Emergency evacuation path generation method, system and device under multi-target conflict and medium

Abstract

The invention relates to an emergency evacuation path generation method, system and device under multi-target conflict and a medium. The method comprises the following steps: acquiring video monitoring data and environment sensing data of a target building, and performing fire disaster detection according to the video monitoring data to obtain a fire detection result; in response to the fire detection result that the fire disaster exists, performing fire evolution prediction according to the environment sensing data based on the environment topological graph of the target building to obtain fire evolution parameters; performing risk assessment and latest passable time marking on each space unit node in the environment topological graph according to the fire evolution parameters to obtain a risk graph layer; and in response to the obtained positions of the to-be-evacuated persons, generating an exclusive emergency evacuation path of each to-be-evacuated person in combination with the environment topological graph and the risk map layer. According to the method, fire detection and fire spreading deduction are carried out through continuous video image frames, time sequence judgment is carried out on traffic effectiveness, and path design is prevented from lagging behind risk change.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of path planning, and particularly relates to an emergency evacuation path generation method, system, device and medium under multi-target conflict. BACKGROUND

[0002] With the development of smart city and public safety field technology, emergency evacuation route planning technology appears, which detects public safety events through sensor perception data or video image, and plans evacuation paths according to real BIM models.

[0003] In the traditional emergency evacuation system, the visual features such as open fire and smoke in the image are identified by combining a deep learning model, or the data features of sensor perception data are extracted, which has the advantages of non-contact, wide coverage and fast response. At the same time, BIM or grid map is combined with environmental modeling method to generate emergency evacuation paths.

[0004] However, the above method relies on video monitoring and sensors to obtain the evolution of the fire, and when the fire is serious, it will damage the circuit or monitoring equipment, so that the monitoring of the evolution of the real fire source is lost, and it is difficult to ensure the effectiveness of the planned route, and it is difficult to accurately reflect the danger level of different spaces changing over time. The existing path generation method only considers the shortest distance, and does not effectively integrate multiple factors such as personnel distribution, risk level evolution and exit load, resulting in path congestion, traffic conflict or entering an already dangerous area, etc., which is difficult to adapt to ensure the practicality of the evacuation path. SUMMARY

[0005] Therefore, it is necessary to provide an emergency evacuation path generation method, system, device and medium under multi-target conflict, which can perform time-effectiveness analysis and path space planning on the fire.

[0006] In a first aspect, the application provides an emergency evacuation path generation method under multi-target conflict, comprising:

[0007] Obtain video monitoring data and environmental perception data of a target building, and perform fire disaster detection according to the video monitoring data to obtain a fire detection result; the video monitoring data includes a plurality of video image frames and corresponding time stamps; the fire detection result includes the existence of fire disaster and non-fire;

[0008] In response to the fire detection result being the existence of fire disaster, perform fire evolution prediction based on the environmental topology map of the target building according to the environmental perception data to obtain fire evolution parameters; the fire evolution parameters include thermal hazard parameters of each space unit node and corresponding evolution time points;

[0009] According to the fire evolution parameter, each space unit node in the environment topology graph is evaluated in danger and the latest passable time is marked to obtain a risk layer;

[0010] In response to the acquired position of the evacuation personnel, the environment topology graph and the risk layer are combined to generate an exclusive emergency evacuation path for each evacuation personnel.

[0011] In one of the embodiments, the environment topology graph is obtained by the following method:

[0012] Obtain the building structure data of the target building; the building structure data includes a passable area and an exit position;

[0013] According to the building structure data, the target building is divided into a plurality of grid units;

[0014] Each grid unit is mapped to a space unit node;

[0015] According to the passable area, a connected relationship edge is constructed between two adjacent space unit nodes, and an environment topology graph is obtained; the environment topology graph includes a plurality of space unit nodes and corresponding connected relationship edges; the space unit node includes an exit node corresponding to the exit position.

[0016] In one of the embodiments, the fire disaster detection is performed according to the video monitoring data to obtain a fire detection result, including:

[0017] The video image frames are input into a pre-trained fire target detection model frame by frame according to the timestamp to obtain a set of potential flame regions; the set of potential flame regions includes a plurality of candidate bounding boxes; the fire target detection model detects the fire target according to a pre-set confidence threshold;

[0018] According to the video image frames, the set of potential flame regions is verified for flame features to obtain a fire detection result.

[0019] In one of the embodiments, the set of potential flame regions is verified for flame features according to the video image frames to obtain a fire detection result, including:

[0020] According to the spatial position consistency of the plurality of candidate bounding boxes in the continuous video image frames, a plurality of flame objects are divided;

[0021] According to the continuous video image frames distributed according to the timestamp, the flame objects are detected in color space, and compared with a pre-set flame color threshold to obtain a flame color result; the flame color result includes flame pixel distribution and flame color percentage; the flame pixel distribution includes flame color pixel position distributed according to the timestamp;

[0022] According to the pixel distribution, the change feature of the pixel contour of the flame object is extracted to obtain a morphological disturbance feature;

[0023] If the flame color percentage exceeds the preset threshold value, and the morphological disturbance feature meets the preset change trend condition, it is determined that the fire detection result is that there is a fire disaster;

[0024] If the flame color percentage is below the preset threshold value, or the morphological disturbance feature does not meet the preset change trend condition, it is determined that the fire detection result is not a fire.

[0025] In one of the embodiments, based on the environmental topology map of the target building, fire evolution prediction is performed according to the environmental perception data to obtain fire evolution parameters, including:

[0026] According to the position of the flame object detected as existing a fire disaster in the video image, the corresponding space unit node is located, and a trusted fire source area is obtained;

[0027] Based on the environmental perception data, and taking the trusted fire source area as an initial heat source, fire spread simulation is performed on the environmental topology map through a preset fluid dynamics equation to obtain heat radiation, convective temperature value and smoke concentration value of each space unit node at an evolution time point;

[0028] Based on each evolution time point, heat hazard parameters of each space unit node are calculated according to the heat radiation, convective temperature value and smoke concentration value to obtain the fire evolution parameters.

