A subway multi-line transfer station flood multi-objective dynamic evacuation path planning method

Abstract

This invention discloses a multi-objective dynamic evacuation path planning method for floods at multi-line transfer stations in subways. The method includes: establishing a three-dimensional spatial model by acquiring subway station structural information and abstracting it into a network representation; acquiring real-time hydrological information using hydrological sensors and video surveillance, preprocessing and fusing it to generate hydrological perception data under a unified spatiotemporal reference; acquiring real-time crowd information within the station through video target detection, and calculating crowd density and passage capacity to form crowd flow situation perception data; combining hydrological conditions, crowd flow situation, and passability constraints, using a real-time path search algorithm to generate multiple candidate evacuation paths for different types of passengers; classifying evacuation paths according to passenger type, and issuing evacuation guidance through multi-terminal channels to achieve differentiated dynamic guidance for exiting passengers, transferring passengers, and passengers with mobility impairments. This method can respond in real-time to flood disasters and changes in crowd flow, improving the evacuation efficiency and safety of subway transfer stations.

Description

Technical Field

[0001] This invention belongs to the field of emergency evacuation technology for underground transportation facilities, specifically relating to a method for planning dynamic evacuation routes for floods at multi-line transfer stations of subways. Background Technology

[0002] As a vital public transportation infrastructure in cities, subway stations typically feature multiple levels, platforms, transfer passages, and complex vertical transportation components (stairs, escalators, elevators), resulting in dense crowds and strong spatial coupling. In the event of flooding or water intrusion, local passageways may be obstructed, visibility limited, people stranded and congestion exacerbated, and even secondary disasters such as stampedes may occur, seriously threatening passenger safety and the city's emergency response capabilities.

[0003] Existing subway station evacuation systems largely rely on static evacuation maps, pre-set routes, and centralized broadcast guidance. Their planning and guidance strategies are typically based on the evacuation logic of traditional emergencies such as fires, focusing on single indicators such as the shortest path or fewest turns, lacking real-time perception and integration of the unique physical characteristics and evolution trends of floods. Furthermore, current practices often use static or artificially assigned capacity values ​​for passageway availability assessments, failing to reflect sudden changes in capacity caused by flood intrusion, pump room / gate status, and backflow from municipal pipelines. Consequently, in actual flood situations, it is difficult to guarantee the feasibility and safety of evacuation routes.

[0004] In the prior art, Chinese patent CN116187608B discloses a method, system, and equipment for deciding evacuation routes of underground transportation facilities under flood conditions, including the following steps: S1: Using a BP neural network algorithm to establish a node passage time prediction model, calculating the time for passengers to pass through turnstiles, stairs, and / or escalators; S2: Based on the time prediction model described in S1, using an objective function to calculate the total evacuation time, total evacuation risk, and total congestion of passengers under flood conditions, constructing a multi-objective optimization model for passenger evacuation routes, and determining recommended evacuation routes for passengers; S3: Using simulation software that realistically simulates pedestrian movement to build an evacuation simulation system for underground transportation facilities under flood conditions, integrating the multi-objective optimization model for passenger evacuation routes described in S2, simulating passenger evacuation behavior, and evaluating the safety of the evacuation optimization strategy.

[0005] However, this method has the following limitations: 1. The prediction of node passage time and path cost mainly relies on offline-trained time prediction models and simulation parameters, making it difficult to reflect the rapid evolution and sudden changes in the water situation within the station in a timely manner. In the event of short-term heavy rainfall or sudden backflow in the pipeline network, water depth, water surface coverage, and local flow velocity may change abruptly. Time prediction models based on historical training or static assumptions cannot guarantee prediction accuracy and timeliness, resulting in inaccurate or infeasible path selection in real-world scenarios. 2. Although path cost and multi-objective optimization consider overall evacuation time, risk, and congestion, the cost construction lacks direct coupling with real-time water conditions from multiple sources and fine-grained passenger flow data, making it difficult to avoid high-risk areas in a timely manner when channels are flooded or congestion occurs. 3. This method does not perform grouped modeling and differentiated cost assignment for the differentiated needs of various passenger types. Path recommendation mainly focuses on overall objective optimization and lacks a mechanism for issuing grouped evacuation plans according to passenger categories. As a result, in mixed scenarios with multiple types of passengers, it is easy to cause path overlap, local congestion, and allocation conflicts, reducing the actual evacuation accuracy and efficiency.

[0006] In summary, existing publicly available methods still have limitations in rapidly evolving flood scenarios, including insufficient ability to perceive multi-source real-time water conditions and fine-grained crowd flow, inaccurate assessment of passage feasibility and remaining safety time, lack of differentiated path allocation for different groups, and an incomplete closed loop between simulation and on-site execution. These limitations restrict their ability to ensure passenger safety and achieve efficient evacuation in real flood emergencies. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a method for multi-objective dynamic evacuation path planning for floods at subway multi-line transfer stations.

[0008] The objective of this invention can be achieved through the following technical solutions:

[0009] This invention provides a method for multi-objective dynamic evacuation path planning for floods at multiple subway line transfer stations, comprising the following steps:

[0010] Obtain the structural information of the target subway station, and establish a three-dimensional spatial model of the station based on the structural information, and abstract the three-dimensional spatial model into a network representation;

[0011] Real-time hydrological information is obtained through hydrological sensors and video monitoring.

[0012] Hydrological information is preprocessed and fused to obtain hydrological perception data under a unified spatiotemporal reference; the hydrological perception data includes water depth, water flow velocity vector, and water coverage area.

