A fire escape route intelligent recommendation method and system

Abstract

The application relates to the technical field of fire safety, and discloses a fire escape route intelligent recommendation method and system, which integrates a building information model, real-time sensor network data and video monitoring data, constructs a virtual digital twin environment which is synchronously updated with a physical fire scene, pre-processes multi-source signals, identifies a fire position, a smoke concentration distribution and a personnel gathering condition through a convolutional neural network based on the pre-processed multi-source signals, generates a real-time multi-dimensional environment risk map and a personnel distribution map, carries out parallel deduction in the digital twin environment, plans an escape route by adopting an A* algorithm, and obtains an optimal personalized escape path, continuously monitors changes in the environment and the personnel state, dynamically plans the escape route when new congestion occurs, and pushes the escape route in real time, and the application improves the safety and efficiency of personnel evacuation under a fire scene.

Description

Technical Field

[0001] This invention relates to the field of fire safety technology, specifically to a method and system for intelligently recommending fire escape routes. Background Technology

[0002] In fire accidents, the dynamic adaptability of escape routes is directly related to the safety of people's lives. Existing escape route recommendation technologies have significant limitations. They mostly rely on preset fixed routes and do not take into account the dynamic characteristics of fire spread and smoke diffusion. As a result, routes are often blocked by the fire. The data processing dimension is also limited, only integrating a small number of sensor signals and failing to effectively integrate key data such as video surveillance and personnel status, leading to one-sided environmental perception. In addition, the lack of a virtual mapping environment synchronized with the physical scene makes it difficult to predict and extrapolate routes in advance. Therefore, there is an urgent need for an intelligent method for recommending fire escape routes. Summary of the Invention

[0003] The purpose of this invention is to solve the above-mentioned problems by designing an intelligent method and system for recommending fire escape routes.

[0004] The first aspect of this invention provides a method for intelligently recommending fire escape routes, the method comprising the following steps: By integrating building information modeling, real-time sensor network data, and video surveillance data, a virtual digital twin environment that is updated synchronously with the physical fire scene is constructed; The system integrates signals from temperature sensors, smoke sensors, carbon monoxide concentration sensors, infrared human body detectors, and video surveillance cameras in real time to obtain multi-source signals, and then preprocesses these multi-source signals. Based on the preprocessed multi-source signals, the location of open flames, smoke concentration distribution and personnel gathering are identified through convolutional neural networks, generating real-time, multi-dimensional environmental risk maps and personnel distribution maps; Parallel simulations are performed in a digital twin environment, and the A* algorithm is used to plan escape routes to obtain the optimal personalized escape path; Continuously monitor changes in the environment and personnel status, and when new congestion occurs, dynamically plan escape routes and push them out in real time.

[0005] Optionally, in a first implementation of the first aspect of the present invention, the integration of building information model, real-time sensor network data, and video surveillance data to construct a virtual digital twin environment that is synchronously updated with the physical fire scene includes: The BIM model is preprocessed in a structured manner to extract spatial geometric and attribute information, including the layout of rooms within the building, the size of passageways, the material of walls, and the location of safety exits. The BIM model is then transformed into a standardized spatial data model. Establish the coordinate correspondence between physical space and virtual space, and match the installation positions of various sensors and cameras in the building to the corresponding nodes of the BIM model; Real-time sensor network data and video surveillance data are streamed and aligned according to the collection time to construct a virtual digital twin environment that is synchronously updated with the physical fire scene.

