Emergency evacuation design and rescue method for large-scale comprehensive building

Abstract

This invention discloses an emergency evacuation design and rescue method for large-scale integrated buildings, specifically in the field of emergency evacuation and rescue design. It addresses the problem of disconnect between rescue route planning and pedestrian guidance caused by the complex internal spatial structure of existing integrated buildings and the lack of visual cognitive support in traditional evacuation designs. By collecting pedestrian movement trajectories from multiple integrated buildings, high-frequency dwelling areas are identified, consensus-based spatial memory features are extracted, and a spatial memory anchor point network is constructed. Evacuation route nodes are marked in the network, and weak visual connections are identified. Miniature fire stations are deployed to form new memory anchor point markers. The guidance utility coefficient is calculated based on fire drill data and converted into dynamic conflict weights. When a fire occurs, a rescue route search is performed in the anchor point network with the goal of minimizing the cumulative dynamic conflict weights. This achieves evacuation guidance optimization and dynamic collaborative design of rescue routes based on visual memory features.

Description

Technical Field

[0001] This invention relates to the field of emergency evacuation and rescue design technology, and more specifically, to an emergency evacuation design and rescue method for a large-scale comprehensive building. Background Technology

[0002] With the increasing number of large-scale integrated buildings, their complex internal spatial structures, diverse functional formats, and dense population flow have made them one of the most challenging scenarios in urban emergency management systems. Traditional evacuation designs typically rely on building floor plans and single-directional signage, lacking a systematic consideration of the actual behavioral characteristics and visual perception patterns of people. This leads to problems such as misjudging directions and choosing routes in unfamiliar environments during sudden emergencies, thus affecting overall evacuation efficiency and the timeliness of rescue operations.

[0003] Existing research largely focuses on crowd density monitoring based on sensor data, failing to fully utilize people's natural spatial habits and visual memory characteristics as a basis for evacuation decision support. Spatial landmarks within complex buildings, such as large billboards, brightly colored columns, service counters, or shop entrances, often form stable visual anchors in people's memories, potentially influencing their direction of movement and path recognition. However, these visual anchors have not been effectively quantified and structured in traditional evacuation design, and there is a lack of dynamic optimization mechanisms based on combining crowd behavior data with building topological characteristics. Furthermore, during fire rescue operations, rescuers often face complex internal spatial networks; traditional rescue route planning lacks dynamic correlation with actual accessibility and evacuation guidance status, potentially leading to conflicts or delays between evacuation and rescue routes.

[0004] Therefore, there is an urgent need for a comprehensive method that integrates spatial memory feature recognition, evacuation route topology analysis, and rescue route dynamic optimization to build an emergency evacuation and rescue system that is more in line with the cognitive patterns of human behavior and improve the safety response capability of complex buildings in emergencies. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide an emergency evacuation design and rescue method for large-scale integrated buildings to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: An emergency evacuation design and rescue method for a large-scale integrated building includes the following steps: S1. Analyze the movement trajectory data of people in multiple complex buildings, identify high-frequency natural dwelling areas, perform visual feature analysis on the spatial landmarks of the dwelling areas, and extract consensus spatial memory features; S2. Based on consensus-based spatial memory features, match specific locations within the target complex building, and construct a spatial memory anchor point network in conjunction with the building structure of the target complex building; S3. Mark the node sequence of the preset evacuation path in the spatial memory anchor network, calculate the topological overlap between the evacuation path nodes and the spatial memory anchors, and identify the weak visual connection sections. S4. Deploy mini fire stations along the preset evacuation route in areas with weak visual connections as new memory anchor points, and connect the mini fire stations to form an evacuation guidance route. S5. Extract fire drill data, calculate the guidance utility coefficient by statistically analyzing the pedestrian flow turning rate around each mini fire station, and convert the guidance utility coefficient into a dynamic conflict weight between evacuation and rescue. S6. When a fire incident occurs, locate the location of the incident in the spatial memory anchor network, and search for a rescue path in the topology of the spatial memory anchor network with the goal of minimizing the cumulative dynamic conflict weight.

[0007] As a further aspect of the present invention, in step S1, analyzing the movement trajectory data of people within multiple integrated buildings, identifying high-frequency natural dwelling areas, clustering these dwelling areas using spatial landmarks, and extracting consensus-based spatial memory features specifically includes: Collect pedestrian movement trajectory data in multiple complex buildings, extract natural rest points from the movement trajectories, and use density-based spatial clustering method to identify high-frequency clustering areas for natural rest points; Analyze the visual characteristics of landmarks in high-frequency clustering areas and extract visual feature data in three dimensions: geometric shape, facade color, and scale. By comparing visual feature data of high-frequency clusters of multiple complex buildings across projects, repeating visual feature combination patterns are selected to construct a consensus-based spatial memory feature set.