[0029] In one of the embodiments, according to the fire evolution parameters, a risk layer is obtained by performing risk assessment and latest passable time marking on each space unit node in the environmental topology map, including:

[0030] When the heat hazard parameter exceeds the preset passable threshold value, the corresponding space unit node is determined as a dangerous node;

[0031] The evolution time point corresponding to the space unit node converted into the dangerous node is determined as the latest expiration time;

[0032] Taking the dangerous node as a center point, the space unit node is expanded outward according to the connected relationship edge, and the safety value of each space unit node is determined based on a preset safety value strategy to obtain a safety layer;

[0033] The safety layer corresponding to each latest expiration time is arranged in turn according to the time flow order of the latest expiration time to obtain the risk layer; the risk layer includes multiple safety layers and their corresponding latest expiration times.

[0034] In one of the embodiments, in response to the obtained position of the personnel to be evacuated, an exclusive emergency evacuation path of each personnel to be evacuated is generated in combination with the environmental topology map and the risk layer, including:

[0035] According to the space unit node corresponding to the position of the to-be-evacuated person, the load of each space unit node at the current time is obtained;

[0036] Based on the multi-target satisfaction principle, an exclusive emergency evacuation path is generated for each to-be-evacuated person according to the connected relationship edge; the multi-target principle includes a moving time principle, a shorter distance principle, a safety distance principle, and a load balancing principle;

[0037] The moving time principle corresponds to the following steps:

[0038] The safety layer corresponding to the current time and the latest expiration time of the corresponding safety layer are obtained;

[0039] If the time difference between the current time and the latest expiration time is less than the preset moving time length, the next space unit node of the current space unit node is calculated according to the safety layer corresponding to the current time; the moving time length is the expected moving time length of the personnel between two adjacent space unit nodes of the exclusive emergency evacuation path;

[0040] If the time difference between the current time and the latest expiration time is greater than or equal to the preset moving time length, the next space unit node of the current space unit node is calculated according to the safety layer of the next latest expiration time of the safety layer corresponding to the current time;

[0041] The update state of the safety layer is rechecked when the next space unit node searches for the path;

[0042] The shorter distance principle corresponds to the map distance value of the next space unit node of the current space unit node being less than the current space unit node; the map distance value is determined by a distance map; the distance map is generated by taking all exit nodes as starting points and calculating the shortest distance from each space unit node to the nearest exit node; the distance map includes the map distance value corresponding to each space unit node;

[0043] The safety distance principle corresponds to the safety value of the next space unit node being greater than the preset safety threshold value when the next space unit node is calculated in the safety layer; the safety threshold value is the safety value of the immediately adjacent dangerous node;

[0044] The load balancing principle corresponds to the load corresponding to the next space unit node being less than the load of the current space unit node.

[0045] In a second aspect, the application also provides an emergency evacuation path generation system under multi-target conflict, comprising:

[0046] A fire checking module is configured to obtain video monitoring data and environmental perception data of a target building, and perform fire disaster detection according to the video monitoring data to obtain a fire detection result; the fire detection result includes a fire disaster and a non-fire disaster;

[0047] a fire evolution prediction module, configured to, in response to the fire detection result being that there is a fire disaster, perform fire evolution prediction based on the environmental topology map of the target building according to the environmental perception data, to obtain fire evolution parameters; the fire evolution parameters include thermal hazard parameters of each space unit node and corresponding evolution time points;

[0048] a building safety evolution module, configured to perform danger assessment and latest passable time marking on each space unit node in the environmental topology map according to the fire evolution parameters, to obtain a risk layer;

[0049] a path generation module, configured to, in response to the obtained positions of the to-be-evacuated personnel, generate exclusive emergency evacuation paths for the to-be-evacuated personnel in combination with the environmental topology map and the risk layer.

[0050] In a third aspect, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the emergency evacuation path generation method under multi-target conflict as described above when executing the computer program.

[0051] In a fourth aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the computer program implements the steps of the emergency evacuation path generation method under multi-target conflict as described above when executed by a processor.

[0052] The emergency evacuation path generation method, system, device and medium under multi-target conflict as described above realize autonomous perception and rapid response to complex disaster scenarios through full-process closed-loop processing of fire source identification, dynamic spread simulation and path generation decision. Through the construction of the latest passable time and the dynamic risk layer, the path planning is no longer limited to the current state judgment, but has the ability to predict the future risk evolution, significantly improving the path stability and personnel safety in the high dynamic fire environment, and enhancing the real-time response ability of path generation to fire changes. Based on the expandable environmental topology map structure and the hierarchical path construction mechanism, it can adapt to different building layouts and personnel distribution modes, realize individualized path customization in high-density environments, and ensure that each evacuating personnel has an exclusive and dynamically updated safe route. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments or the related art, the following will briefly introduce the drawings needed to be used in the embodiment or related art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0054] Figure 1 It is a flowchart of the emergency evacuation path generation method under multi-target conflict of the present application.

[0055] Figure 2 a sub-step flowchart of step S102;

[0056] Figure 3 a sub-step flowchart of step S103;

[0057] Figure 4 a composition structure of the emergency evacuation path generation system under multi-target conflict of the application. DETAILED DESCRIPTION

[0058] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0059] In one embodiment, as shown in Figure 1 a method for generating an emergency evacuation path under multi-target conflict is provided, and the embodiment takes the method applied to a terminal as an example. It should be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is realized through the interaction of the terminal and the server. In the embodiment, the method includes the following steps:

[0060] S101, acquiring video monitoring data and environment perception data of a target building, and detecting a fire disaster according to the video monitoring data to obtain a fire detection result; the video monitoring data includes a plurality of video image frames and corresponding time stamps; the fire detection result includes the presence of a fire disaster and non-fire.

