A subway station crowd evacuation path planning method and system based on density detection

Abstract

The present application relates to the technical field of subway station crowd dynamic evacuation, in particular to a subway station crowd evacuation path planning method and system based on density detection. The evacuation path planning method acquires crowd density data of the platform waiting area of the subway transfer station in real time through accurate monitoring, and dynamically divides the exclusive evacuation area of the transfer channel and the regular stair exit based on the data as the core, thereby scientifically and accurately distributing the passenger flow between the two types of exits from the root. This method fully taps the potential utilization value of the transfer channel as an auxiliary evacuation exit, breaks the inherent limitations of the traditional static guidance mode, effectively avoids the evacuation congestion bottleneck problem caused by the excessive aggregation of passenger flow at a single exit, significantly improves the overall efficiency and on-site safety of emergency evacuation in subway transfer stations, and can flexibly adapt to the evacuation needs of different density scenarios such as daily, peak, and extreme passenger flow, with strong adaptability and practicality in actual application.

Description

Technical Field

[0001] This invention relates to the field of dynamic evacuation technology for subway stations, specifically a method and system for planning evacuation routes for subway stations based on density detection. Background Technology

[0002] In recent years, with the rapid expansion of urban subway networks, the inherent characteristics of subway stations, especially transfer stations, such as enclosed spaces, high passenger flow, and complex structures, have led to increasingly prominent issues regarding passenger evacuation safety in emergencies such as fires and sudden surges in passenger volume. Conventional evacuation exits (staircases) in subway stations are mostly located at both ends of the platform. In scenarios involving large passenger flows, areas near the staircases are prone to bottlenecks and congestion, and the congestion boundary continues to shift towards the center of the platform as passenger flow increases. This results in the obstruction of evacuation routes for passengers away from the staircases, significantly reducing overall platform evacuation efficiency and even potentially causing secondary safety risks such as stampedes. To improve platform evacuation efficiency and alleviate congestion, existing technologies mainly offer the following three typical solutions: One is an evacuation scheme based on dynamic signage guidance (such as invention patent CN119721422A): by deploying dynamic signage devices in the station and combining simulation software to conduct real-time simulation analysis of pedestrian flow, the optimal evacuation route adapted to the site is generated, and the direction of signage guidance is dynamically adjusted to optimize personnel distribution and shorten the overall evacuation time.

[0003] Second, there are evacuation schemes based on multi-objective optimization models (such as invention patent CN114298438A): by collecting data on the station's structure, personnel distribution, and hazardous sources, a multi-objective path optimization model is constructed, and the optimal evacuation strategy adapted to multiple hazardous source scenarios is obtained, thereby improving evacuation efficiency and safety.

[0004] Thirdly, there is the evacuation scheme based on dynamic lighting guidance (such as invention patent CN116110175A): the safe passage is determined according to the station structure and disaster source signals, control commands are generated to drive the evacuation indicator lights to adjust the guide arrows, and the guidance effect is verified by the image acquisition device. In case of abnormality, manual intervention and early warning are triggered.

[0005] A comprehensive analysis of the existing technologies reveals that all three solutions focus on maximizing evacuation efficiency within the station through means such as signage optimization, model solving, or lighting control, without altering the layout of conventional evacuation exits. They fail to fully explore the potential of cross-line transfer passages unique to subway transfer stations as auxiliary evacuation exits and lack a coordinated evacuation allocation strategy for transfer passages and conventional staircase exits. Summary of the Invention

[0006] To address the technical problem that existing emergency evacuation technologies in subway stations struggle to dynamically divide platform waiting areas based on real-time crowd density, thus hindering the rational allocation of evacuation passenger flow between transfer passages and regular stair exits and easily leading to platform congestion and reduced overall evacuation efficiency, this invention discloses a crowd evacuation path planning method for subway stations based on density detection. Based on this path planning method, this invention also discloses a crowd evacuation path planning system for subway stations based on density detection.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for planning crowd evacuation routes in subway stations based on density detection includes the following planning steps: At the outset of an emergency evacuation, real-time crowd density data is obtained for the transfer passage and the platform waiting area between it and the nearest regular staircase exit. Based on real-time crowd density data, determine the length of the evacuation zone for the appropriate transfer passage within the current platform waiting area. Based on the length of the evacuation section of the transfer passage, the current platform waiting area is divided into the first evacuation zone and the second evacuation zone, which correspond to the transfer passage and the regular staircase exit, respectively. Passengers in the first evacuation zone are guided to the transfer passage, while passengers in the second evacuation zone are guided to the regular stairwell exit.