[0008] As a further aspect of the present invention, in S2, the specific location within the target complex building is matched based on consensus spatial memory features, and the spatial memory anchor network is constructed in conjunction with the building structure of the target complex building, specifically including: Select a target complex building, divide the interior of the complex building into points based on the shops and supporting service points, and extract the actual visual feature data of the landmarks at each point; The visual feature data of each location marker is matched with the consensus spatial memory feature set to calculate the matching degree index between each location and the consensus spatial memory feature. Specific points with matching scores higher than the preset matching threshold are selected as spatial memory anchor points for the target complex building; Based on the architectural structural connectivity layout of the target integrated building, the spatial topological relationship between each spatial memory anchor point is analyzed, and a spatial memory anchor point network is constructed with spatial memory anchor points as nodes and spatial topological relationships as connections.

[0009] As a further aspect of the present invention, in step S3, marking the node sequence of the preset evacuation path in the spatial memory anchor network, calculating the topological overlap between the evacuation path nodes and the spatial memory anchors, and identifying visually weak sections specifically includes: The turning points in the preset emergency evacuation routes are extracted, integrated into a path node sequence, and marked in the spatial memory anchor network; Calculate the spatial distance from each path node to the nearest spatial memory anchor point, and convert the spatial distance into the topological overlap of each path node; The topological overlap of each path node is connected according to the spatial order of the preset evacuation path to form a path matching curve. The fluctuations in the path alignment curve are retrieved, and path segments in which the topological overlap is consistently below a set alignment threshold are identified and marked as segments with weak visual connections.

[0010] As a further aspect of the present invention, in step S4, deploying mini fire stations as new memory anchor points along the preset evacuation route in areas with weak visual connections, and connecting the various mini fire stations to form an evacuation guidance route specifically includes: Mini fire stations are deployed along the preset evacuation route in areas with weak visual connections as new memory anchor points. The initial layout density of the mini fire stations has a negative linear relationship with the visual connection strength value corresponding to the areas with weak visual connections. Establish a spatial sequence among mini fire stations, form a progressive point layout according to the evacuation direction, connect each mini fire station to construct an evacuation guidance path that coincides with the preset evacuation path, and update the spatial memory anchor point network.

[0011] As a further aspect of the present invention, the visual connection strength value is the average topological overlap of all path nodes within the corresponding visual connection weak segment.

[0012] As a further aspect of the present invention, in step S5, extracting fire drill data, calculating the pedestrian flow turning rate around each mini fire station to determine the guidance utility coefficient, and converting the guidance utility coefficient into a dynamic conflict weight between evacuation and rescue specifically includes: During fire drills, data on the movement trajectories of personnel at mini fire stations are extracted, and the ratio of the number of people moving along the evacuation guidance path at each mini fire station to the total number of people is calculated as the turning rate of the flow of people. By combining the visual connection strength values ​​of the locations of each mini fire station with the pedestrian turning rate and the visual connection strength values, the guidance effectiveness coefficient of each mini fire station is obtained by weighting the pedestrian turning rate with the visual connection strength values. The guidance utility coefficient of each mini fire station is normalized and weighted to serve as the dynamic conflict weight for evacuation and rescue routes.

[0013] As a further aspect of the present invention, in step S6, when a fire incident occurs, the location of the incident is determined in the spatial memory anchor network, and a rescue path search is performed in the topology of the spatial memory anchor network with the goal of minimizing the cumulative dynamic conflict weight. This specifically includes: When a fire occurs, the exact location of the incident is located in the spatial memory anchor network, and that location is set as the target point for the rescue path search. Starting from the entrance of the complex building, the optimal rescue path is determined by searching the topology of the spatial memory anchor network for the connected path with the minimum cumulative dynamic conflict weight.

[0014] The technical effects and advantages of the emergency evacuation design and rescue method for large-scale integrated buildings of the present invention are as follows: By introducing the concept of a spatial memory anchor network, this invention combines the natural dwelling behavior of people within a complex building with the building's spatial topological features to construct an emergency evacuation and rescue system that combines visual recognizability and path accessibility. This method can extract consensus-based spatial memory features from multi-project crowd movement data, achieving cognitive coupling between evacuation signage and building structure, making evacuation route design more aligned with people's actual visual memory patterns. Within the evacuation routes, identifying weak visual connections and deploying mini fire stations to form new spatial anchors not only improves the continuity and recognizability of evacuation guidance but also enhances the stability of personnel's directional judgment under high-pressure environments. Simultaneously, this invention utilizes fire drill data to calculate the guidance utility coefficient of the mini fire stations, transforming crowd response characteristics into quantifiable dynamic conflict weights, providing data support for intelligent search of rescue routes.