[0013] By detecting objects in the video, real-time crowd information within the station is obtained, and real-time crowd flow situation awareness data is generated based on the crowd information; the crowd flow situation awareness data includes crowd density and passage capacity.

[0014] Combining water sentiment perception data, pedestrian flow situation perception data, and preset accessibility constraints, differentiated access costs are set for each grid cell in the three-dimensional spatial model for different types of passengers. Based on the starting point, ending point, and differentiated access costs of various types of passengers, the A* algorithm is used in conjunction with the horizontal and vertical connected edge sets of the three-dimensional spatial model to perform real-time path search and generate multiple candidate schemes for group evacuation routes.

[0015] Multiple candidate evacuation routes are categorized by passenger type and corresponding evacuation instructions are issued through various terminal channels.

[0016] Furthermore, the structural information of the target subway station includes the location, geometric dimensions, and elevation information of the station hall, platform, transfer passage, stairs, elevator, turnstile, entrance / exit, maintenance passage, pipe trench, and pump room.

[0017] Furthermore, the step of establishing a three-dimensional spatial model within the station based on the structural information, and abstracting the three-dimensional spatial model into a network representation, specifically includes:

[0018] Based on the structural information, at a resolution on the horizontal plane. Discretize the space within the station to form a grid set:

[0019]

[0020] in, The set of grid cells obtained after discretizing the space represents the total number of grid cells. Indicates the first n Each unit of yuan, n Indicates the total number of cells;

[0021] Each cell Has a plane coordinate center Elevation value and area:

[0022]

[0023] in, These represent the spatial resolution of the grid in two horizontal directions, determined based on the channel width and facility density. Representing lattice elements The area;

[0024] Based on structural information, in the grid set G Establish horizontal domain set and vertically connected sets , used to describe connectivity in the horizontal and vertical directions:

[0025]

[0026]

[0027] in, Representing lattice elements The set of horizontal domains; Representing lattice elements A vertically connected set;

[0028] Based on grid set G and its horizontal domain set and vertically connected sets Construct a three-dimensional spatial model within the station. M , is represented as:

[0029]

[0030] in, It is the set of horizontally connected edges, including the set of horizontal neighborhoods of all lattice elements; It is a vertically connected edge set, which includes the vertically connected set of all lattice elements.

[0031] Furthermore, the hydrological information includes the water depth of each grid cell. With image frame ;

[0032] The acquisition of real-time hydrological information through hydrological sensors and video monitoring specifically includes:

[0033] The water depth of the corresponding grid cell in the three-dimensional spatial model is obtained by hydrological sensors deployed in each grid cell within the station. ;

[0034] Image frames covering the grid were acquired using the station's video surveillance equipment. .

[0035] Furthermore, the preprocessing and fusion of hydrological information to obtain hydrological perception data under a unified spatiotemporal reference specifically includes:

[0036] The water depth of the corresponding grid cell is obtained using a hydrological sensor. ;

[0037] Image frames obtained through in-station video surveillance Image processing is performed to extract the polygon boundaries of the water surface. and water flow velocity vector ;

[0038] The water sentiment perception data is per cell. water depth Water flow velocity vector With the boundary of the water surface polygon , is represented as:

[0039]

[0040] in, Representing lattice elements At any moment t Water sentiment data Indicates the first i Individual cells; Representing lattice elements At any moment t The water depth; For grid element At any moment t The velocity vector of the water flow, including the magnitude and direction of the velocity; Representing lattice elements At any moment t The polygonal boundary of the water surface is used to represent the current water coverage area;

[0041] The water sentiment data for the entire station is represented as follows:

[0042]

[0043] in, This represents the water sentiment data for the entire station.

[0044] Furthermore, the image frames obtained through in-station video surveillance... Image processing is performed to extract the polygon boundaries of the water surface. and water flow velocity vector Specifically, it includes:

[0045] Image frames are segmented using image segmentation algorithms. The process involves extracting the water surface area, specifically including:

[0046] A lightweight convolutional neural network is used for semantic segmentation of the water surface region to obtain the water surface mask corresponding to each cell. ; Mask the water surface As the boundary of the water surface polygon ;in, Represents the cells obtained through video surveillance. At any moment t Image frames; For grid element At any moment t The water surface mask represents the set of pixels representing the water surface region in the image;

[0047] Utilizing motion information between images, an optical flow estimation algorithm is used to extract the water flow velocity vector for each cell, specifically including:

[0048] For consecutive image frames and Optical flow estimation is performed by calculating the pixel displacement of the water surface region in the image to obtain the velocity information of each pixel. And based on speed information Obtain the water flow velocity vector for each cell. ;

[0049]

[0050] in, For grid element At any moment t The water flow velocity vector; For grid element The corresponding water surface polygon boundary, including lattice elements The pixel set of the mid-water surface region; For pixels p The weights are determined based on the pixel brightness gradient intensity and region connectivity. For pixels p At any moment t The optical flow displacement between consecutive frames is given by the formula:

[0051]

[0052] in, This is the optical flow estimation function, used to calculate the two-dimensional motion vector of a pixel by considering the brightness consistency and local smoothing constraints of two frames of images; This indicates the sampling time interval between two adjacent image frames; , Representing lattice elements At any moment t With time Image frames.

[0053] Furthermore, the real-time acquisition of crowd information within the station through video target detection specifically includes:

[0054] Obtain the image frames corresponding to each cell within the station. ;

[0055] For the image frame The system performs object detection processing to identify the set of pedestrian targets in the image, and counts the number of pedestrians in each cell based on the detection results. Obtain crowd information within the station; the crowd information includes the number of people in each cell. .