[0006] Optionally, in a second implementation of the first aspect of the present invention, the real-time fusion of signals from a temperature sensor, a smoke sensor, a carbon monoxide concentration sensor, an infrared human body detector, and a video surveillance camera to obtain multi-source signals, and the preprocessing of the multi-source signals, includes: For each sensor data in the multi-source signal, initialize the filtering parameters, and predict the current state value and the corresponding error covariance based on the state estimate and sensor data of the previous time step. The actual sensor readings at the current moment are compared with the prediction results, the Kalman gain is calculated, and the prediction state is corrected by the Kalman gain to obtain the denoised sensor data. Gradient calculation is performed on blurred video frames from multi-source signals to extract image edge features. A blur kernel estimation model is constructed based on the edge features, and the blur kernel parameters of the current image are obtained by iterative optimization. The estimated blur kernel is used to perform deconvolution operation on the video frame to obtain a clear image. The clear image is divided into multiple non-overlapping sub-blocks, and the gray-level histogram of each sub-block is calculated. The histogram of each sub-block is equalized, and the boundary of adjacent sub-blocks is smoothed by bilinear interpolation algorithm. By integrating the denoised sensor data and the processed video images, a preprocessed multi-source signal is obtained.

[0007] Optionally, in a third implementation of the first aspect of the present invention, the step of identifying the location of open flames, smoke concentration distribution, and personnel gathering based on preprocessed multi-source signals using a convolutional neural network to generate a real-time, multi-dimensional environmental risk map and personnel distribution map includes: The sensor features and visual features in the preprocessed multi-source signals are unified in dimensionality to form a fusion feature tensor, which is then input into a convolutional neural network. By performing sliding convolution operation on the fused feature tensor using a 3×3 convolution kernel, low-level features including edges and textures from visual and sensory features are extracted. Then, the feature map is downsampled through a max pooling layer to generate a high-level feature map containing semantic information about fire and personnel. The high-level feature map is input into the region suggestion branch, and multiple candidate regions are generated through a sliding window. After cropping and normalizing the features of each candidate region, the data is fed into the classification subnetwork to determine whether there is an open flame in the region. The coordinate position of the candidate region is corrected by the regression subnetwork, and the region with a confidence level higher than the preset threshold is selected as the final open flame location identification result. The high-level feature map is input into the semantic segmentation branch. The spatial resolution of the high-level feature map is restored through upsampling operation. The low-level features are then fused together, and the smoke concentration level is classified for each pixel through the softmax activation function to obtain the smoke concentration distribution data. The anchor frame mechanism is used to match human targets of different sizes on high-level feature maps. The classification layer determines whether the anchor frame contains a human body, and the regression layer adjusts the anchor frame coordinates to obtain the human body detection box. The number and distribution density of human bodies in each area of ​​the building are counted to form data on human gathering. The results of open flame location identification, smoke concentration distribution data, and personnel gathering data are integrated and aligned with the preset building space structure data to output a real-time multi-dimensional environmental risk map and personnel distribution map.

[0008] Optionally, in the fourth implementation of the first aspect of the present invention, the step of performing parallel simulation in a digital twin environment and using the A* algorithm to plan the escape route to obtain the optimal personalized escape path includes: Based on the generated environmental risk map and personnel distribution map, a dynamic path planning graph model is constructed in the digital twin environment. The dynamic path planning graph model uses the passage nodes, room nodes, and safety exit nodes in the building as vertices of the graph and the connecting passages between nodes as edges. Initiate a parallel simulation mechanism in a digital twin environment to simultaneously simulate the passage of multiple potential escape routes; The A* algorithm is used to calculate the comprehensive cost of each path, and the path with the lowest comprehensive cost is selected as the optimal personalized escape path. The environmental risk weight and personnel adaptation factor are incorporated into the heuristic function of the A* algorithm.

[0009] Optionally, in the fifth implementation of the first aspect of the present invention, the step of calculating the comprehensive cost of each path using the A* algorithm and selecting the path with the lowest comprehensive cost as the optimal personalized escape path includes: From the dynamic path planning graph model of the digital twin environment, obtain the node sequence of each candidate path and the travel distance between nodes; Obtain node risk level data from the environmental risk map, as well as personnel characteristic information associated with the personnel distribution map; The node risk level data is converted into quantitative weight values ​​of 0.5, 1, 2, and 3, and the comprehensive risk weight of the path is calculated. Based on the personnel characteristic information of the path, the personnel matching factor is determined. The actual cost of each candidate path is calculated by incorporating comprehensive risk weights and personnel suitability factors. Then, the actual cost is added to the heuristic cost of each node to obtain the comprehensive cost of each candidate path. Sort all candidate paths by their combined cost in ascending order and extract the path with the lowest cost as the initial optimal path. Check if there are any unmarked temporary obstacles on the initial optimal path. If so, remove the current path and select the second lowest cost path. Iterate to obtain the optimal personalized escape path.

