast delivery strategy so that the maximum reward can be obtained under various environmental conditions; the policy network is a neural network model used to generate broadcast strategies (i.e., actions) based on state space features; the initialization parameters of the network are usually random, and the strategy may not be perfect at the beginning; the Monte Carlo method is a numerical calculation method based on random sampling; in reinforcement learning, it is often used to estimate the expected reward of a strategy; by simulating different broadcast strategies, calculating the coverage effect under each strategy, and then estimating the expected reward of the strategy through these simulation results; this process helps to obtain the initial policy network parameters; incremental learning is a method of gradually updating the learning model; in this step, the parameters of the policy network are gradually updated according to the simulation and actual feedback Positive; through incremental learning, the policy network can continuously adjust and optimize according to the newly acquired data; each update will make the policy network more adaptable to the current environment; as the learning process proceeds, the state space features will also be dynamically updated as the environment changes; exploration-exploitation balance is a core issue in reinforcement learning; exploration refers to trying different actions and strategies to discover potential better solutions, while exploitation refers to taking the best action under a known strategy to obtain the maximum reward; in the initial stage, the system may explore more and try a variety of different strategies; as learning deepens, the system will make greater use of existing knowledge to select the most effective strategy; maintaining this balance helps prevent falling into local optimal solutions and ensures that the strategy remains highly adaptable in the face of a changing environment; through continuous correction and optimization, a policy adjustment model is formed, which can analyze environmental changes in real time and continuously adjust the broadcast cycle, power and direction according to the current state and feedback to achieve the best broadcast effect.

[0037] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. Fire emergency broadcast data information collection system, characterized by: include: Sound source localization unit, used to obtain the original data of the fire scene environmental sound; Performing sound source localization analysis on the original data to obtain sound source distribution data; a sound source analysis unit, configured to perform sound source area analysis based on the sound source distribution data to obtain sound source distribution density data; and to mark the sound source movement trajectory based on the sound source distribution density data to obtain sound source movement trajectory data; a loss estimation unit, configured to estimate the sound propagation loss based on the sound source motion trajectory data to obtain sound propagation loss data; Performing a periodic benchmark simulation assessment of emergency broadcast coverage requirements based on the sound propagation loss data to obtain periodic benchmark data; An environmental identification unit is configured to obtain real-time environmental temperature and humidity data at the fire scene; identify environmental impact factors on the sound propagation loss data based on the real-time environmental temperature and humidity data to obtain propagation impact factors; and perform coverage requirement environmental correction on the periodic baseline data based on the propagation impact factors to obtain environmental correction data; The data acquisition unit is used to construct an emergency broadcast delivery strategy adjustment model through the environmental correction data according to the policy gradient algorithm, and send the emergency broadcast delivery strategy adjustment model to the fire command cloud platform to perform fire emergency broadcast data information collection and analysis.

2. The fire emergency broadcast data information collection system according to claim 1, characterized in that: Performing sound source localization analysis on the original data to obtain sound source distribution data, including: Performing multi-channel signal synchronous acquisition on the raw data using a microphone array device to obtain a synchronized sound wave signal; performing consistency calibration on the synchronized sound wave signal and extracting phase features; wherein the characteristic dimension of the phase feature is determined by the number of array elements of the microphone array and the phase dimension of each array element, thereby obtaining multi-channel phase feature data; Performing feature sampling to a target dimension based on the multi-channel phase feature data to obtain a first feature; performing feature extraction based on physical layout attribute information of the microphone array to obtain a second feature of the target dimension; wherein the attribute information includes the array arrangement and the number of array elements, and the target dimension is a unified feature dimension; The first feature and the second feature are combined to obtain a target feature, which is input into a pre-trained sound source localization model; the sound source localization model is trained based on a sample audio set, and the spatial coordinates of the sound source are predicted using the target feature to obtain the sound source position; Performing time-frequency analysis on the sound source position, constructing a time-frequency-space correlation matrix of the sound source based on short-time Fourier transform, and obtaining time-frequency-space distribution data of the sound source; marking the dynamic trajectory of the sound source based on the time-frequency-space distribution data of the sound source, and obtaining trajectory data; Ambient noise analysis is performed based on the trajectory data, and background noise unrelated to the sound source trajectory is removed using an adaptive noise cancellation algorithm to obtain corrected trajectory data. Density clustering of the spatial distribution of sound sources is performed based on the corrected trajectory data to identify spatial aggregation areas of different sound sources to obtain sound source distribution data.

3. The fire emergency broadcast data information collection system according to claim 2, characterized in that: Performing sound source area analysis based on the sound source distribution data to obtain sound source distribution density data includes: Constructing a three-dimensional spatial grid model based on the sound source distribution data, dividing the fire scene into multi-scale adaptive grid units, and obtaining spatial grid data; Performing sound source position mapping based on the spatial grid data, matching the spatial coordinates of each sound source to a corresponding grid unit, and counting the number of sound sources in each grid unit to obtain sound source basic density data; Performing time correction on the basic density data of the sound source according to the trajectory data, counting the residence time and movement frequency of the sound source in each grid unit through a sliding time window, and calculating the time-weighted density to obtain the dynamic density data of the sound source; The sound source dynamic density data is corrected for environmental impacts, the density calculation range is adjusted using a sound propagation attenuation model associated with smoke concentration, and the density calculation time interval is corrected using a sound propagation speed model associated with temperature to obtain sound source distribution density data.