[0056] Furthermore, the generation of real-time crowd flow situation awareness data based on the crowd information specifically includes:

[0057] Based on the number of people in each cell row Area of ​​grid cells Calculate pedestrian density :

[0058]

[0059] in, Representing lattice elements At any moment t Crowd density; Representing lattice elements At any moment t The number of pedestrians; Representing lattice elements i The area;

[0060] Based on the passable width of each cell and population density Calculate the passage capacity of each cell. :

[0061]

[0062] in, Representing lattice elements At any moment t Traffic capacity; Representing lattice elements At any moment t The passable width; It is a non-zero constant;

[0063] population density With traffic capacity As real-time pedestrian flow situation awareness data:

[0064]

[0065] in, Indicates the time of the entire station t Human flow situation awareness data.

[0066] Furthermore, by combining water sentiment perception data, pedestrian flow situation perception data, and preset accessibility constraints, differentiated passage costs are set for each cell in the 3D spatial model for different types of passengers. Based on the starting point, ending point, and differentiated passage costs of various types of passengers, the A* algorithm is used in conjunction with the horizontal and vertical connected edge sets of the 3D spatial model to perform real-time path search, generating multiple candidate schemes for group evacuation routes, specifically including:

[0067] Passengers are categorized into exiting passengers, connecting passengers, and passengers with mobility impairments.

[0068] For various types of passengers, based on water sentiment perception data and passenger flow situation perception data, corresponding passage costs are set for each cell in the three-dimensional spatial model;

[0069] Set corresponding starting and ending points for different types of passengers;

[0070] Based on the set start and end points, and the toll cost, the A* algorithm is used to determine the horizontally connected edge set in the constructed 3D spatial model. Vertically connected edge set Perform path search to obtain candidate evacuation routes for various groups of people;

[0071] The A* algorithm is used to determine the horizontally connected edge set in the constructed 3D spatial model. Vertically connected edge set Performing path search specifically includes:

[0072] For each cell Set a value , indicating the distance from the starting point to the cell The total cost is calculated using the following formula:

[0073]

[0074] in, Indicates the distance from the starting point to the cell. At any moment t The value of the substitute; From the starting point to the grid cell The actual cost is expressed as:

[0075]

[0076] in, From the starting point to the grid cell The path, including the set of cells along the path; For grid element The cost of passage;

[0077] For grid element The estimated cost to the destination is calculated using a heuristic estimation method, employing Euclidean distance:

[0078]

[0079] in, Representing lattice elements and the finish line The Euclidean distance between them; This is the maximum walking speed within the station;

[0080] Start grid Add to the open list and calculate the starting cell. The value of the substitute;

[0081] Select the cell with the lowest cost from the open list. and move it to the closed list;

[0082] Grid element The adjacent cells are evaluated, and the adjacent cells include cells. horizontal domain set and vertically connected sets ;

[0083] Calculate the cost of each adjacent cell and add it to the open list; if an adjacent cell is already in the closed list, skip that cell.

[0084] The path search ends when the endpoint cell is added to the closed list;

[0085] Reconstruct the evacuation path based on the path traced back from the endpoint cell to the starting cell;

[0086] For each type of passenger, the evacuation route is calculated using the A* algorithm to obtain multiple candidate evacuation routes for different groups.

[0087] Furthermore, the toll cost is calculated using the following formula:

[0088]

[0089] in, For passenger type k At any moment t In Ge Yuan The cost of passage; passenger types include exiting passengers, connecting passengers, and passengers with mobility impairments. For passengers exiting the station, For connecting passengers, For passengers with mobility impairments; , , , , Representing lattice elements At any moment t Water depth, water velocity, pedestrian density, traffic capacity, and water surface area; water velocity Based on the water flow velocity vector Obtain; water surface area Based on the polygonal boundary of the water surface Sure; Representing lattice elements i The area; , , , These represent the maximum water depth, maximum flow velocity, maximum pedestrian density, and maximum passage capacity, respectively. , , , , Passenger type k The weighting coefficients are different for different types of passengers.

[0090] Compared with the prior art, the present invention has the following advantages:

[0091] (1) In the prior art, the evacuation route planning of subway transfer stations usually relies on static models or traditional route planning algorithms, which cannot cope with sudden flood disasters and dense crowds, nor do they take into account the needs of different types of passengers, resulting in low evacuation efficiency during disasters and even serious safety problems such as stampedes and stranding. This invention can dynamically update evacuation routes in real time by acquiring water information, passenger flow data and other sensor data, and provide differentiated evacuation guidance for different types of passengers (such as passengers with mobility impairments, transfer passengers, etc.), thus solving the safety hazards and evacuation efficiency problems in the prior art.

[0092] (2) Existing path search methods are typically based on static traffic capacity and preset paths, which cannot dynamically adjust evacuation routes in real time according to changes in water conditions and pedestrian flow, nor can they flexibly handle the evacuation needs of people in different environments. This invention dynamically updates the passage cost of grid cells using real-time water condition perception data and pedestrian flow situation perception data, realizing path search based on these real-time data. This method can automatically optimize evacuation route planning according to environmental changes and population needs, avoiding the neglect of sudden changes in traditional methods, and significantly improving the safety and efficiency of evacuation.