[0010] Optionally, in the sixth implementation of the first aspect of the present invention, the continuous monitoring of changes in the environment and personnel status, and the dynamic planning of escape routes when new congestion occurs, and the real-time push notifications, includes: A sliding window algorithm is used to continuously collect environmental sensor data and personnel positioning data, and the data differences within adjacent time windows are compared. Set a congestion judgment threshold. When the population density in a certain passage area exceeds the preset congestion judgment threshold, it is determined that new congestion has occurred. When congestion is detected, the A* algorithm and parallel simulation process are repeated based on the updated environmental risk map and personnel distribution map to generate a new optimal escape route. The updated escape route information is marked as an emergency message and converted into a format for different terminals for real-time push.

[0011] A second aspect of the present invention provides an intelligent fire escape route recommendation system, the system comprising: The building module is used to integrate building information model, real-time sensor network data, and video surveillance data to build a virtual digital twin environment that is synchronized with the physical fire scene. The preprocessing module is used to fuse signals from temperature sensors, smoke sensors, carbon monoxide concentration sensors, infrared human body detectors, and video surveillance cameras in real time to obtain multi-source signals and preprocess the multi-source signals. The identification module is used to identify the location of open flames, smoke concentration distribution, and personnel gathering based on preprocessed multi-source signals through a convolutional neural network, generating real-time, multi-dimensional environmental risk maps and personnel distribution maps; The planning module is used to perform parallel simulations in a digital twin environment, using the A* algorithm to plan escape routes and obtain the optimal personalized escape path; The real-time push module is used to continuously monitor changes in the environment and personnel status. When new congestion occurs, it dynamically plans escape routes and pushes them in real time.

[0012] A third aspect of the present invention provides a fire escape route intelligent recommendation device, the fire escape route intelligent recommendation device including a memory and at least one processor, the memory storing instructions; the at least one processor calling the instructions in the memory to cause the fire escape route intelligent recommendation device to perform the various steps of the fire escape route intelligent recommendation method as described in any of the preceding claims.

[0013] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed by a processor, implement the steps of the intelligent fire escape route recommendation method as described in any of the preceding claims.

[0014] The technical solution provided by this invention integrates building information modeling, real-time sensor network data, and video surveillance data to construct a virtual digital twin environment that is synchronously updated with the physical fire scene. It fuses signals from temperature sensors, smoke sensors, carbon monoxide concentration sensors, infrared human body detectors, and video surveillance cameras in real time to obtain multi-source signals, which are then preprocessed. Based on the preprocessed multi-source signals, a convolutional neural network is used to identify the location of open flames, smoke concentration distribution, and personnel gathering, generating a real-time, multi-dimensional environmental risk map and personnel distribution map. Parallel simulations are performed within the digital twin environment, and the A* algorithm is used to plan escape routes to obtain the optimal personalized escape route. The invention provides a digital twin environment that enables real-time synchronization between the physical site and the virtual model, providing precise scenario support for route planning. Multi-source signal fusion and preprocessing enhance data reliability, while convolutional neural networks enable more accurate identification of fire conditions and personnel status, and more timely generation of environmental and personnel maps. An improved A* algorithm incorporates personalized and risk factors, making the planned escape routes more adaptable to different personnel needs. The dynamic monitoring mechanism can quickly respond to emergencies such as congestion, updating and pushing routes in real time, significantly reducing escape risks and improving the safety and efficiency of personnel evacuation in fire scenarios. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0016] Figure 1 A flowchart of the intelligent fire escape route recommendation method provided in this embodiment of the invention; Figure 2 This is a schematic diagram of the structure of the intelligent fire escape route recommendation system provided in an embodiment of the present invention; Figure 3A schematic diagram of the structure of the intelligent fire escape route recommendation device provided in an embodiment of the present invention. Detailed Implementation