[0015] Compared with traditional static evacuation design, this invention achieves coordinated optimization of evacuation guidance and rescue routes, and can dynamically balance the distribution of people and the priority of rescue passage in complex spaces, which greatly improves the efficiency of emergency response and the scientific nature of safety management, and has significant practical value and promotion significance. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of an emergency evacuation design and rescue method for a large-scale comprehensive building according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0018] Figure 1 This invention provides an emergency evacuation design and rescue method for a large-scale integrated building, which includes the following steps: S1. Analyze the movement trajectory data of people in multiple complex buildings, identify high-frequency natural dwelling areas, perform visual feature analysis on the spatial landmarks of the dwelling areas, and extract consensus spatial memory features; S2. Based on consensus-based spatial memory features, match specific locations within the target complex building, and construct a spatial memory anchor point network in conjunction with the building structure of the target complex building; S3. Mark the node sequence of the preset evacuation path in the spatial memory anchor network, calculate the topological overlap between the evacuation path nodes and the spatial memory anchors, and identify the weak visual connection sections. S4. Deploy mini fire stations along the preset evacuation route in areas with weak visual connections as new memory anchor points, and connect the mini fire stations to form an evacuation guidance route. S5. Extract fire drill data, calculate the guidance utility coefficient by statistically analyzing the pedestrian flow turning rate around each mini fire station, and convert the guidance utility coefficient into a dynamic conflict weight between evacuation and rescue. S6. When a fire incident occurs, locate the location of the incident in the spatial memory anchor network, and search for a rescue path in the topology of the spatial memory anchor network with the goal of minimizing the cumulative dynamic conflict weight.

[0019] In step S1, the movement trajectory data of people within multiple integrated buildings are analyzed to identify high-frequency natural dwelling areas. Spatial landmark clustering is performed on these dwelling areas to extract consensus-based spatial memory features, specifically including: Collect pedestrian movement trajectory data in multiple complex buildings, extract natural rest points from the movement trajectories, and use density-based spatial clustering method to identify high-frequency clustering areas for natural rest points; Analyze the visual characteristics of landmarks in high-frequency clustering areas and extract visual feature data in three dimensions: geometric shape, facade color, and scale. By comparing visual feature data of high-frequency clusters of multiple complex buildings across projects, repeating visual feature combination patterns are selected to construct a consensus-based spatial memory feature set.

[0020] A crowd movement data collection device, combining video surveillance and Bluetooth beacons, was deployed inside multiple mixed-use buildings to record people's movement paths and stopping positions in real time. Each collection point was assigned a trajectory sequence with timestamps and coordinate information, with a sampling frequency controlled at once per second to ensure trajectory continuity and spatial resolution. After obtaining complete trajectory data, the changes in people's movement speed over continuous time periods were analyzed to identify segments with speeds below 0.2 meters per second and durations exceeding 6 seconds as natural stopping points, and their locations were automatically marked. False stopping points with durations shorter than this were removed. Each natural stopping point included information such as its floor, planar coordinates, and corresponding time period to describe the natural stopping behavior patterns of people in commercial spaces. Through the overlay analysis of multi-day and multi-time period data, a dataset of stopping point distribution covering various levels and areas was formed.

[0021] Based on the obtained dataset of natural dwelling points, density clustering was used for spatial clustering analysis to identify areas where people frequently stayed across multiple time periods. During clustering, a minimum sample size of 10 and a neighborhood radius of 3 meters were set to ensure the identified clusters had practical spatial significance. By judging the spatial proximity and density levels between dwelling points, several high-frequency clusters were automatically formed, each representing a space where people concentrated. To improve the accuracy of the results, edge points were verified after clustering. If the spatial distance between an edge point and the core area exceeded a set range, that point was removed to avoid excessive expansion of the cluster boundaries. The clustering results were processed by floor mapping to create a visual heatmap, reflecting the high-frequency dwelling distribution characteristics within the complex building. This method clearly identifies key locations where people naturally gather in different functional areas.