[0093] (3) Existing evacuation route planning methods often neglect the special needs of different types of passengers. For example, passengers with mobility impairments may need to avoid crowded areas, while transfer passengers are more concerned with the speed of the route, lacking personalized evacuation strategies. By using the crowd label classification and differentiated passage cost setting technology of this invention, different passage costs are set for different passenger types, which can optimize evacuation routes according to passenger needs. Exiting passengers, transfer passengers, and passengers with mobility impairments will choose the optimal route according to their own needs, avoiding route congestion and suboptimal evacuation results, and significantly improving the accuracy and efficiency of evacuation.

[0094] (4) Existing route planning methods often neglect the actual water surface coverage, especially during floods, where the size and distribution of the water surface area significantly impacts traffic capacity. By introducing polygonal boundaries for the water surface and calculating its area, this invention can more accurately simulate the actual impact of floods on evacuation routes within stations. In areas with large water surfaces, the route planning system can intelligently avoid them, preventing traffic obstructions caused by flooding. This technology solves the problem of existing route planning methods neglecting water surface area, achieving more precise evacuation route planning.

[0095] (5) In existing route planning methods, routes are usually selected based on static capacity values, ignoring the actual availability of passageways and areas within stations. The dynamic passage cost model in this invention, based on pedestrian density, capacity, and water quality awareness data, allows for dynamic adjustment of routes according to the actual passage conditions of different grid cells. This innovative feature avoids the impact of congested areas and poorly managed passageways on evacuation route planning, ensuring the efficiency and safety of evacuation routes.

[0096] (6) Existing path planning based on A* graph search typically does not incorporate remaining safety time, dynamic passage cost, and horizontal / vertical connectivity into the expansion and pruning strategies. The search results are prone to producing theoretically reachable but time-consuming or potentially dangerous paths. To address this, this invention implements an improved A* search on the horizontal and vertical connected edge sets represented by a 3D mesh: when expanding nodes, the real-time value of the grid cell is used as the basis for expansion and pruning, and differentiated processing is supported for vertical channels (stairs, elevators) and horizontal channels. This enables proactive pruning and rapid re-searching of potentially failed paths, thereby improving search efficiency and ensuring the safe and executable nature of the obtained paths under dynamic flood and pedestrian flow conditions. Attached Figure Description

[0097] Figure 1 This is a flowchart of the dynamic evacuation route planning method according to an embodiment of the present invention;

[0098] Figure 2 This is a model diagram of the dynamic evacuation route planning system according to an embodiment of the present invention. Detailed Implementation

[0099] 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, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0100] Example 1:

[0101] This embodiment provides a method for multi-objective dynamic evacuation path planning for floods at multiple subway line transfer stations, such as... Figure 1 As shown, it includes the following steps:

[0102] Step S1: Obtain the structural information of the target subway station, and build a three-dimensional spatial model of the station based on the structural information, and abstract the three-dimensional spatial model into a network representation;

[0103] The structural information of the target subway station includes the location, geometric dimensions, and elevation information of the station hall, platform, transfer passage, stairs, elevators, turnstiles, entrances and exits, maintenance passages, pipe trenches, and pump room. The elevation information refers to the height.

[0104] Step S1 specifically includes:

[0105] Based on structural information, at a resolution on the horizontal plane. Discretize the space within the station to form a grid set:

[0106]

[0107] in, The set of grid cells obtained after discretizing the space represents the total number of grid cells. Indicates the first n Each unit of yuan, n Indicates the total number of cells;

[0108] Each cell Has a plane coordinate center Elevation value and area:

[0109]

[0110] in, These represent the spatial resolution of the grid in the two horizontal directions, determined based on the channel width and facility density. Representing lattice elements The area;

[0111] Based on structural information, in the grid set G Establish horizontal domain set and vertically connected sets , used to describe connectivity in the horizontal and vertical directions:

[0112]

[0113]

[0114] in, Representing lattice elements The set of horizontal domains; Representing lattice elements A vertically connected set; where lattice elements With Ge Yuan Vertical connectivity refers to two cells being directly connected by stairs, escalators, elevators, shafts, or other vertical transportation components, and structurally enabling passengers to move continuously from one cell to another.

[0115] Based on grid set G and its horizontal domain set and vertically connected sets Construct a three-dimensional spatial model within the station. M , is represented as:

[0116]

[0117] in, It is the set of horizontally connected edges, including the set of horizontal neighborhoods of all lattice elements; It is a vertically connected edge set, which includes the vertically connected set of all lattice elements.

[0118] This step involves refining the structural information of multiple subway line transfer stations, establishing a grid-based discrete 3D spatial model, and realizing the digital and networked representation of the complex spatial structure within the station.

[0119] Traditional evacuation route planning is mostly based on two-dimensional planar schematic diagrams or node connectivity models, which makes it difficult to accurately reflect the spatial relationships between different floors, passages and vertical components in subway transfer stations. This leads to discrepancies between the route search results and the actual passable routes. In particular, in flood disaster scenarios, the water level distribution and passage height differences have a significant impact on traffic capacity, and two-dimensional models cannot support accurate analysis.

[0120] This invention discretizes the station space into a set of grid cells and defines horizontal neighborhoods and vertically connected sets, enabling the model to simultaneously describe the horizontal connectivity and vertical accessibility of the subway space, thus forming a computable and dynamically updatable three-dimensional network structure representation. This structured representation not only supports seamless navigation in multi-level spaces for subsequent path search processes based on the A* algorithm, but also allows for dynamic accessibility calculations for different floors, passageways, and three-dimensional spatial areas by combining water level data and pedestrian flow patterns.