[0061] Illustratively, the video monitoring data inside the target building can be sourced from a visible light camera, and the environment perception data can include multi-source information collected by temperature sensors, smoke sensors, air quality sensors, etc. Further, to ensure the continuity and real-time nature of the image recognition process, the video data is processed in the form of image frame sequences and bound with corresponding time stamps to construct a complete video timing structure. Optionally, a pre-trained fire target detection model is used to detect potential flame areas in the video image frame by frame. Illustratively, the fire target detection model can use spatial convolution and feature pyramid strategy to extract texture, color and edge features of the flame from high-level semantic information to identify open fire or smoke signs.

[0062] S102, in response to the fire detection result being the presence of a fire disaster, performing fire evolution prediction according to the environment perception data based on the environment topology graph of the target building to obtain fire evolution parameters; the fire evolution parameters include thermal hazard parameters of each spatial unit node and corresponding evolution time points.

[0063] After confirming the fire disaster in the building, in order to realize the predictable extension of the disaster space, the environment topology map of the target building is called out schematically, and a fire evolution prediction model is constructed in combination with the environment perception data. The environment topology map can be established in the form of a regular grid modeling, that is, the building space is divided into discrete space unit nodes according to the preset size, and the node connection relationship is established according to the physical channel information, so as to reflect the spatial constraint of the building traffic structure. The fire evolution prediction takes the credible fire source as the initial heat source node, uses the simulation mechanism based on computational fluid dynamics (CFD), simulates the coupling dynamics of heat propagation, smoke diffusion and ventilation disturbance, and inputs the parameters such as ventilation state, room temperature gradient, obstacle material and the like as boundary conditions into the fire evolution model, so that each space unit node can output the key thermal hazard indexes such as heat radiation intensity, convective temperature, smoke concentration and visibility within a given time period.

[0064] S103, risk assessment and latest passable time marking of each space unit node in the environment topology map according to the fire evolution parameters, to obtain a risk layer.

[0065] Further, according to the fire evolution parameters, the risk level of each space unit node in the topology map is evaluated and the time attribute is marked, and a dynamic risk layer is constructed. The layer takes the space unit node as the basic unit, takes the maximum risk value and evolution trend in the fire parameters as the evaluation basis, marks the latest passable time of each node, that is, the latest time point at which the node can still be safely crossed in time. When the heat radiation, smoke concentration or toxic dose of a node first exceeds the preset safe passing threshold at a certain time, the time is defined as the expiration time of the node. Starting from the credible fire source area, the topology connection relationship between nodes is expanded layer by layer outward, a safety layer structure based on time sequence is established, and the structure reflects the risk propagation and boundary of each space unit area in the process of dynamic evolution of the fire. In combination with the safety layer level, the latest passable time and the state of the adjacent nodes of each node, a dynamic risk layer is formed, which is used as the basis for space partitioning and path avoidance.

[0066] S104, in response to the acquired positions of the personnel to be evacuated, in combination with the environment topology map and the risk layer, an exclusive emergency evacuation path of each personnel to be evacuated is generated.

[0067] To realize the individualized response of personnel and the non-congestion guidance of dynamic path, illustratively, after obtaining the position of the current evacuation personnel, the node where the personnel is located is taken as the starting point of the path, and the path planning and generation are performed under multi-objective conflict in combination with the risk layer and the environmental topology map. Illustratively, the path generation algorithm is based on an improved multi-level breadth-first search mechanism, and three types of objective functions are fused for comprehensive decision-making based on the latest passable time label, respectively for the shortest path distance, the optimal path safety, and the path load balancing. In the path generation process, the optimal adjacent node that meets all the objective conditions is selected step by step, and whether the path has an overtime risk and a congestion trend is evaluated in each step, so as to ensure that each path generated finally has high safety, high pass efficiency, and good group scheduling effect. Optionally, the generated path can guide the user to implement evacuation through voice guidance, emergency light control, or the terminal device of the user in necessary cases.

[0068] In the above-mentioned emergency evacuation path generation method under multi-objective conflict, fire detection is performed through continuous video image frames, which not only can quickly identify the sudden fire source, but also can analyze the dynamic trajectory of the flame object in combination with the time stamp, reducing the false alarm and missed alarm problems caused by similar accidental visual features in traditional image detection, and significantly improving the accuracy and robustness of fire identification. With the aid of environmental perception data and building topology information, the processes of heat conduction and smoke diffusion are simulated through a fluid dynamics model, which can predict the fire risk level of each space region at different time points in the future, realize the forward-looking understanding of fire evolution, and avoid the static prediction. The concept of the latest passable time is introduced, which can accurately map the fire influence state of each space element node to a dynamic time label, and has the time sequence judgment ability of node passability in path planning, avoiding the lag of path design behind the risk change. The multi-level risk layer formed based on the thermal hazard parameters and the evolution time points not only reflects the spatial hierarchical structure of the space danger level, but also records when each node evolves into an impassable state, forming a time-space risk database that can be queried in real time by the path algorithm. In the path generation, multi-objective optimization mechanisms such as distance priority, safety redundancy, load balancing, and time window control are introduced, which can flexibly switch between different evacuation strategies, dynamically generate evacuation paths with individual adaptability and global coordination according to the building layout, personnel distribution, and fire development.

[0069] In one of the embodiments, the environmental topology map is obtained by the following method:

[0070] S11, obtain building structure data of a target building; the building structure data includes passable areas and exit positions.

[0071] Illustratively, the building structure data can come from a BIM (Building Information Modeling) model of the building, engineering drawings, or a three-dimensional point cloud of the building reconstructed based on a laser radar, and is preprocessed and converted into two-dimensional planar structured information containing passable areas and exit locations. In a specific data structure, the passable areas mark the scope in which personnel are allowed to pass, including corridors, staircases, open halls, etc., and the exit locations mark the actual layout points of evacuation passages or escape doors, which are used as end point constraints in subsequent path planning.

[0072] S12, grid division of the target building according to the building structure data, to obtain a plurality of grid cells.