[0017] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0018] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 A flowchart of the intelligent fire escape route recommendation method provided in this embodiment of the invention, the method specifically includes the following steps: Step 101: Integrate building information model, real-time sensor network data, and video surveillance data to construct a virtual digital twin environment that is synchronously updated with the physical fire scene; In this embodiment, the BIM model undergoes structured preprocessing to extract spatial geometry and attribute information, including room layout, passageway dimensions, wall materials, and safety exit locations within the building. The BIM model is then transformed into a standardized spatial data model. A coordinate correspondence between the physical and virtual spaces is established, and the installation positions of various sensors and cameras within the building are matched to the corresponding nodes in the BIM model. Real-time sensor network data and video surveillance data are streamed and received, aligning data from different sources according to their acquisition time to construct a virtual digital twin environment that is synchronously updated with the physical fire scene.

[0019] Step 102: In real time, fuse signals from temperature sensor, smoke sensor, carbon monoxide concentration sensor, infrared human body detector, and video surveillance camera to obtain multi-source signals, and preprocess the multi-source signals. In this embodiment, for each sensor data in the multi-source signal, filtering parameters are initialized, and the current state value and corresponding error covariance are predicted based on the state estimate and sensor data from the previous moment. The actual sensor acquisition value at the current moment is compared with the prediction result, and the Kalman gain is calculated. The predicted state is corrected using the Kalman gain to obtain the denoised sensor data. Gradient calculation is performed on the blurred video frames in the multi-source signal to extract image edge features. A blur kernel estimation model is constructed based on the edge features, and the blur kernel parameters of the current image are obtained through iterative optimization. The estimated blur kernel is used to perform deconvolution on the video frame to obtain a clear image. The clear image is divided into multiple non-overlapping sub-blocks, and the grayscale histogram of each sub-block is calculated. The histogram of each sub-block is equalized, and the boundaries of adjacent sub-blocks are smoothed using a bilinear interpolation algorithm. The denoised sensor data and the processed video image are integrated to obtain the preprocessed multi-source signal.

[0020] Step 103: Based on the preprocessed multi-source signals, the location of open flames, smoke concentration distribution, and personnel gathering are identified through a convolutional neural network to generate a real-time, multi-dimensional environmental risk map and personnel distribution map. In this embodiment, the sensor and visual features in the preprocessed multi-source signals are unified in dimensionality to form a fused feature tensor, which is then input into a convolutional neural network. A sliding convolution operation is performed on the fused feature tensor using a 3×3 convolutional kernel to extract low-level features, including edges and textures, from the visual and sensor features. The feature map is then downsampled using a max pooling layer to generate a high-level feature map containing semantic information about fire and personnel. This high-level feature map is input into a region suggestion branch, which generates multiple candidate regions through a sliding window. After cropping and normalizing the features of each candidate region, they are fed into a classification subnetwork to determine whether an open flame exists within the region. A regression subnetwork corrects the coordinates of the candidate regions, and regions with confidence scores higher than a preset threshold are selected as the most likely candidates. The final result of open flame location identification is obtained. High-level feature maps are input into the semantic segmentation branch, and upsampling operations are used to restore the spatial resolution of the high-level feature maps. Low-level features are then fused, and each pixel is classified for smoke concentration level using the softmax activation function to obtain smoke concentration distribution data. Anchor boxes are used to match human targets of different sizes on the high-level feature maps. A classification layer determines whether a human is contained within the anchor box, and a regression layer adjusts the anchor box coordinates to obtain the human detection box. The number and distribution density of human figures in each area of ​​the building are statistically analyzed to form data on human gathering. The open flame location identification results, smoke concentration distribution data, and human gathering data are fused and aligned with the preset building spatial structure data to output a real-time multi-dimensional environmental risk map and a human distribution map.