[0121] Step S2: Obtain real-time hydrological information through hydrological sensors and video monitoring;

[0122] The hydrological information includes the water depth of each grid cell. With image frame ;

[0123] Real-time hydrological information is obtained through hydrological sensors and video surveillance, specifically including:

[0124] The water depth of the corresponding grid cell in the three-dimensional spatial model is obtained by hydrological sensors deployed in each grid cell within the station. Hydrological sensors are sensing devices that can monitor water depth, water level changes and water flow status in real time, including but not limited to ultrasonic level gauges, pressure level sensors, radar level sensors or fiber optic water level monitoring devices. They are installed on the ground, walls or drainage ditches of each grid cell and upload the monitoring data to the control system in real time through a communication network to update the water information of the corresponding grid cell in the three-dimensional spatial model.

[0125] Image frames covering the grid were acquired using the station's video surveillance equipment. .

[0126] Step S3: Preprocess and fuse the hydrological information to obtain hydrological perception data under a unified spatiotemporal reference, specifically including:

[0127] The water depth of the corresponding grid cell is obtained using a hydrological sensor. ;

[0128] Water depth observations and image frames obtained through in-station video surveillance Image processing is performed to extract the polygon boundaries of the water surface. and water flow velocity vector Specifically, it includes:

[0129] Image frames are segmented using image segmentation algorithms. The process involves extracting the water surface area, specifically including:

[0130] A lightweight convolutional neural network is used for semantic segmentation of the water surface region to obtain the water surface mask corresponding to each cell. ; Mask the water surface As the boundary of the water surface polygon ;in, Represents the cells obtained through video surveillance. At any moment t Image frames; For grid element At any moment t The water surface mask represents the set of pixels in the water surface region of the image; the lightweight convolutional neural network is a deep learning model that can achieve real-time segmentation with limited computing resources, including but not limited to MobileNet, ShuffleNet, EfficientNet-Lite or a custom lightweight semantic segmentation network based on depthwise separable convolution. Its network structure includes an encoding layer, a feature compression layer and a decoding layer. It introduces a set of water surface feature samples for supervised learning during the training phase to identify the water surface boundary in the image frame and output the corresponding water surface mask.

[0131] Utilizing motion information between images, an optical flow estimation algorithm is used to extract the water flow velocity vector for each cell, specifically including:

[0132] For consecutive image frames and Optical flow estimation is performed by calculating the pixel displacement of the water surface region in the image to obtain the velocity information of each pixel. And based on speed information Obtain the water flow velocity vector for each cell. ;

[0133]

[0134] in, For grid element At any moment t The water flow velocity vector; For grid element The corresponding water surface polygon boundary, including lattice elements The pixel set of the mid-water surface region; For pixels p The weights are determined based on the pixel brightness gradient intensity and region connectivity. For pixels p At any moment t The optical flow displacement between consecutive frames is given by the formula:

[0135]

[0136] in, This is the optical flow estimation function, used to calculate the two-dimensional motion vector of a pixel by considering the brightness consistency and local smoothing constraints of two frames of images; This indicates the sampling time interval between two adjacent frames.

[0137] Water perception data for each cell water depth Water flow velocity With water surface polygon , is represented as:

[0138]

[0139] in, Representing lattice elements At any moment t Water sentiment data Indicates the first i Individual cells; Representing lattice elements At any moment t The water depth; For grid element At any moment tThe velocity vector of the water flow, including the magnitude and direction of the velocity; Representing lattice elements At any moment t The polygonal boundary of the water surface is used to represent the current water coverage area;

[0140] The water sentiment data for the entire station is represented as follows:

[0141]

[0142] in, This represents the water sentiment data for the entire station.

[0143] Step S4: Real-time crowd information within the station is acquired through video object detection, and real-time crowd flow situation awareness data is generated based on this information. Specifically, this includes:

[0144] Obtain the image frames corresponding to each cell within the station. ;

[0145] For image frames The system performs object detection processing to identify the set of pedestrian targets in the image, and counts the number of pedestrians in each cell based on the detection results. To obtain information about the people within the station.

[0146] Based on the information of each grid member Area of ​​grid cells Calculate pedestrian density :

[0147]

[0148] in, Representing lattice elements At any moment t Crowd density; Representing lattice elements At any moment t The number of pedestrians; Representing lattice elements i The area;

[0149] Based on the passable width of each cell and population density Calculate the passage capacity of each cell. :

[0150]

[0151] in, Representing lattice elements At any moment t Traffic capacity; Representing lattice elements At any momentt The passable width; It is a non-zero constant;

[0152] population density With traffic capacity As real-time pedestrian flow situation awareness data:

[0153]

[0154] in, Indicates the time of the entire station t Human flow situation awareness data.

[0155] Step S5: Combining water sentiment perception data, pedestrian flow situation perception data, and preset accessibility constraints, differentiated passage costs are set for each cell in the 3D spatial model for different types of passengers; based on the starting point, ending point, and differentiated passage costs of various types of passengers, the A* algorithm is used in conjunction with the horizontal and vertical connected edge sets of the 3D spatial model to perform real-time path search, generating multiple candidate schemes for group evacuation routes, specifically including:

[0156] Passengers are categorized into exiting passengers, connecting passengers, and passengers with mobility impairments.