[0073] Further, the entire building plan area is grid divided to form a plurality of rectangular or hexagonal basic cells with the same side length, referred to as grid cells. The division granularity can be dynamically set according to the building area size and the path planning accuracy requirement, and illustratively, is set to 1 meter x 1 meter per cell.

[0074] S13, mapping of each grid cell to a spatial cell node.

[0075] Each valid grid cell is mapped to a spatial cell node. The validity is determined by whether the grid is in a passable area, and specifically, a grid located in a wall, obstacle, or closed space will not participate in node construction. Each spatial cell node has a unique number and position coordinates, and can have additional attribute identifiers, such as whether it is an exit node. If a grid cell covers an exit area, the corresponding node will be marked as an exit node, which will be given priority as an evacuation endpoint in subsequent path planning.

[0076] S14, construction of a connected relationship edge between two adjacent spatial cell nodes according to the passable area, to obtain an environment topology graph; the environment topology graph includes a plurality of spatial cell nodes and corresponding connected relationship edges; the spatial cell nodes include exit nodes corresponding to exit locations.

[0077] Illustratively, if two adjacent grid cells are both located in a passable area and share an edge or a corner point, a connected relationship edge will be established between the corresponding spatial cell nodes. The weight of the edge can be set as a unit distance, a floor conversion cost, or an inverse of the passing speed, etc. according to actual needs, so as to be used for cost calculation in subsequent path planning. The finally constructed environment topology graph can be formally represented as G=(V, E), where V is a set of spatial cell nodes, and E is a set of connected edges between nodes. This graph not only accurately describes the structural connection of the building space, but also reflects the building function partition and the passable accessibility through node attributes and edge attributes, providing a unified data basis for fire evolution modeling and multi-objective path planning.

[0078] In one of the embodiments, fire disaster detection is performed according to the video monitoring data, and a fire detection result is obtained, including:

[0079] S21, inputting the video image frames into a pre-trained fire target detection model frame by frame according to timestamps to obtain a set of potential flame regions; the set of potential flame regions includes a plurality of candidate bounding boxes; the fire target detection model performs fire target detection on the video image frames according to a pre-set confidence threshold.

[0080] Illustratively, based on the collected building video monitoring data, image sequences therein are input into a pre-trained fire target detection model frame by frame according to timestamps. Illustratively, the model can be a convolutional neural network structure based on YOLOv8, and is pre-set with class labels for detecting typical fire image features such as open flames and smoke. Specifically, a plurality of candidate bounding boxes are output for each image frame, for representing regions where flames may exist. The determination of the candidate regions is based on the confidence score output by the model, and a low confidence threshold is set to retain a plurality of suspicious bounding boxes, thereby generating a set of potential flame regions composed of a plurality of candidate bounding boxes.

[0081] S22, verifying the set of potential flame regions according to the video image frames for flame features to obtain a fire detection result.

[0082] A flame feature verification process is introduced to filter the candidate results based on time continuity and visual disturbance features. Illustratively, based on the dual verification mechanism in the time dimension, dynamic trajectory matching, morphological disturbance analysis, and color space consistency judgment are performed on each candidate flame region, thereby reducing false positives caused by factors such as changes in lighting and objects with flame colors. When the candidate region exhibits a dynamic evolution trend consistent with the flame features in multiple time frames, and the color proportion meets the prior distribution condition of the flame, it is determined as a credible fire source, and a fire detection result is output.

[0083] In one of the embodiments, fire disaster detection is performed according to the video image frames, and a fire detection result is obtained, including:

[0084] S31, dividing the plurality of candidate bounding boxes into a plurality of flame objects according to spatial position consistency in consecutive video image frames.

[0085] Illustratively, if the positions of two bounding boxes in consecutive image frames are highly close and their sizes do not change significantly, it can be inferred that they belong to the same physical entity, i.e., constitute the same flame object. In this way, a plurality of flame objects are constructed on the time axis, and each object corresponds to a stable tracking sequence of candidate regions.

[0086] S32, color space detection is performed on the flame object according to the continuous video image frames distributed according to the time stamp, and compared with a preset flame color threshold value to obtain a flame color result; the flame color result includes flame pixel distribution and flame color percentage; the flame pixel distribution includes flame color pixel positions distributed according to the time stamp.

[0087] The corresponding region of the flame object in each frame is extracted as an image block, and is converted to a color space such as HSV (color space) or YCbCr (color space), and the distribution range of the flame color pixel thereof is analyzed. By comparison with a preset color threshold value model, for example, the color threshold value can be compared according to the characteristics of the high-brightness orange yellow and medium-high brightness area of the flame, and further, the number of flame color pixels contained in each image frame of the flame object is quantified, so as to calculate the color percentage. If a flame object shows a high proportion of flame color pixels in multiple time frames, it means that it has the color continuity of fire characteristics.

[0088] S33, the change feature of the pixel contour of the flame object is extracted according to the pixel distribution to obtain a morphological disturbance feature.

[0089] The deformation feature of the edge contour of the flame object is extracted according to the flame pixel distribution, mainly analyzing the change frequency, contour fluctuation degree and stability index thereof with respect to the time frame, and referring to the dynamic morphological features of the natural jumping and irregularity of the flame. If an object shows an edge change trend consistent with the disturbance feature of the flame in continuous image frames, such as boundary expansion and contraction or unstable flickering in a short period, the credibility of the flame is significantly improved.

[0090] S34, if the flame color percentage exceeds the preset threshold value, and the morphological disturbance feature meets the preset change trend condition, it is determined that the fire detection result is that there is a fire disaster.

[0091] When the color percentage of a flame object exceeds the set threshold value, and the morphological disturbance index shows a change trend consistent with the natural morphology of the flame, such as edge standard deviation jumping within a preset range, it is determined as a real fire source, and the fire detection result is that there is a fire disaster.

[0092] S35, if the flame color percentage is below the preset threshold value, or the morphological disturbance feature does not meet the preset change trend condition, it is determined that the fire detection result is not a fire.