[0021] Step 104: Perform parallel simulation in the digital twin environment, use the A* algorithm to plan the escape route, and obtain the optimal personalized escape path; In this embodiment, a dynamic path planning graph model is constructed in the digital twin environment based on the generated environmental risk map and personnel distribution map. The dynamic path planning graph model uses the passage nodes, room nodes, and safety exit nodes in the building as vertices of the graph, and the connecting passages between nodes as edges. A parallel inference mechanism is initiated in the digital twin environment to simulate the passage process of multiple potential escape routes simultaneously. The comprehensive cost of each path is calculated using the A* algorithm, and the path with the lowest comprehensive cost is selected as the optimal personalized escape route. The environmental risk weight and personnel adaptation factor are incorporated into the heuristic function of the A* algorithm.

[0022] In this embodiment, the node sequence and travel distance between nodes of each candidate path are obtained from the dynamic path planning graph model of the digital twin environment. At the same time, node risk level data from the environmental risk map and characteristic information such as age, heart rate, and mobility of the personnel suitable for the path associated with the personnel distribution map are called and organized into a structured algorithm input dataset. The risk level mapping algorithm is used to convert the node risk level, such as low, medium, high, and extremely high, into quantitative weight values ​​of 0.5, 1, 2, and 3. For multiple consecutive high-risk nodes in the path, the comprehensive risk weight of the road segment is calculated through the weight superposition rule. That is, the comprehensive weight of n consecutive nodes of the same risk level is n times the weight of a single node, ensuring that high-risk areas occupy a higher proportion in the cost calculation. Based on personnel characteristics, a matching factor calculation rule is constructed. A healthy adult is used as the baseline matching factor of 1. If the person's age is ≥65 years or ≤12 years, the matching factor is increased to 1.3; if the heart rate is ≥120 beats / min or the movement speed is ≤0.8m / s, the matching factor is further increased by 0.2. Finally, the personalized matching factor for each person is obtained through feature aggregation. The larger the matching factor value, the higher the person's requirements for the safety and smoothness of the path. Based on the traditional distance estimation cost, environmental risk weights and personnel matching factors are incorporated. The new heuristic function is: Heuristic Cost = Base Distance Cost from Node to Safety Exit × Comprehensive Environmental Risk Weight × Personnel Matching Factor. The base distance cost is calculated using Euclidean distance. This formula transforms safety risks and personnel needs into quantifiable heuristic costs. The improved A* algorithm's cost calculation module is invoked to first calculate the actual cost of each candidate path, which is the product of the travel distance from the starting point to each node and the real-time congestion coefficient. The congestion coefficient is determined by the real-time population density and is set to 1.3 during congestion. Then, the actual cost is added to the heuristic cost of each node to obtain the complete comprehensive cost of each candidate path. The cost sorting algorithm is then activated to sort the comprehensive costs of all candidate paths in ascending order, and the path with the lowest cost is extracted as the initial optimal path. At the same time, it is checked whether there are unmarked temporary obstacles or sudden congestion on the path. If so, the path is eliminated and the second lowest cost path is selected. Finally, the optimal personalized escape path that combines low cost and high feasibility is determined.

[0023] Step 105: Continuously monitor changes in the environment and personnel status. When new congestion occurs, dynamically plan escape routes and push the information in real time.

[0024] In this embodiment, a sliding window algorithm is used to continuously collect environmental sensor data and personnel positioning data, and compare the data differences within adjacent time windows; a congestion judgment threshold is set, and when the personnel density in a certain passage area exceeds the preset congestion judgment threshold, it is determined that new congestion has occurred; when congestion is detected, based on the updated environmental risk map and personnel distribution map, the A* algorithm and parallel inference process are repeated to generate a new optimal escape route; the updated escape route information is marked as an emergency message and converted into the format of different terminals for real-time push.