[0157] For various types of passengers, based on water sentiment perception data and passenger flow situation perception data, corresponding passage costs are set for each cell in the three-dimensional spatial model;

[0158] Set corresponding starting and ending points for different types of passengers;

[0159] Based on the set start and end points, and the toll cost, the A* algorithm is used to determine the horizontally connected edge set in the constructed 3D spatial model. Vertically connected edge set Perform path search to obtain candidate evacuation routes for various groups of people;

[0160] Using the A* algorithm, based on the horizontally connected edge set in the constructed 3D spatial model... Vertically connected edge set Performing path search specifically includes:

[0161] For each cell Set a value , indicating the distance from the starting point to the cell The total cost is calculated using the following formula:

[0162]

[0163] in, Indicates the distance from the starting point to the cell. At any moment tTotal agency value; From the starting point to the grid cell The actual cost is expressed as:

[0164]

[0165] in, From the starting point to the grid cell The path, including the set of cells along the path; For grid element The cost of passage;

[0166] For grid element The estimated cost to the destination is calculated using a heuristic estimation method, employing Euclidean distance:

[0167]

[0168] in, Representing lattice elements and the finish line The Euclidean distance between them; This is the maximum walking speed within the station;

[0169] Start grid Add to the open list and calculate the starting cell. The value of the substitute;

[0170] Select the cell with the lowest cost from the open list. and move it to the closed list;

[0171] Grid element The adjacent cells are evaluated, and the adjacent cells include cells. horizontal domain set and vertically connected sets ;

[0172] Calculate the cost of each adjacent cell and add it to the open list; if an adjacent cell is already in the closed list, skip that cell.

[0173] The path search ends when the endpoint cell is added to the closed list;

[0174] Reconstruct the evacuation path based on the path traced back from the endpoint cell to the starting cell;

[0175] For each type of passenger, the evacuation route is calculated using the A* algorithm to obtain multiple candidate evacuation routes for different groups.

[0176] The toll cost is calculated using the following formula:

[0177]

[0178] in, For passenger type k At any moment t In Ge Yuan The cost of passage; passenger types include exiting passengers, connecting passengers, and passengers with mobility impairments. For passengers exiting the station, For connecting passengers, For passengers with mobility impairments; , , , , Representing lattice elements At any moment t Water depth, water velocity, pedestrian density, traffic capacity, and water surface area; water velocity Based on the water flow velocity vector Obtain; water surface area Based on the polygonal boundary of the water surface Sure; Representing lattice elements i The area; , , , These represent the maximum water depth, maximum flow velocity, maximum pedestrian density, and maximum passage capacity, respectively. , , , , Passenger type k The weighting coefficients are different for different types of passengers.

[0179] For passengers exiting the station, they may be more concerned about factors such as water depth and current velocity than water conditions and passenger density. Therefore, a weighting coefficient is set to balance these factors, and the weighting coefficient is:

[0180]

[0181] For connecting passengers, travel time and passenger flow are of greater concern; therefore, capacity and passenger density are given higher weights, with the following weighting coefficients:

[0182]

[0183] For passengers with mobility impairments, who are more sensitive to water depth and current velocity, water condition data (such as water depth and current velocity) have a greater impact on passage costs, and therefore have a correspondingly higher weighting coefficient. The weighting coefficient is:

[0184]

[0185] Step S5 aims to dynamically analyze and search the station space and generate candidate evacuation routes for different types of passengers during floods at multi-line transfer stations. This is achieved by comprehensively utilizing real-time water situation data and passenger flow data to dynamically analyze the station space and search for possible routes. By classifying passengers as exiting, transferring, and those with mobility impairments, and setting differentiated travel costs and route search strategies for each group, this step automatically adjusts evacuation routes based on real-time factors such as flood spread, water depth changes, passageway accessibility, and passenger density. This ensures that all types of passengers can obtain the optimal safe evacuation route. The technical effect of this step is to achieve dynamic and intelligent evacuation route planning, significantly improving the safety and efficiency of subway station evacuation during floods and avoiding problems such as fixed routes, delayed response, and crowd congestion in traditional methods.

[0186] Step S6: The multiple candidate evacuation routes are categorized by passenger type and corresponding evacuation instructions are issued through various terminal channels. These terminal channels include:

[0187] The fixed broadcasting system in the subway station is used to broadcast voice evacuation information to all passengers;

[0188] Electronic guidance screens in the station hall, platform and passageway areas are used to display real-time directional guidance for group evacuation routes, emergency avoidance areas and transfer passage status;

[0189] Staff members' mobile terminal devices are used to receive evacuation plans for various passenger types pushed by the system, and to assist in on-site evacuation.

[0190] Passenger mobile devices, including mobile apps, WeChat mini-programs, or subway travel clients, are used to push personalized evacuation routes and guidance information to passengers.

[0191] Example 2:

[0192] This embodiment provides a multi-objective dynamic evacuation path planning system for floods at multiple subway line transfer stations, such as... Figure 2 As shown, it includes:

[0193] The structural information acquisition module is used to acquire the structural information of the target subway station, including the location, geometric dimensions and elevation information of the station hall, platform, transfer passage, stairs, escalators, elevators, turnstiles, entrances and exits, maintenance passages, pipe trenches, pump rooms and other fixed facilities, and to build a three-dimensional spatial model of the station based on the structural information.

[0194] The water sentiment perception module acquires real-time water information through hydrological sensors, video monitoring, and meteorological data interfaces. It also preprocesses and fuses data such as water depth, water flow velocity, and water surface coverage to generate water sentiment perception data under a unified spatiotemporal reference.