[0093] If the color feature does not meet the threshold requirement, or the disturbance morphology lacks periodicity or does not have the jumping feature, the system judges that it is not a fire, effectively suppressing false positives.

[0094] The method fully integrates the efficiency of depth image target detection and the dynamic feature verification mechanism in time dimension, and greatly improves the accuracy and robustness of building interior fire identification through multi-level and multi-modal analysis means.

[0095] In one embodiment, as shown in Figure 2 Based on the environmental topology map of the target building, fire evolution prediction is performed according to the environmental perception data to obtain fire evolution parameters, including:

[0096] S201, according to the position of the flame object detected as existing fire disaster in the video image, the corresponding space unit node is located, and the credible fire source area is obtained.

[0097] Illustratively, according to the position of the flame object detected as existing fire disaster in the video image, it is mapped to the corresponding space unit node in the environmental topology map, so as to locate the initial heat source area of the fire. Since the environmental topology map is a grid representation of the building space, the center coordinates of the flame object can be converted to node index through image-space mapping. If the flame object covers multiple grid areas, all the nodes involved can be formed into a set, which is defined as a credible fire source area. This area will be used as the initial heat source for fire spread simulation and become the boundary condition and source term in the heat dynamics model.

[0098] S202, based on the environmental perception data, and taking the credible fire source area as the initial heat source, fire spread simulation is performed on the environmental topology map through a preset fluid dynamics equation to obtain the heat radiation, convective temperature value and smoke concentration value of each space unit node at the evolution time point.

[0099] Illustratively, the environmental perception data usually includes wind speed and direction for convection term modeling, indoor initial temperature for heat balance initial condition, and air flow for smoke migration path adjustment. The obstacle type and combustible material distribution of the target building itself can also be used as the initial condition for fire spread simulation. Specifically, a numerical simulation mechanism based on a pre-designed computational fluid dynamics (CFD) equation is used, which can select FDS (Fire Dynamics Simulator) or other lightweight models that can embed path simulation. Based on the construction of the environmental topology map, each space unit node is regarded as the basic diffusion unit of heat, gas and smoke. The heat source node injects heat flow and combustible gas source term at t=0. Through multi-step iterative calculation of heat conduction, radiation exchange, convection transport and mass migration process, the temperature, radiation intensity and smoke concentration of each node at multiple evolution time points are generated.

[0100] Optionally, due to the high heterogeneity of the building structure and the heat conduction path, the difference in connectivity weight between nodes can be considered in the simulation. For example, wall thickness, ventilation opening state, etc. can all have a significant impact on the propagation of thermal parameters.

[0101] S203, based on each evolution time point, the thermal hazard parameters of each space unit node are calculated according to the heat radiation, convective temperature value and smoke concentration value, and the fire evolution parameters are obtained.

[0102] Further, the simulation results are converted into thermal hazard parameters that can be used for evacuation path evaluation. For example, the heat radiation value, convective temperature value and smoke concentration value of each space unit node at different evolution time points are analyzed comprehensively. Specifically, the thermal parameters can be converted into unit time dose values by the FED (Fractional Effective Dose) model, the cumulative heat dose received by each node is calculated in time sequence, and the heat hazard parameters are compared with the human body tolerance threshold, so as to determine the safety of passing through at any time, and the obtained thermal hazard parameters should include spatial distribution node position and time distribution, thereby forming a space-time joint of fire evolution parameters.

[0103] In one embodiment, as shown in Figure 3 According to the fire evolution parameters, the risk layer is obtained by evaluating the risk of each space unit node in the environment topology graph and marking the latest passable time, including:

[0104] S301, when the thermal hazard parameter exceeds the preset passable threshold, the corresponding space unit node is determined as a dangerous node.

[0105] Illustratively, the passability of each space unit node is determined according to the thermal hazard parameters thereof. Specifically, a set of passable thresholds based on the human body heat exposure safety limit is preset, if any thermal hazard parameter of a space unit node at a certain evolution time point exceeds the passable threshold corresponding thereto, the node is regarded as entering an impassable state, and the node is marked as a dangerous node at the time point accordingly.

[0106] S302, the evolution time point corresponding to the dangerous node converted from the space unit node is determined as the latest expiration time.

[0107] Further, the time point at which each dangerous node is converted from a passable state to a dangerous state is recorded, i.e. the evolution time point corresponding to the node is defined as the latest expiration time, thereby forming a time window constraint condition in path scheduling.

[0108] S303, taking the dangerous node as the center point, expanding the space unit node outward according to the connectivity relationship edge, and determining the safety value of each space unit node based on the preset safety value strategy, to obtain a safety layer.

[0109] Centered on the dangerous node, according to its connectivity relationship in the topology graph, a plurality of nested safety layers are formed by expanding outward layer by layer. Each safety layer is composed of space unit nodes directly or indirectly connected to the dangerous node, and the connectivity depth between nodes is taken as the hierarchical basis. Further, a safety value attribute is introduced in each safety layer to measure the distance or buffer period of the current node relative to the dangerous area. For example, the safety value of the dangerous node can be defined as 0, the safety value of the dangerous adjacent node directly connected to the dangerous node is 1, the safety value of the node connected to the dangerous adjacent node is 2, and so on, increasing outwardly. The safety value can also be determined according to the minimum connectivity step of the node to the dangerous area, or the shortest remaining time required to reach the dangerous state.

[0110] S304, arrange the safety layers corresponding to each latest expiration time in the time sequence of the latest expiration time in turn to obtain a risk layer; the risk layer includes a plurality of safety layers and the corresponding latest expiration time of each safety layer.

[0111] Illustratively, the safety layer structures corresponding to all different latest expiration times are arranged in time sequence in turn to form a risk layer with time sequence logic. The risk layer not only contains the known high-risk area information in the current space structure, but also clearly indicates when each safety layer will evolve into an impassable area.

[0112] In one embodiment, in response to the obtained positions of the personnel to be evacuated, an exclusive emergency evacuation path for each personnel to be evacuated is generated in combination with the environmental topology graph and the risk layer, including:

[0113] S41, obtain the load of each space unit node at the current time according to the space unit node corresponding to the position of the personnel to be evacuated.