[0022] Within each high-frequency clustering area, visual imaging equipment was used to capture images of key landmarks, including advertising pillars inside buildings, shop entrances, facade decorations, and service counters—structures with high visual recognizability. Geometric analysis was performed on the captured image data to extract the outlines and main shape features of the landmarks, classifying them into geometric types such as rectangles, circles, and polygons. In terms of color, a standardized color space model was used to extract the mean value of image pixels, recording the hue and saturation of the dominant color. In terms of scale, the relative height and width of the landmarks in the on-site environment were obtained through on-site measurements or image scaling. Data from all three dimensions were stored quantitatively and identified by area numbers to ensure comparability between data from different mixed-use buildings. After uniformly formatting the visual feature data extracted from multiple mixed-use buildings, a cross-project visual feature comparison matrix was constructed, calculating similarity indices for geometric shape, facade color, and scale. In similarity assessment, landmarks with identical shapes or a similarity exceeding 80% are grouped into the same visual category. For color features, a primary hue deviation of no more than 10 degrees and a saturation difference of no more than 0.15 are set as similarity criteria. Regarding scale, landmarks with a relative size difference of less than 15% are classified as belonging to the same scale category. By combining feature labels across these three dimensions, visual feature patterns that repeatedly appear in multiple mixed-use buildings are filtered out. Recurring landmarks such as "red rectangular columns" and "blue and white entrance structures" are aggregated into the same visual category. The geometric, color, and scale data of landmarks within the same category are standardized, and their average values ​​are calculated to obtain the average feature vector for that visual category. All average feature vectors together form a consensus spatial memory feature set, stored in multi-dimensional vector form, used for continuous spatial matching and feature similarity retrieval within the target mixed-use building.

[0023] In step S2, specific locations within the target complex building are matched based on consensus-based spatial memory features, and a spatial memory anchor point network is constructed in conjunction with the building structure of the target complex building. Specifically, this includes: Select a target complex building, divide the interior of the complex building into points based on the shops and supporting service points, and extract the actual visual feature data of the landmarks at each point; The visual feature data of each location marker is matched with the consensus spatial memory feature set to calculate the matching degree index between each location and the consensus spatial memory feature. Specific points with matching scores higher than the preset matching threshold are selected as spatial memory anchor points for the target complex building; Based on the architectural structural connectivity layout of the target integrated building, the spatial topological relationship between each spatial memory anchor point is analyzed, and a spatial memory anchor point network is constructed with spatial memory anchor points as nodes and spatial topological relationships as connections.

[0024] Based on the interior floor plan and business layout of the target complex, the commercial space is functionally divided into retail and service areas, with each category further subdivided into multiple point units. Retail areas are defined by the storefront locations, while service areas are marked by visually distinctive nodes such as elevator lobbies, service counters, and restroom entrances. After point division, visual feature data of the target complex is obtained at each designated point using the consistent visual acquisition process in S1. The geometric shape, facade color, and scale data of each point are compared item by item with a consensus-based spatial memory feature set. During the comparison, the matching degree index is determined by calculating the similarity of feature dimensions, and shape type, main color parameters, and size proportions are jointly encoded into multi-dimensional visual feature vectors. Then, a full similarity calculation is performed between these point feature vectors and reference vectors in the consensus-based spatial memory feature set. The similarity is represented by the spatial distance between feature vectors; a higher similarity value indicates that the point is closer to a visual anchor point in the general public's memory. A similarity threshold of 0.8 was set. When the similarity between a point and any reference vector exceeded this value, the point was identified as a spatial memory anchor. The selected spatial memory anchors formed stable visual recognition nodes within the building, providing a precise set of nodes for the subsequent construction of the spatial memory anchor network.

[0025] Based on the architectural design drawings and 3D structural model, the spatial location of selected spatial memory anchor points is analyzed, and the accessibility relationships between anchor points are calculated. Connectivity paths between anchor points are established using walkways, escalators, elevators, and ventilation shafts on the floor plan as connection criteria. Connectivity is determined by walkability; two anchor points are considered directly connected when there are no obstructions between them. Inter-floor anchor points are connected via elevators, escalators, or stairs. An anchor point association matrix is ​​formed by organizing all connected pairs between anchor points, and a topological network structure is generated based on this matrix. In the network, each anchor point is a node, and each connection is an edge. To enhance the completeness of the structural representation, the topological network is hierarchically divided. Anchor points on the same floor are connected to form local subnets, and vertical connections are established between different floors through inter-floor passages, forming a three-dimensional composite network, namely the spatial memory anchor point network. This network accurately reflects the spatial connectivity structure within the complex building.