[0025] Please see Figure 2 A schematic diagram of the structure of the intelligent fire escape route recommendation system provided in this embodiment of the invention. The system includes: The building module is used to integrate building information model, real-time sensor network data, and video surveillance data to build a virtual digital twin environment that is synchronized with the physical fire scene. The preprocessing module is used to fuse signals from temperature sensors, smoke sensors, carbon monoxide concentration sensors, infrared human body detectors, and video surveillance cameras in real time to obtain multi-source signals and preprocess the multi-source signals. The identification module is used to identify the location of open flames, smoke concentration distribution, and personnel gathering based on preprocessed multi-source signals through a convolutional neural network, generating real-time, multi-dimensional environmental risk maps and personnel distribution maps; The planning module is used to perform parallel simulations in a digital twin environment, using the A* algorithm to plan escape routes and obtain the optimal personalized escape path; The real-time push module is used to continuously monitor changes in the environment and personnel status. When new congestion occurs, it dynamically plans escape routes and pushes them in real time.

[0026] Figure 3This is a schematic diagram of the structure of an intelligent fire escape route recommendation device 300 provided in an embodiment of the present invention. The intelligent fire escape route recommendation device 300 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 310 (e.g., one or more processors) and a memory 320, and one or more storage media 330 (e.g., one or more mass storage devices) for storing application programs 333 or data 332. The memory 320 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the intelligent fire escape route recommendation device 300. Furthermore, the processor 310 may be configured to communicate with the storage media 330 and execute the series of instruction operations in the storage media 330 on the intelligent fire escape route recommendation device 300 to implement the method provided in the above embodiment.

[0027] The intelligent fire escape route recommendation device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3 The fire escape route intelligent recommendation device structure shown does not constitute a limitation on the computer device provided by the present invention. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0028] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform each step of the intelligent fire escape route recommendation method provided in the above embodiments.

[0029] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described equipment or apparatus / unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0030] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0031] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A method for intelligently recommending fire escape routes, characterized in that, The method includes the following steps: By integrating building information modeling, real-time sensor network data, and video surveillance data, a virtual digital twin environment that is updated synchronously with the physical fire scene is constructed; The system integrates signals from temperature sensors, smoke sensors, carbon monoxide concentration sensors, infrared human body detectors, and video surveillance cameras in real time to obtain multi-source signals, and then preprocesses these multi-source signals. Based on the preprocessed multi-source signals, the location of open flames, smoke concentration distribution and personnel gathering are identified by convolutional neural networks, generating real-time, multi-dimensional environmental risk maps and personnel distribution maps; Parallel simulations are performed in a digital twin environment, and an improved A* algorithm is used to plan escape routes to obtain the optimal personalized escape path. Continuously monitor changes in the environment and personnel status, and when new congestion occurs, dynamically plan escape routes and push them in real time; The process of parallel simulation in a digital twin environment, using an improved A* algorithm to plan escape routes and obtain the optimal personalized escape path, includes: Based on the generated environmental risk map and personnel distribution map, a dynamic path planning graph model is constructed in the digital twin environment. The dynamic path planning graph model uses the passage nodes, room nodes, and safety exit nodes in the building as vertices of the graph and the connecting passages between nodes as edges. Initiate a parallel simulation mechanism in a digital twin environment to simultaneously simulate the passage of multiple potential escape routes; The improved A* algorithm is used to calculate the comprehensive cost of each path, and the path with the lowest comprehensive cost is selected as the optimal personalized escape path, including: From the dynamic path planning graph model of the digital twin environment, obtain the node sequence of each path and the travel distance between nodes; Obtain risk level data for nodes in the environmental risk map, as well as personnel characteristic information associated with the personnel distribution map; The risk level data of nodes is converted into quantitative weight values, and the weights of multiple consecutive nodes with the same risk level in the path are superimposed to obtain the comprehensive risk weight of the path. Based on path-related personnel characteristic information, personnel adaptation factors are determined, which are related to personnel's age, heart rate, and mobility. The new heuristic function is: Heuristic cost = Basic distance cost from the current node to the safe exit node × Comprehensive risk weight of the path × Personnel suitability factor; The improved A* algorithm's cost calculation module is invoked to calculate the actual cost of each path by multiplying the travel distance from the starting point to each node by the real-time congestion coefficient, which is determined by the real-time population density. The actual cost is then added to the heuristic cost of each node to obtain the comprehensive cost of each path. A cost ranking algorithm is then initiated to rank the comprehensive costs of all paths, extracting the path with the lowest cost as the initial optimal path. Simultaneously, it is checked whether the path contains unmarked temporary obstacles or sudden congestion. If so, the path is removed, and the next lowest cost path is selected. This iterative process yields the optimal personalized escape path.