[0114] Illustratively, the current position of the personnel to be evacuated is obtained and mapped to the corresponding space unit node in the environmental topology graph. Based on the current time point, the safety layer and the latest expiration time corresponding to each node in the risk layer are retrieved, and the distribution of the registered evacuation personnel on each node is counted, thereby constructing a load layer at the current time. The layer records the number of personnel or the passing frequency per unit time in the space unit, reflecting the possibility of path congestion and providing necessary support for load balancing in path generation.

[0115] S42, generate an exclusive emergency evacuation path for each personnel to be evacuated based on a multi-objective satisfaction principle according to the connectivity relationship edge; the multi-objective principle includes a moving time principle, a shorter distance principle, a safety distance principle, and a load balancing principle.

[0116] The moving time principle corresponds to the following steps:

[0117] obtain a safety layer corresponding to the current time and a latest expiration time of the corresponding safety layer.

[0118] If a time difference between the current time and the latest expiration time is less than a preset moving time length, a next space unit node of the current space unit node is calculated according to the safety layer corresponding to the current time; the moving time length is a predicted moving time length of personnel between two adjacent space unit nodes of the exclusive emergency evacuation path.

[0119] If the time difference between the current time and the latest expiration time is greater than or equal to the preset moving time length, a next space unit node of the current space unit node is calculated according to a safety layer of a next latest expiration time of the safety layer corresponding to the current time.

[0120] The next space unit node rechecks the update state of the safety layer when performing path search.

[0121] The shorter distance principle corresponds to a map distance value of the next space unit node of the current space unit node being less than the current space unit node; the map distance value is determined by a distance map; the distance map is generated by taking all exit nodes as starting points and calculating the shortest distance from each space unit node to the nearest exit node; and the distance map includes the map distance value corresponding to each space unit node.

[0122] The safety distance principle corresponds to the safety value of the next space unit node being greater than a preset safety threshold value when the next space unit node is calculated in the safety layer; the safety threshold value is the safety value of the immediately adjacent dangerous node.

[0123] The load balancing principle corresponds to the load corresponding to the next space unit node being less than the load of the current space unit node.

[0124] Illustratively, the four multi-objective satisfaction principles are used to start the personalized path search process for each person to be evacuated. In this process, the next hop space unit node decision of each space unit node of the path not only needs to meet the single shortest path objective, but also needs to comprehensively consider the time safety window, space safety redundancy and traffic dispersion demand to ensure that the path has high feasibility and low conflict rate under the dynamic evolution of the fire.

[0125] Specifically, the moving time principle ensures consistency between the time rationality of the path and the passability of the node. In the path search, the safety layer in the risk layer according to the current time difference is found, and the latest expiration time of the current node is evaluated. If the difference between the current time and the latest expiration time is less than the predicted moving time length of the evacuation object to the next node, the current node will be invalid, and the safety layer of the next time period needs to be switched in advance to reposition the next node of the path, so as to avoid the situation of encountering node failure in the middle of the way.

[0126] The shorter distance principle guarantees the overall evacuation timeliness of the path. A distance map is constructed, i.e., taking each exit node as a starting point, the shortest path cost from each spatial unit node to the nearest exit is calculated by using the reverse breadth-first or Dijkstra algorithm. When the path jumps, the next spatial unit node with a map distance value smaller than the current node is preferentially selected to ensure that the overall path monotonically approaches the evacuation target, thereby avoiding unreasonable situations such as regional detours or path backtracking.

[0127] Further, the safety distance principle requires that each hop in the path operates outside the preset minimum safety redundancy. In the path jumping process, the safety value of the next node must be higher than the minimum safety value of the adjacent dangerous area, and the path jumping is allowed. For example, the safety value of the dangerous adjacent node adjacent to the dangerous area is 1, and the safety value of the next node must be > 1 to be included in the path selection range, effectively avoiding the path close to the high-risk area, improving the path robustness and personnel safety margin.

[0128] The load balancing principle avoids high-density passing nodes in path selection by real-time reading of the current load layer. If there are candidate nodes with a load less than the current node, the node with a lower load will be selected as the jumping target, thereby reducing the risk of congestion in panic, optimizing the passing efficiency, and improving the path distinguishability when multiple personnel evacuate at the same time.

[0129] Optionally, if all conditions cannot be met, safety and distance are ensured first, and then the load is adjusted, always taking safe evacuation as the core goal.

[0130] It should be understood that although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately executed with other steps or steps or stages in other steps.

[0131] Based on the same inventive concept, the embodiments of the present application also provide a system for realizing the emergency evacuation path generation method under multi-target conflict as described above. The system provides a solution to the implementation scheme as described in the above method, and therefore the specific limitations in one or more emergency evacuation path generation system embodiments under multi-target conflict provided below can refer to the limitations of the emergency evacuation path generation method under multi-target conflict described above, which will not be repeated here.

[0132] In one exemplary embodiment, as shown in Figure 4 An emergency evacuation path generation system under multi-target conflict is provided, comprising:

[0133] A fire inspection module 401 is configured to acquire video monitoring data and environmental perception data of a target building, and perform fire disaster detection based on the video monitoring data to obtain a fire detection result. The fire detection result includes the presence of a fire disaster and no fire.

[0134] A fire evolution prediction module 402 is configured to, in response to the fire detection result being the presence of a fire disaster, perform fire evolution prediction based on the environmental topology map of the target building according to the environmental perception data to obtain fire evolution parameters. The fire evolution parameters include thermal hazard parameters of each space unit node and corresponding evolution time points.

[0135] A building safety evolution module 403 is configured to perform hazard assessment and latest passable time labeling on each space unit node in the environmental topology map according to the fire evolution parameters to obtain a risk layer.

[0136] A path generation module 404 is configured to, in response to the acquired position of the evacuees, combine the environmental topology map and the risk layer to generate exclusive emergency evacuation paths for each evacuee.