[0026] In step S3, the node sequence of the preset evacuation path is marked in the spatial memory anchor network, the topological overlap between the evacuation path nodes and the spatial memory anchors is calculated, and weak visual connection sections are identified. Specifically, this includes: The turning points in the preset emergency evacuation routes are extracted, integrated into a path node sequence, and marked in the spatial memory anchor network; Calculate the spatial distance from each path node to the nearest spatial memory anchor point, and convert the spatial distance into the topological overlap of each path node; The topological overlap of each path node is connected according to the spatial order of the preset evacuation path to form a path matching curve. The fluctuations in the path alignment curve are retrieved, and path segments in which the topological overlap is consistently below a set alignment threshold are identified and marked as segments with weak visual connections.

[0027] In the architectural structural model of the target complex building, an emergency evacuation route that has passed safety review is selected as the preset route. This route is set along the main passageway, escalator connection area, and exit direction, with clear continuity and accessibility. All turning points are extracted through path geometric data analysis, including polyline nodes with a direction change angle exceeding 30 degrees and vertical conversion nodes for inter-floor passage. These points are numbered sequentially to form a path node sequence. The spatial coordinates of each node are determined by the building plan coordinate system and floor identification. To ensure the spatial correspondence between the path and the spatial memory anchor point network, existing spatial memory anchor point coordinates are loaded into the building information model, and spatial overlay analysis is used to determine whether the path node is within the anchor point's neighborhood. The neighborhood range is set based on the width of the internal passageway of the complex building, with a default spatial search radius of 20 meters. When the distance between the path node and the anchor point is less than this radius, the corresponding relationship of the node is marked in the anchor point network; if there is no corresponding anchor point, it is marked as a non-anchor point node.

[0028] After the path node sequence is labeled, the 3D spatial distance from each node to its nearest spatial memory anchor is calculated sequentially. Distance calculation is based on the coordinate information of the building structure data, including both horizontal and vertical distances. When the path node and anchor are on the same floor, the planar distance is used directly; when they are on different floors, the 3D distance is calculated by considering the floor height difference. To ensure comparability of commercial spaces at different scales, all spatial distance values ​​are standardized and mapped to a continuous interval of 0 to 1. Based on the standardization results, topological overlap is defined as a numerical quantification index that is inversely proportional to distance; the shorter the spatial distance, the higher the topological overlap, indicating that the path node and memory anchor are closer in spatial layout and visual perception.

[0029] The obtained path node topological overlap is arranged according to the order of the nodes in the evacuation route, and a path consistency curve is generated through linear connection. The horizontal axis of the curve represents the spatial sequence position of the path nodes, and the vertical axis represents the corresponding topological overlap. The path consistency curve can intuitively reflect the visual coverage continuity of the path in the spatial memory anchor network. By observing the fluctuation pattern of the curve, sections with obvious declines or persistent low values ​​are identified. To ensure the consistency of the identification results, a consistency threshold of 0.5 is set. When the overlap of five consecutive nodes on the curve is lower than the consistency threshold, the path segment is automatically identified as a visually weak segment. This segment usually corresponds to a location in the building space with insufficient visual guidance or complex structural transitions, such as closed passages, unmarked corners, or multiple exit junctions. After identifying the weak segment, its start and end node positions are spatially labeled, and the segment length and floor information are recorded for use in the deployment of mini fire stations and visual guidance enhancement. In this way, the process from evacuation route structural data to visual correlation quantification and weak segment location is completed.

[0030] In step S4, mini fire stations are deployed along the preset evacuation route in areas with weak visual connections as new memory anchor points, connecting the various mini fire stations to form an evacuation guidance route, specifically including: Mini fire stations are deployed along the preset evacuation route in areas with weak visual connections as new memory anchor points. The initial layout density of the mini fire stations has a negative linear relationship with the visual connection strength value corresponding to the areas with weak visual connections. Establish a spatial sequence among mini fire stations, form a progressive point layout according to the evacuation direction, connect each mini fire station to construct an evacuation guidance path that coincides with the preset evacuation path, and update the spatial memory anchor point network.