2. The intelligent fire escape route recommendation method as described in claim 1, characterized in that, The integrated building information model, real-time sensor network data, and video surveillance data construct a virtual digital twin environment that is synchronously updated with the physical fire scene, including: The BIM model is preprocessed in a structured manner to extract spatial geometric and attribute information, including the layout of rooms inside the building, the size of passageways, the material of walls, and the location of safety exits. The BIM model is then converted into a standardized spatial data model. Establish the coordinate correspondence between physical space and virtual space, and match the installation positions of various sensors and cameras in the building to the corresponding nodes of the BIM model; Real-time sensor network data and video surveillance data are streamed and aligned according to the collection time to construct a virtual digital twin environment that is synchronously updated with the physical fire scene.

3. The intelligent fire escape route recommendation method as described in claim 1, characterized in that, The real-time fusion of signals from temperature sensors, smoke sensors, carbon monoxide concentration sensors, infrared human body detectors, and video surveillance cameras yields multi-source signals, which are then preprocessed, including: For each sensor data in the multi-source signal, initialize the filtering parameters, and predict the current state value and the corresponding error covariance based on the state estimate and sensor data of the previous time step. The actual sensor readings at the current moment are compared with the prediction results, the Kalman gain is calculated, and the prediction state is corrected by the Kalman gain to obtain the denoised sensor data. Gradient calculation is performed on blurred video frames from multi-source signals to extract image edge features. A blur kernel estimation model is constructed based on the edge features, and the blur kernel parameters of the current image are obtained by iterative optimization. The estimated blur kernel is used to perform deconvolution operation on the video frame to obtain a clear image. The clear image is divided into multiple non-overlapping sub-blocks, and the gray-level histogram of each sub-block is calculated. The histogram of each sub-block is equalized, and the boundary of adjacent sub-blocks is smoothed by bilinear interpolation algorithm. By integrating the denoised sensor data and the processed video images, a preprocessed multi-source signal is obtained.

4. The intelligent fire escape route recommendation method as described in claim 1, characterized in that, Based on the preprocessed multi-source signals, a convolutional neural network is used to identify the location of open flames, smoke concentration distribution, and personnel gathering, generating a real-time, multi-dimensional environmental risk map and personnel distribution map, including: The sensor features and visual features in the preprocessed multi-source signals are unified in dimensionality to form a fusion feature tensor, which is then input into a convolutional neural network. By performing sliding convolution operation on the fused feature tensor using a 3×3 convolution kernel, low-level features including edges and textures from visual and sensory features are extracted. Then, the feature map is downsampled through a max pooling layer to generate a high-level feature map containing semantic information about fire and personnel. The high-level feature map is input into the region suggestion branch, and multiple candidate regions are generated through a sliding window. After cropping and normalizing the features of each candidate region, the data is fed into the classification subnetwork to determine whether there is an open flame in the region. The coordinate position of the candidate region is corrected by the regression subnetwork, and the region with a confidence level higher than the preset threshold is selected as the final open flame location identification result. The high-level feature map is input into the semantic segmentation branch. The spatial resolution of the high-level feature map is restored through upsampling operation. The low-level features are then fused together, and the smoke concentration level is classified for each pixel through the softmax activation function to obtain the smoke concentration distribution data. The anchor frame mechanism is used to match human targets of different sizes on high-level feature maps. The classification layer determines whether the anchor frame contains a human body, and the regression layer adjusts the anchor frame coordinates to obtain the human body detection box. The number and distribution density of human bodies in each area of ​​the building are counted to form data on human gathering. The results of open flame location identification, smoke concentration distribution data, and personnel gathering data are integrated and aligned with the preset building space structure data to output a real-time, multi-dimensional environmental risk map and personnel distribution map.