[0137] In one embodiment, the system further comprises a building model construction module configured to:

[0138] Acquire building structure data of the target building. The building structure data includes passable areas and exit positions.

[0139] Perform grid division on the target building according to the building structure data to obtain a plurality of grid units.

[0140] Map each grid unit to a space unit node.

[0141] Construct a connected relationship edge between two adjacent space unit nodes according to the passable areas to obtain an environmental topology map. The environmental topology map includes a plurality of space unit nodes and corresponding connected relationship edges. The space unit nodes include exit nodes corresponding to the exit positions.

[0142] In one of the embodiments, the fire inspection module 401 is further configured to:

[0143] input the video image frames into a pre-trained fire target detection model frame by frame according to timestamps to obtain a set of potential flame regions; the set of potential flame regions includes a plurality of candidate bounding boxes; the fire target detection model detects fire targets according to a pre-set confidence threshold for the video image frames;

[0144] verify the set of potential flame regions according to flame features of the video image frames to obtain a fire detection result.

[0145] In one of the embodiments, the fire inspection module 401 is further configured to:

[0146] divide the plurality of candidate bounding boxes into a plurality of flame objects according to spatial position consistency of the plurality of candidate bounding boxes in continuous video image frames;

[0147] detect the flame objects according to continuous video image frames distributed according to timestamps, and compare the flame objects with a pre-set flame color threshold to obtain a flame color result; the flame color result includes flame pixel distribution and flame color percentage; the flame pixel distribution includes flame color pixel positions distributed according to timestamps;

[0148] extract a change feature of a pixel contour of the flame object according to the pixel distribution to obtain a morphological disturbance feature;

[0149] if the flame color percentage exceeds a pre-set threshold, and the morphological disturbance feature meets a pre-set change trend condition, determine that the fire detection result is that there is a fire disaster;

[0150] if the flame color percentage is below the pre-set threshold, or the morphological disturbance feature does not meet the pre-set change trend condition, determine that the fire detection result is that there is no fire.

[0151] In one of the embodiments, the fire evolution prediction module 402 is further configured to:

[0152] locate corresponding spatial unit nodes according to positions of the flame objects detected as having the fire disaster in the video images to obtain a trusted fire source area;

[0153] simulate fire spread on an environment topology graph based on environment perception data and taking the trusted fire source area as an initial heat source through a pre-set fluid dynamics equation to obtain heat radiation, convection temperature values and smoke concentration values of the spatial unit nodes distributed according to evolution time points;

[0154] calculate heat hazard parameters of the spatial unit nodes according to the heat radiation, the convection temperature values and the smoke concentration values based on the evolution time points to obtain fire evolution parameters.

[0155] In one of the embodiments, the building safety evolution module 403 is further configured to:

[0156] When the thermal hazard parameter exceeds the preset passable threshold, the corresponding space unit node is determined as a dangerous node;

[0157] The evolution time point corresponding to the conversion of the space unit node to the dangerous node is determined as the latest expiration time;

[0158] Taking the dangerous node as a center point, the space unit nodes are expanded outward according to the connection relationship edges, and the safety values of the space unit nodes are determined based on a preset safety value strategy, to obtain a safety layer;

[0159] The safety layers corresponding to the latest expiration times are arranged in sequence according to the time flow order of the latest expiration times, to obtain a risk map layer; the risk map layer includes the safety layers and the corresponding latest expiration times.

[0160] In one of the embodiments, the path generation module 404 is further configured to:

[0161] According to the space unit node corresponding to the position of the to-be-evacuated person, the load of each space unit node at the current time is obtained;

[0162] Based on a multi-target satisfaction principle, an exclusive emergency evacuation path is generated for each to-be-evacuated person according to the connection relationship edges; the multi-target principle includes a moving time principle, a shorter distance principle, a safety distance principle, and a load balancing principle.

[0163] In one embodiment, a computer device is provided, including a memory and a processor, the memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0164] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the steps in the above method embodiments.

[0165] For the device embodiments, since they basically correspond to the method embodiments, the related parts are described in the method embodiments. The device embodiments described above are only schematic, and the components described as separate components can or can not be physically separate, and the components displayed as a unit can or can not be a physical unit, i.e., they can be located in one place, or distributed on multiple network units. According to actual needs, some or all of the modules can be selected to achieve the purpose of the present disclosure. Those skilled in the art can understand and implement it without creative labor.

[0166] The above-described embodiments only express several implementation manners of the application, the description is more specific and detailed, but it cannot be understood as the limitation of the patent scope of the application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the application, several modifications and improvements can be made, which are within the protection scope of the application.

Claims

1. A method for generating emergency evacuation routes under multi-objective conflict, characterized in that, The method includes: The system acquires video surveillance data and environmental perception data of the target building, and performs fire detection based on the video surveillance data to obtain fire detection results. The video surveillance data includes multiple video image frames and their corresponding timestamps. The fire detection results include whether a fire exists or not. In response to the fire detection result indicating the presence of a fire, fire evolution prediction is performed based on the environmental topology map of the target building and the environmental perception data to obtain fire evolution parameters; the fire evolution parameters include the thermal hazard parameters of each spatial unit node and the corresponding evolution time points. Based on the fire evolution parameters, a risk layer is obtained by performing a hazard assessment and marking the latest passable time for each spatial unit node in the environmental topology map. In response to the obtained location of the personnel to be evacuated, a dedicated emergency evacuation route is generated for each of the personnel to be evacuated, based on the environmental topology map and the risk layer.

2. The method according to claim 1, characterized in that, The environmental topology map was obtained through the following method: Obtain the building structure data of the target building; the building structure data includes accessible areas and exit locations; The target building is divided into multiple grid cells based on the building structure data. Map each of the aforementioned grid cells to the aforementioned spatial cell nodes; Based on the passable area, a connection edge is constructed between two adjacent spatial unit nodes to obtain the environmental topology graph; the environmental topology graph includes multiple spatial unit nodes and their corresponding connection edges; the spatial unit nodes include the exit node corresponding to the exit location.