[0031] After identifying weak visual connection zones, the spatial extent, number of nodes, and topological length of each weak zone are first quantitatively analyzed. The visual connection strength value of the weak zone is determined based on the mean of the topological overlap sequence, ranging from 0 to 1, with lower values ​​indicating poorer visual continuity. The deployment density of mini fire stations is set according to different intensity ranges to ensure that areas with insufficient visual guidance are compensated for with higher density. In practice, the length and number of nodes of the weak zone are first determined, and a density adjustment coefficient is automatically generated based on the visual connection strength value. This coefficient varies between 0 and 1 and is used to control the deployment spacing. When the visual connection strength is high, the deployment spacing increases accordingly; when the visual connection strength is low, the spacing automatically decreases, making the fire stations more densely deployed. To avoid visual redundancy or resource waste, the minimum deployment spacing is set to be no less than twice the width of the passage, and the maximum spacing is set to be no more than one-tenth of the length of the passage. The specific locations of the deployment points are selected at nodes with visibility on both sides of the evacuation route to ensure uniform and continuous visual coverage. Taking into account the internal passageway space, ceiling height, and power supply conditions of the complex building, priority was given to selecting corners, intersections, and line-of-sight turning points as deployment locations. Each mini fire station is equipped with independent lighting signals and directional indicators, and its exterior adopts a unified color and form design to ensure that it is visually distinct from surrounding anchor points. After deployment, the spatial coordinates of each new fire station are marked, and its location is written into the anchor point database as a new memory anchor point, thereby locally expanding and visually reinforcing the original spatial memory anchor point network.

[0032] After deploying mini fire stations in all vulnerable sections, these newly added anchor points are sequentially associated. Using the exit direction of the evacuation route as the sequence direction, the fire stations are ordered by spatial coordinates, forming a progressive spatial sequence along the evacuation flow. The sequence is established following the principle of continuous passage; when there is spatial obstruction between adjacent fire stations, the order is automatically adjusted or intermediate auxiliary nodes are inserted to ensure the guidance path is continuous without breaks. Subsequently, a spatial connection algorithm connects all fire stations as path segments, generating a guidance path model consistent with the preset evacuation path geometry. The guidance path spatially covers the main passageway of the original evacuation route, and fire stations replace or supplement some of the original anchor points at node positions, structurally strengthening the visual guidance chain. After the path construction is completed, a consistency check is performed on the updated anchor point network to ensure that the new nodes maintain correct connectivity and topological integrity with the original network. The final spatial memory anchor point network consists of original anchor point nodes and newly added fire station nodes, presenting a coherent spatial path formed by interconnected highly recognizable nodes.

[0033] In step S5, fire drill data is extracted, and the pedestrian flow turning rate around each mini fire station is statistically analyzed to calculate the guidance utility coefficient. This guidance utility coefficient is then transformed into a dynamic conflict weight between evacuation and rescue, specifically including: During fire drills, data on the movement trajectories of personnel at mini fire stations are extracted, and the ratio of the number of people moving along the evacuation guidance path at each mini fire station to the total number of people is calculated as the turning rate of the flow of people. By combining the visual connection strength values ​​of the locations of each mini fire station with the pedestrian turning rate and the visual connection strength values, the guidance effectiveness coefficient of each mini fire station is obtained by weighting the pedestrian turning rate with the visual connection strength values. The guidance utility coefficient of each mini fire station is normalized and weighted to serve as the dynamic conflict weight for evacuation and rescue routes. The visual connection strength value is the average topological overlap of all path nodes within the corresponding visually weak segment.

[0034] In fire drills at comprehensive buildings, the movement trajectories of participants are continuously recorded to capture changes in their positions during evacuation. All movement trajectory data is stored as a time series and spatially matched with the coordinates of each mini fire station node. When a person's trajectory passes a mini fire station node, their direction of movement is identified to determine whether they should continue along the evacuation guidance path. Subsequently, the number of people passing through each mini fire station node along the evacuation guidance path is counted, and the ratio of this number to the total number of people passing through that node is calculated as the pedestrian turning rate. The pedestrian turning rate reflects the actual degree to which the node guides the directional behavior of the crowd during evacuation. After obtaining the pedestrian turning rate for each mini fire station, the visual connection strength value of its corresponding path segment is extracted. To comprehensively evaluate the effectiveness of mini fire stations under different visual conditions, the pedestrian turning rate and visual connection strength value are weighted and fused to form a guidance utility coefficient. Two adjustment coefficients are introduced in the weighted calculation: directional response weight and visual perception weight, the sum of which is fixed at 1. The specific weights are determined based on the building structure of the commercial complex and the specifications and appearance of the mini fire stations. The directional response weight is set slightly higher than the visual perception weight, making the actual behavior of the crowd more prominent in the results. Through weighted calculations, each mini fire station obtains a stable guidance utility coefficient, which comprehensively reflects its directional guidance efficiency and the degree of environmental visual enhancement in the evacuation route. After obtaining the guidance utility coefficients of each mini fire station, all guidance utility coefficients are normalized to ensure a unified measurement standard across different route segments. The normalized value is defined as the dynamic conflict weight, used to represent the degree of influence of this node on the passage priority in rescue route planning. The dynamic conflict weight serves as the path cost in rescue route calculation; a higher weight indicates potential congestion and evacuation conflicts at this node during rescue. When multiple evacuation routes occur concurrently, the dynamic conflict weight is used to allocate the passage priority order of the rescue priority passage.