[0195] The crowd flow situation awareness module acquires real-time crowd information within the station through video target detection, and calculates the crowd density and passage capacity of each grid cell based on this information, forming real-time crowd flow situation awareness data.

[0196] The passage cost calculation module combines water sentiment perception data, pedestrian flow situation perception data, and preset passability constraints to set differentiated passage costs for different types of passengers (exiting passengers, transfer passengers, and passengers with mobility impairments) in each cell of the three-dimensional spatial model.

[0197] The route planning module, based on the starting point, destination, and differentiated travel costs of various types of passengers, uses the A* algorithm combined with the horizontal and vertical connected edge sets of the 3D spatial model to perform real-time route search and generate multiple candidate schemes for group evacuation routes.

[0198] The evacuation guidance module categorizes multiple candidate evacuation routes according to passenger type and issues corresponding evacuation guidance to passengers through various terminal channels, thereby achieving intelligent and hierarchical on-site evacuation guidance.

[0199] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0200] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A subway multi-line transfer station flood multi-objective dynamic evacuation path planning method, characterized in that, Includes the following steps: Obtain the structural information of the target subway station, and establish a three-dimensional spatial model of the station based on the structural information, and abstract the three-dimensional spatial model into a network representation; Real-time hydrological information is obtained through hydrological sensors and video monitoring. Hydrological information is preprocessed and fused to obtain hydrological perception data under a unified spatiotemporal reference; the hydrological perception data includes water depth, water flow velocity vector, and water coverage area. By detecting objects in the video, real-time crowd information within the station is obtained, and real-time crowd flow situation awareness data is generated based on the crowd information; the crowd flow situation awareness data includes crowd density and passage capacity. Combining water sentiment perception data, pedestrian flow situation perception data, and preset accessibility constraints, differentiated access costs are set for each grid cell in the three-dimensional spatial model for different types of passengers. Based on the starting point, ending point, and differentiated access costs of various types of passengers, the A* algorithm is used in conjunction with the horizontal and vertical connected edge sets of the three-dimensional spatial model to perform real-time path search and generate multiple candidate schemes for group evacuation routes. Multiple candidate evacuation routes are categorized by passenger type and corresponding evacuation instructions are issued through various terminal channels. The method combines water sentiment perception data, pedestrian flow situation perception data, and preset accessibility constraints to set differentiated access costs for different types of passengers in each cell of the three-dimensional spatial model, specifically including: Passengers are categorized into exiting passengers, connecting passengers, and passengers with mobility impairments. For various types of passengers, based on water sentiment perception data and pedestrian flow situation perception data, a corresponding passage cost is set for each cell in the three-dimensional spatial model; the passage cost is calculated using the following formula: wherein, is the passenger type k at time t in the cell ; the passenger type includes an outbound passenger, a transfer passenger, and a passenger with mobility impairment, is the outbound passenger, is the transfer passenger, is the passenger with mobility impairment; , , , , respectively represent the water depth, the water velocity, the crowd density, the passing capacity, and the water surface area of the cell at time t ; the water velocity is obtained according to the water flow velocity vector ; the water surface area is determined according to the water surface polygon boundary ; represents the area of the cell i ; , , , respectively represent the maximum water depth, the maximum flow velocity, the maximum crowd density, and the maximum passing capacity; , , , , are weight coefficients of the passenger types k , wherein the weight coefficients of the passenger types are different.

2. The method for multi-objective dynamic evacuation path planning for floods at multi-line transfer stations in subways according to claim 1, characterized in that, The structural information of the target subway station includes the location, geometric dimensions, and elevation information of the station hall, platform, transfer passage, stairs, elevator, turnstile, entrance / exit, maintenance passage, pipe trench, and pump room.

3. The method for multi-objective dynamic evacuation path planning for floods at multi-line transfer stations in subways according to claim 1, characterized in that, The step of establishing a three-dimensional spatial model within the station based on the structural information, and abstracting the three-dimensional spatial model into a network representation, specifically includes: Based on the structural information, at a resolution on the horizontal plane. Discretize the space within the station to form a grid set: in, The set of grid cells obtained after discretizing the space represents the total number of grid cells. Indicates the first n Each unit of yuan, n Indicates the total number of cells; Each cell Has a plane coordinate center Elevation value and area: in, These represent the spatial resolution of the grid in two horizontal directions, determined based on the channel width and facility density. Representing lattice elements The area; Based on structural information, in the grid set G Establish horizontal domain set and vertically connected sets , used to describe connectivity in the horizontal and vertical directions: in, Representing lattice elements The set of horizontal domains; Representing lattice elements A vertically connected set; Based on grid set G and its horizontal domain set and vertically connected sets Construct a three-dimensional spatial model within the station. M , represented as: in, It is the set of horizontally connected edges, including the set of horizontal neighborhoods of all lattice elements; It is a vertically connected edge set, which includes the vertically connected set of all lattice elements.

4. The method for multi-objective dynamic evacuation path planning for floods at multi-line transfer stations in subways according to claim 1, characterized in that, The hydrological information includes the water depth of each cell. With image frame ; The acquisition of real-time hydrological information through hydrological sensors and video monitoring specifically includes: The water depth of the corresponding grid cell in the three-dimensional spatial model is obtained by hydrological sensors deployed in each grid cell within the station. ; Image frames covering the grid were acquired using the station's video surveillance equipment. .