5. The intelligent fire escape route recommendation method as described in claim 1, characterized in that, The system continuously monitors changes in the environment and personnel status, dynamically plans escape routes when new congestion occurs, and pushes the information in real time, including: A sliding window algorithm is used to continuously collect environmental sensor data and personnel positioning data, and the data differences within adjacent time windows are compared. Set a congestion judgment threshold. When the population density in a certain passage area exceeds the preset congestion judgment threshold, it is determined that new congestion has occurred. When congestion is detected, an improved A* algorithm and parallel simulation process are executed based on the updated environmental risk map and personnel distribution map to generate a new optimal escape route; The updated escape route information is marked as an emergency message and converted into a format for different terminals for real-time push.

6. A fire escape route intelligent recommendation system, characterized in that, The system includes: The building module is used to integrate building information model, real-time sensor network data, and video surveillance data to build a virtual digital twin environment that is synchronized with the physical fire scene. The preprocessing module is used to fuse signals from temperature sensors, smoke sensors, carbon monoxide concentration sensors, infrared human body detectors, and video surveillance cameras in real time to obtain multi-source signals and preprocess the multi-source signals. The identification module is used to identify the location of open flames, smoke concentration distribution, and personnel gathering based on preprocessed multi-source signals through a convolutional neural network, generating real-time, multi-dimensional environmental risk maps and personnel distribution maps; The planning module is used to perform parallel simulations in a digital twin environment, using an improved A* algorithm to plan escape routes and obtain the optimal personalized escape path; The real-time push module is used to continuously monitor changes in the environment and personnel status. When new congestion occurs, it dynamically plans escape routes and pushes them in real time. The planning module is used for parallel simulation in a digital twin environment, employing an improved A* algorithm to plan escape routes and obtain the optimal personalized escape path, including: Based on the generated environmental risk map and personnel distribution map, a dynamic path planning graph model is constructed in the digital twin environment. The dynamic path planning graph model uses the passage nodes, room nodes, and safety exit nodes in the building as vertices of the graph and the connecting passages between nodes as edges. Initiate a parallel simulation mechanism in a digital twin environment to simultaneously simulate the passage of multiple potential escape routes; The improved A* algorithm is used to calculate the comprehensive cost of each path, and the path with the lowest comprehensive cost is selected as the optimal personalized escape path, including: From the dynamic path planning graph model of the digital twin environment, obtain the node sequence of each path and the travel distance between nodes; Obtain risk level data for nodes in the environmental risk map, as well as personnel characteristic information associated with the personnel distribution map; The risk level data of nodes is converted into quantitative weight values, and the weights of multiple consecutive nodes with the same risk level in the path are superimposed to obtain the comprehensive risk weight of the path. Based on path-related personnel characteristic information, personnel adaptation factors are determined, which are related to personnel's age, heart rate, and mobility. The new heuristic function is: Heuristic cost = Basic distance cost from the current node to the safe exit node × Comprehensive risk weight of the path × Personnel suitability factor; The improved A* algorithm's cost calculation module is invoked to calculate the actual cost of each path by multiplying the travel distance from the starting point to each node by the real-time congestion coefficient, which is determined by the real-time population density. The actual cost is then added to the heuristic cost of each node to obtain the comprehensive cost of each path. A cost ranking algorithm is then initiated to rank the comprehensive costs of all paths, extracting the path with the lowest cost as the initial optimal path. Simultaneously, it is checked whether the path contains unmarked temporary obstacles or sudden congestion. If so, the path is removed, and the next lowest cost path is selected. This iterative process yields the optimal personalized escape path.

7. A fire escape route intelligent recommendation device, characterized in that, The intelligent fire escape route recommendation device includes a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the intelligent fire escape route recommendation device to perform the various steps of the intelligent fire escape route recommendation method as described in any one of claims 1-5.

8. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the steps of the intelligent fire escape route recommendation method as described in any one of claims 1-5.

Images