4. The fire emergency broadcast data information collection system according to claim 3, characterized in that: Marking the sound source movement trajectory according to the sound source distribution density data to obtain sound source movement trajectory data includes: Constructing a tracking grid according to the sound source distribution density data to obtain tracking grid data; performing density gradient analysis on the sound source distribution density data according to the tracking grid data, calculating the sound source density difference between adjacent grid cells, identifying the possible movement direction of the sound source, and obtaining candidate data of the sound source movement direction; Counting the continuity of the sound source density in each grid unit in the candidate data of the sound source movement direction according to the sliding time window, marking the nodes of the sound source trajectory to obtain sound source trajectory node data; A trajectory prediction model is constructed based on the sound source trajectory node data. The historical spatiotemporal pattern of sound source movement is learned through a long short-term memory network, the trajectory node position of the next time window is predicted, and the actually collected sound source position data is matched and verified to obtain the sound source motion trajectory data.

5. The fire emergency broadcast data information collection system according to claim 4, characterized in that: Estimating sound propagation loss based on the sound source motion trajectory data to obtain sound propagation loss data includes: Constructing a spatial model of the sound source movement path based on the sound source motion trajectory data, and marking key environmental nodes on the trajectory path to obtain associated model data; Performing sound propagation analysis based on the correlation model data, identifying the sound propagation mode on the trajectory path, calculating the initial propagation loss coefficient under each propagation mode, and obtaining initial propagation loss coefficient data; Correcting the initial propagation loss coefficient by adjusting the loss coefficient in each propagation mode using a sound wave attenuation model associated with smoke concentration and a sound velocity correction model associated with temperature to obtain corrected propagation loss coefficient data; The corrected propagation loss coefficient data is regionally verified according to the sensor, the error between the theoretical loss and the actual loss is calculated, and the loss coefficient is corrected by an adaptive filtering algorithm to obtain the sound propagation loss data.

6. The fire emergency broadcast data information collection system according to claim 5, characterized in that: Conduct a periodic benchmark simulation assessment of emergency broadcast coverage requirements based on the sound propagation loss data to obtain periodic benchmark data, including: Constructing a coverage requirement model according to the sound propagation loss data to obtain coverage requirement grid data; According to the sound source motion trajectory data and the coverage requirement grid data, the frequency of each grid being covered by the sound source in different time periods is counted to obtain coverage requirement distribution data; Performing coverage simulation on the coverage demand distribution data, calculating the number of high-priority grids that each device can cover in different time periods, to obtain coverage efficiency data; A coverage demand cycle simulation model is constructed based on the coverage efficiency data. The total coverage demand and total coverage supply of high-priority grids in each cycle are counted through a sliding time window, and the matching degree of coverage demand and supply is calculated to obtain cycle benchmark data.

7. The fire emergency broadcast data information collection system according to claim 6, characterized in that: Identifying environmental impact factors on the sound propagation loss data based on the real-time environmental temperature and humidity data to obtain propagation impact factors includes: Real-time environmental temperature and humidity data collection is performed based on the wireless sensor network to obtain the original temperature and humidity data; The raw temperature and humidity data are cleaned and aligned, sensor noise is removed through Kalman filtering, and the temperature and humidity data collected by the drone are mapped to a ground grid coordinate system to obtain temperature and humidity data; Calculating a correlation index between temperature and humidity and propagation loss based on the temperature and humidity data and the sound propagation loss data to obtain correlation index data; An environmental impact factor model is constructed based on the correlation index data, and the temperature correlation coefficient and the humidity correlation coefficient are mapped into impact factor values ​​through a fuzzy logic reasoning system to obtain a propagation impact factor.

8. The fire emergency broadcast data information collection system according to claim 7, characterized in that: Correcting the periodic baseline data for coverage requirements according to the propagation impact factor to obtain environmental correction data includes: An environmental impact map is constructed based on the propagation impact factor data, and the fire scene is divided into multi-level dynamic grids. Each grid is labeled with a temperature and humidity impact level to obtain impact grid data; wherein the grids include high-impact grids, medium-impact grids, and low-impact grids; Calculating the coverage demand correction coefficient of each grid under the influence of temperature and humidity based on the periodic benchmark data and the affected grid data to obtain coverage supply data; The coverage supply data is corrected twice, and the equipment coverage supply is adjusted through a line of sight obstruction model associated with smoke concentration and a sound propagation deviation model associated with wind speed to obtain environmental correction data.

9. The fire emergency broadcast data information collection system according to claim 8, characterized in that: Constructing an emergency broadcast delivery strategy adjustment model using the environmental correction data according to a policy gradient algorithm, including: constructing a state space according to the environmental correction data to obtain a state space feature; Defining an action space of the emergency broadcast delivery strategy to obtain action space parameters; wherein the action space parameters include a delivery period, a delivery power, and a delivery direction, and each action combination corresponds to a strategy parameter vector; A multi-objective reward function is designed, combining coverage requirement satisfaction and environmental adaptability to obtain the reward function calculation rules.

10. The fire emergency broadcast data information collection system according to claim 9, characterized in that: Constructing an emergency broadcast delivery strategy adjustment model using the environmental correction data according to a policy gradient algorithm, further comprising: Constructing a policy gradient learning model based on the state space characteristics, the action space parameters, and the reward function calculation rule, initializing the policy network, and simulating coverage effects under different strategies using the Monte Carlo method to obtain initial policy network parameters; The initial policy network parameters are modified, the state space features are updated through an incremental learning mechanism, and the policy adaptability is maintained through exploration-exploitation balance to obtain a policy adjustment model.

Images