5. The method for multi-objective dynamic evacuation path planning for floods at multi-line transfer stations in subways according to claim 1, characterized in that, The preprocessing and fusion of hydrological information to obtain hydrological perception data under a unified spatiotemporal reference specifically includes: The water depth of the corresponding grid cell is obtained using a hydrological sensor. ; Image frames obtained through in-station video surveillance Image processing is performed to extract the polygon boundaries of the water surface. and water flow velocity vector ; The water sentiment perception data is per cell. water depth Water flow velocity vector With the boundary of the water surface polygon , represented as: in, Representing lattice elements At any moment t Water sentiment data Indicates the first i Individual cells; Representing lattice elements At any moment t The water depth; For grid element At any moment t The velocity vector of the water flow, including the magnitude and direction of the velocity; Representing lattice elements At any moment t The polygonal boundary of the water surface is used to represent the current water coverage area; The water sentiment data for the entire station is represented as follows: in, This represents the water sentiment data for the entire station.

6. The method for multi-objective dynamic evacuation path planning for floods at multi-line transfer stations in subways according to claim 5, characterized in that, The image frames obtained through station video surveillance Image processing is performed to extract the polygon boundaries of the water surface. and water flow velocity vector Specifically, it includes: Image frames are segmented using image segmentation algorithms. The process involves extracting the water surface area, specifically including: A lightweight convolutional neural network is used for semantic segmentation of the water surface region to obtain the water surface mask corresponding to each cell. ; Mask the water surface As the boundary of the water surface polygon ;in, Represents the cells obtained through video surveillance. At any moment t Image frames; For grid element At any moment t The water surface mask represents the set of pixels representing the water surface region in the image; Utilizing motion information between images, an optical flow estimation algorithm is used to extract the water flow velocity vector for each cell, specifically including: For consecutive image frames and Optical flow estimation is performed by calculating the pixel displacement of the water surface region in the image to obtain the velocity information of each pixel. And based on speed information Obtain the water flow velocity vector for each cell. ; in, For grid element At any moment t The water flow velocity vector; For grid element The corresponding water surface polygon boundary, including lattice elements The pixel set of the mid-water surface region; For pixels p The weights are determined based on the pixel brightness gradient intensity and region connectivity. For pixels p At any moment t The optical flow displacement between consecutive frames is given by the formula: in, This is the optical flow estimation function, used to calculate the two-dimensional motion vector of a pixel by considering the brightness consistency and local smoothing constraints of two frames of images; This indicates the sampling time interval between two adjacent image frames; , Representing lattice elements At any moment t With time Image frames.

7. The method for multi-objective dynamic evacuation path planning for floods at multi-line transfer stations in subways according to claim 1, characterized in that, The method of obtaining real-time crowd information within the station through video target detection specifically includes: Obtain the image frames corresponding to each cell within the station. ; For the image frame The system performs object detection processing to identify the set of pedestrian targets in the image, and counts the number of pedestrians in each cell based on the detection results. Obtain crowd information within the station; the crowd information includes the number of people in each cell. .

8. The method for multi-objective dynamic evacuation path planning for floods at multi-line transfer stations in subways according to claim 1, characterized in that, The generation of real-time crowd flow situation awareness data based on the crowd information specifically includes: Based on the number of people in each cell row Area of ​​grid cells Calculate pedestrian density : in, Representing lattice elements At any moment t Crowd density; Representing lattice elements At any moment t The number of pedestrians; Representing lattice elements i The area; Based on the passable width of each cell and population density Calculate the passage capacity of each cell. : in, Representing lattice elements At any moment t Traffic capacity; Representing lattice elements At any moment t The passable width; It is a non-zero constant; population density With traffic capacity As real-time pedestrian flow situation awareness data: in, Indicates the time of the entire station t Human flow situation awareness data.

9. The method for multi-objective dynamic evacuation path planning for floods at multi-line transfer stations in subways according to claim 1, characterized in that, Based on the origin, destination, and differentiated travel costs of various passenger types, the method utilizes the A* algorithm combined with the horizontal and vertical connected edge sets of a 3D spatial model to perform real-time path search, generating multiple candidate solutions for clustered evacuation routes, specifically including: Set corresponding starting and ending points for different types of passengers; Based on the set start and end points, and the toll cost, the A* algorithm is used to determine the horizontally connected edge set in the constructed 3D spatial model. Vertically connected edge set Perform path search to obtain candidate evacuation routes for various groups of people; The A* algorithm is used to determine the horizontally connected edge set in the constructed 3D spatial model. Vertically connected edge set Performing path search specifically includes: For each cell Set a value , indicating the distance from the starting point to the cell The total cost is calculated using the following formula: in, Indicates the distance from the starting point to the cell. At any moment t The value of the substitute; From the starting point to the grid cell The actual cost is expressed as: in, From the starting point to the grid cell The path, including the set of cells along the path; For grid element The cost of passage; For grid element The estimated cost to the destination is calculated using a heuristic estimation method, employing Euclidean distance: in, Representing lattice elements and the finish line The Euclidean distance between them; This is the maximum walking speed within the station; Start grid Add to the open list and calculate the starting cell. The value of the substitute; Select the cell with the lowest cost from the open list. and move it to the closed list; Grid element The adjacent cells are evaluated, and the adjacent cells include cells. horizontal domain set and vertically connected sets ; Calculate the cost of each adjacent cell and add it to the open list; if an adjacent cell is already in the closed list, skip that cell. The path search ends when the endpoint cell is added to the closed list; Reconstruct the evacuation path based on the path traced back from the endpoint cell to the starting cell; For each type of passenger, the evacuation route is calculated using the A* algorithm to obtain multiple candidate evacuation routes for different groups.

Images