3. The method according to claim 2, characterized in that, The step of detecting fire based on the video surveillance data and obtaining fire detection results includes: The video image frames are input frame by frame according to the timestamp into a pre-trained fire target detection model to obtain a set of potential flame regions; the set of potential flame regions includes multiple candidate bounding boxes; the fire target detection model performs fire target detection on the video image frames according to a preset confidence threshold; The potential flame region set is verified by analyzing the flame features based on the video image frames to obtain the fire detection results.

4. The method according to claim 3, characterized in that, The step of verifying the flame features of the potential flame region set based on the video image frames to obtain the fire detection result includes: The candidate bounding boxes are divided into multiple flame objects based on the consistency of their spatial positions in consecutive video image frames. The flame object is subjected to color space detection based on consecutive video image frames distributed according to the timestamps, and compared with a preset flame color threshold to obtain a flame color result; the flame color result includes flame pixel distribution and flame color percentage; the flame pixel distribution includes the flame color pixel positions distributed with the timestamps; Based on the pixel distribution, the pixel contour variation features of the flame object are extracted to obtain morphological perturbation features; If the percentage of flame color exceeds a preset threshold and the morphological disturbance features meet preset change trend conditions, the fire detection result is determined to indicate the presence of a fire. If the percentage of flame color is below the preset threshold, or if the morphological disturbance characteristics do not meet the preset change trend conditions, the fire detection result is determined to be a non-fire.

5. The method according to claim 4, characterized in that, The fire evolution prediction based on the environmental topology map of the target building and the environmental perception data yields fire evolution parameters, including: Based on the location of the flame object detected as having a fire in the video image, the corresponding spatial unit node is located to obtain the reliable fire source area; Based on the environmental perception data and using the credible fire source area as the initial heat source, fire spread simulation is performed on the environmental topology map through preset fluid dynamics equations to obtain the thermal radiation, convection temperature and smoke concentration values ​​of each spatial unit node according to the evolution time point. Based on each of the aforementioned evolution time points, the thermal hazard parameters of each of the aforementioned spatial unit nodes are calculated according to the thermal radiation, the convective temperature value, and the smoke concentration value, thereby obtaining the fire evolution parameters.

6. The method according to claim 2, characterized in that, The risk layer is obtained by performing hazard assessment and marking the latest passable time for each spatial unit node in the environmental topology map based on the fire evolution parameters, including: When the thermal hazard parameter exceeds the preset passable threshold, the corresponding spatial unit node is identified as a dangerous node; The evolution time point corresponding to the transformation of the spatial unit node into the dangerous node is determined as the latest expiration time; Using the dangerous node as the center point, the spatial unit nodes are expanded outward according to the connectivity edges, and the safety value of each spatial unit node is determined based on a preset safety value strategy to obtain a safety layer; The security layers corresponding to the latest expiration times are arranged sequentially according to the time flow order of the latest expiration times to obtain the risk layer; the risk layer includes multiple security layers and their corresponding latest expiration times.

7. The method according to claim 6, characterized in that, In response to the acquired locations of the personnel to be evacuated, and in conjunction with the environmental topology map and the risk layer, a dedicated emergency evacuation route is generated for each of the personnel to be evacuated, including: Based on the spatial unit nodes corresponding to the locations of the personnel to be evacuated, the load of each spatial unit node at the current time is obtained; Based on the principle of satisfying multiple objectives, a dedicated emergency evacuation route is generated for each of the persons to be evacuated according to the connectivity edges; the multi-objective principle includes the principle of travel time, the principle of shorter distance, the principle of safe distance, and the principle of load balancing; The aforementioned moving-time principle corresponds to the following steps: Obtain the security layer corresponding to the current time and the latest expiration time of the corresponding security layer; If the time difference between the current time and the latest expiration time is less than the preset movement time, then the next spatial unit node of the current spatial unit node is calculated according to the security layer corresponding to the current time; the movement time is the estimated movement time of personnel between two adjacent spatial unit nodes of the dedicated emergency evacuation path. If the time difference between the current time and the latest expiration time is greater than or equal to the preset moving time, then the next spatial unit node of the current spatial unit node is calculated using the security layer of the next latest expiration time of the security layer corresponding to the current time. The next step involves the spatial unit node re-checking the update status of the security layer during path search; The shorter distance principle corresponds to the map distance value of the next spatial unit node of the current spatial unit node being less than that of the current spatial unit node; the map distance value is determined by a distance map; the distance map is generated by calculating the shortest distance from each spatial unit node to the nearest exit node, taking all exit nodes as the starting point; the distance map includes the map distance value corresponding to each spatial unit node. The safety distance principle corresponds to the requirement that when calculating the next spatial unit node in the safety layer, the safety value of the next spatial unit node must be greater than a preset safety threshold; the safety threshold is the safety value of the adjacent dangerous node. The load balancing principle is that the load of the next spatial unit node is less than the load of the current spatial unit node.

8. An emergency evacuation route generation system under multi-objective conflict, characterized in that, The system includes: The fire detection module is used to acquire video surveillance data and environmental perception data of the target building, and to perform fire detection based on the video surveillance data to obtain fire detection results; the fire detection results include the presence of a fire and the absence of a fire. The fire evolution prediction module is used to respond to the fire detection result indicating the existence of a fire disaster, and to predict the fire evolution based on the environmental topology map of the target building and the environmental perception data to obtain fire evolution parameters; the fire evolution parameters include the thermal hazard parameters of each spatial unit node and the corresponding evolution time points. The building safety evolution module is used to perform hazard assessment and mark the latest passable time for each spatial unit node in the environmental topology map based on the fire evolution parameters, so as to obtain a risk layer. The path generation module is used to generate a dedicated emergency evacuation path for each person to be evacuated in response to the obtained location of the person to be evacuated, in combination with the environmental topology map and the risk layer.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images