[0035] In step S6, when a fire incident occurs, the location of the incident is determined in the spatial memory anchor network. With the goal of minimizing the cumulative dynamic conflict weight, a rescue path search is performed within the topology of the spatial memory anchor network. Specifically, this includes: When a fire occurs, the exact location of the incident is located in the spatial memory anchor network, and that location is set as the target point for the rescue path search. Starting from the entrance of the complex building, the optimal rescue path is determined by searching the topology of the spatial memory anchor network for the connected path with the minimum cumulative dynamic conflict weight.

[0036] When any fire incident occurs within a complex building, fire signals are acquired through multi-sensor devices, generating event location data containing time, coordinates, and floor information. Based on the event coordinates, spatial mapping is performed on the building's 3D model to identify the spatial unit and floor to which the coordinate point belongs. Subsequently, a position comparison is performed within the existing spatial memory anchor point network, selecting the anchor node closest to the event coordinates as the location point. If the event location lies on a connecting edge between two anchor nodes, the spatial interpolated coordinates of that point are calculated based on the distance ratio between the nodes, and this is added as a new temporary node to the anchor point network to ensure the event point's addressability within the topology. To improve positioning accuracy, the boundary of the event area is corrected using real-time image recognition results. When the event area spans multiple anchor nodes, the central anchor point is designated as the primary target node. This primary target node is defined as the endpoint of the rescue path in subsequent path searches. This method ensures a correspondence between the spatial location of the fire incident and the anchor point network, guaranteeing the uniqueness and traceability of the input location for rescue calculations within the topology.

[0037] After the incident location is determined, the main entrance of the complex building or the entrance to the fire escape route is used as the starting point for the path search, and the aforementioned target node is used as the ending point. The path search is performed within the topology of the spatial memory anchor network. Each edge in the network is assigned a dynamic conflict weight, derived from the normalized result of the guidance utility coefficient formed during the preceding fire drills. This weight represents the degree of influence of each anchor point and path segment on traffic efficiency during evacuation and rescue. The search process aims to minimize the cumulative dynamic conflict weight, traversing all reachable paths in the network, accumulating the weight values ​​of nodes and edges for each path, and recording the results. To avoid excessively long paths or cyclical searches, a path length constraint is set to ensure that the total length of candidate paths does not exceed three times the straight-line distance. The path with the smallest cumulative weight value among all feasible paths is selected as the optimal rescue path. This path represents the route that can reach the incident location with minimal conflict cost under the current building layout and pedestrian distribution. After the path is determined, the location information of all nodes in the path is mapped back to the building floor plan and marked as continuous path lines to indicate the route of the rescue team. Simultaneously, the anchor node numbers and their corresponding spatial coordinates involved in the path are recorded as the basis for dispatch instructions, realizing a closed-loop process from spatial identification to rescue decision-making. Through this path search and selection process, the accessibility and conflict minimization characteristics of the rescue channel in complex building structures are ensured, improving the accuracy and timeliness of emergency rescue execution.

[0038] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0039] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An emergency evacuation design and rescue method for a large-scale comprehensive building, characterized in that, Includes the following steps: S1. Analyze the movement trajectory data of people in multiple complex buildings, identify high-frequency natural dwelling areas, perform visual feature analysis on the spatial landmarks of the dwelling areas, and extract consensus spatial memory features; S2. Based on consensus-based spatial memory features, match specific locations within the target complex building, and construct a spatial memory anchor point network in conjunction with the building structure of the target complex building; S3. Mark the node sequence of the preset evacuation path in the spatial memory anchor network, calculate the topological overlap between the evacuation path nodes and the spatial memory anchors, and identify the weak visual connection sections. S4. Deploy mini fire stations along the preset evacuation route in areas with weak visual connections as new memory anchor points, and connect the mini fire stations to form an evacuation guidance route. S5. Extract fire drill data, calculate the guidance utility coefficient by statistically analyzing the pedestrian flow turning rate around each mini fire station, and convert the guidance utility coefficient into a dynamic conflict weight between evacuation and rescue. S6. When a fire incident occurs, locate the location of the incident in the spatial memory anchor network, and search for a rescue path in the topology of the spatial memory anchor network with the goal of minimizing the cumulative dynamic conflict weight. In step S5, fire drill data is extracted, and the pedestrian flow turning rate around each mini fire station is statistically analyzed to calculate the guidance utility coefficient. The guidance utility coefficient is then converted into a dynamic conflict weight for evacuation and rescue, specifically including: During fire drills, data on the movement trajectories of personnel at mini fire stations are extracted, and the ratio of the number of people moving along the evacuation guidance path at each mini fire station to the total number of people is calculated as the turning rate of the flow of people. By combining the visual connection strength values ​​of the locations of each mini fire station with the pedestrian turning rate and the visual connection strength values, the guidance effectiveness coefficient of each mini fire station is obtained by weighting the pedestrian turning rate with the visual connection strength values. The guidance utility coefficient of each mini fire station is normalized and weighted to serve as the dynamic conflict weight for evacuation and rescue routes. The visual connection strength value is the average topological overlap of all path nodes within the corresponding visually weak segment.

2. The emergency evacuation design and rescue method for a large-scale comprehensive building according to claim 1, characterized in that, In step S1, the analysis of crowd movement trajectory data within multiple integrated buildings, identification of high-frequency natural dwelling areas, spatial marker clustering of these dwelling areas, and extraction of consensus-based spatial memory features specifically include: Collect pedestrian movement trajectory data in multiple complex buildings, extract natural rest points from the movement trajectories, and use density-based spatial clustering method to identify high-frequency clustering areas for natural rest points; Analyze the visual characteristics of landmarks in high-frequency clustering areas and extract visual feature data in three dimensions: geometric shape, facade color, and scale. By comparing visual feature data of high-frequency clusters of multiple complex buildings across projects, repeating visual feature combination patterns are selected to construct a consensus-based spatial memory feature set.

3. The emergency evacuation design and rescue method for a large-scale comprehensive building according to claim 1, characterized in that, In step S2, the specific location within the target complex building is matched based on consensus-based spatial memory features, and the spatial memory anchor point network is constructed in conjunction with the building structure of the target complex building, specifically including: Select a target complex building, divide the interior of the complex building into points based on the shops and supporting service points, and extract the actual visual feature data of the landmarks at each point; The visual feature data of each location marker is matched with the consensus spatial memory feature set to calculate the matching degree index between each location and the consensus spatial memory feature. Specific points with matching scores higher than the preset matching threshold are selected as spatial memory anchor points for the target complex building; Based on the architectural structural connectivity layout of the target integrated building, the spatial topological relationship between each spatial memory anchor point is analyzed, and a spatial memory anchor point network is constructed with spatial memory anchor points as nodes and spatial topological relationships as connections.

4. The emergency evacuation design and rescue method for a large-scale comprehensive building according to claim 1, characterized in that, In step S3, marking the node sequence of the preset evacuation path in the spatial memory anchor network, calculating the topological overlap between the evacuation path nodes and the spatial memory anchors, and identifying visually weak segments specifically include: The turning points in the preset emergency evacuation routes are extracted, integrated into a path node sequence, and marked in the spatial memory anchor network; Calculate the spatial distance from each path node to the nearest spatial memory anchor point, and convert the spatial distance into the topological overlap of each path node; The topological overlap of each path node is connected according to the spatial order of the preset evacuation path to form a path matching curve. The fluctuations in the path alignment curve are retrieved, and path segments in which the topological overlap is consistently below a set alignment threshold are identified and marked as segments with weak visual connections.

5. The emergency evacuation design and rescue method for a large-scale comprehensive building according to claim 1, characterized in that, In step S4, deploying mini fire stations along the preset evacuation route in areas with weak visual connections serves as new memory anchor points, and connecting these mini fire stations to form an evacuation guidance route specifically includes: Mini fire stations are deployed along the preset evacuation route in areas with weak visual connections as new memory anchor points. The initial layout density of the mini fire stations has a negative linear relationship with the visual connection strength value corresponding to the areas with weak visual connections. Establish a spatial sequence among mini fire stations, form a progressive point layout according to the evacuation direction, connect each mini fire station to construct an evacuation guidance path that coincides with the preset evacuation path, and update the spatial memory anchor point network.

6. The emergency evacuation design and rescue method for a large-scale comprehensive building according to claim 1, characterized in that, In step S6, when a fire incident occurs, the location of the incident is determined in the spatial memory anchor network. The search for a rescue path within the topology of the spatial memory anchor network, with the goal of minimizing the accumulated dynamic conflict weight, specifically includes: When a fire occurs, the exact location of the incident is located in the spatial memory anchor network, and that location is set as the target point for the rescue path search. Starting from the entrance of the complex building, the optimal rescue path is determined by searching the topology of the spatial memory anchor network for the connected path with the minimum cumulative dynamic conflict weight.

Images