[0008] As a further improvement to the above scheme: the length L of the evacuation zone in the transfer passage c The data satisfy a mapping relationship L with the crowd density data ρ in the corresponding platform waiting area. c =F(ρ), the process of establishing this mapping relationship is as follows: Simulation model construction: Based on the classic Moore neighborhood floor field cellular automaton model, an evacuation simulation model consistent with the actual subway transfer station structure, size, and facility layout is constructed; at the same time, evacuation time, crowd density, and density-velocity ratio are defined as the core evaluation indicators of evacuation safety, where the critical danger crowd density is set to 4 ped / m², and the density-velocity ratio is used to characterize the degree of regional congestion. Simulation variables are designed as follows: the crowd density data ρ in the waiting area of ​​the target platform is used as the independent variable, and the evacuation interval length L is used as the independent variable. c The dependent variable; Simulation calculations were conducted: For a separate evacuation scenario in the waiting area of ​​the target platform, different crowd density data ρ gradients were set, covering density intervals under various conditions. Different L values ​​were tested under each density group. c The values ​​corresponding to evacuation time, maximum population density in the area, and congestion duration are selected to filter out the L values ​​that minimize the evacuation time at the target platform, eliminate areas with supercritical density, and minimize congestion duration. c , which is the optimal value under this density; Establish a mapping list: This involves mapping the optimal L values ​​under different population density data ρ and correction variable combinations. c The values ​​are organized to create a mapping list of crowd density and optimal transfer corridor evacuation interval length for real-time retrieval.

[0009] As a further improvement to the above scheme, the process of acquiring real-time crowd density data is as follows: multiple cameras installed at various locations in the waiting area of ​​the platform perform a full-area scan of the current waiting area of ​​the platform, collect platform image information, and then stitch the image information together. The image recognition algorithm is used to identify passenger targets in the stitched complete platform image and count the total number of people in the current waiting area of ​​the platform. The average crowd density is calculated by combining the actual size of the current waiting area of ​​the platform, and this is used as the real-time crowd density data.

[0010] As a further improvement to the above scheme: after determining the length of the evacuation zone, N is used. c =round (L c / s) will be the length of the evacuation zone L c Convert to indicator light threshold N c ,round is the rounding function, and s is the spacing between the evacuation guidance lights on the platform floor.

[0011] As a further improvement to the above scheme, passengers are guided to evacuate to the corresponding exit channels by evacuation guidance lights installed on the ground in the waiting area of ​​the platform. The evacuation guidance lights are installed at equal intervals along the length of the platform, and each evacuation guidance light is equipped with a unique identification ID value.

[0012] As a further improvement to the above solution: based on the indicator light threshold N c The direction of the evacuation guidance indicator lights is controlled according to the following rule: the ID value of the evacuation guidance indicator light is less than N. c At that time, the indicator light points to the transfer passage; the ID value of the evacuation guidance indicator light is greater than N. c At that time, the indicator light points to the regular stairwell exit.

[0013] As a further improvement to the above scheme, the SIFT algorithm is used to stitch the image information. The specific steps are as follows: Gaussian blurring and downsampling are applied to the acquired platform images to construct a Gaussian pyramid, generating a difference-of-Gaussian pyramid and filtering out extreme points. Stable feature points are obtained through sub-pixel interpolation and edge point removal. The magnitude and direction of the gradient in the neighborhood of the feature points are calculated, the direction distribution is statistically analyzed, and the precise main direction is determined by interpolation. The neighborhood of the feature points is divided into sub-regions based on the main direction, and the gradient magnitude of each sub-region is statistically analyzed to form a feature vector. Multiple platform images are stitched together by matching the feature vectors to obtain a complete platform image.

[0014] As a further improvement to the above scheme, the image recognition algorithm adopts the YOLOv11 algorithm. The specific steps are as follows: the backbone network performs hierarchical feature extraction and downsampling processing on the stitched complete platform image to output multi-scale original feature maps; the neck network performs feature fusion and refinement optimization on the multi-scale original feature maps, fusing shallow detail features and deep semantic features to generate multi-scale fused feature maps; the head network performs target detection prediction on the multi-scale fused feature maps, outputting a prediction tensor containing the bounding box coordinates of passenger targets and the target existence confidence, thus completing passenger target recognition and counting the number of people.

[0015] A density-detection-based crowd evacuation path planning system for subway stations is used to execute a density-detection-based crowd evacuation path planning method for subway stations. It includes a platform data acquisition module, a crowd density calculation module, an evacuation interval determination module, an evacuation area division module, and an evacuation guidance execution module that are connected in sequence. The platform data acquisition module includes a multi-camera shooting unit and an image transmission unit installed at various locations in the platform waiting area. The multi-camera shooting unit is used to perform a full-area scan of the platform waiting area and collect platform image information at the beginning of an emergency evacuation event. The image transmission unit is used to transmit the collected platform image information to the outside. The crowd density calculation module includes an image receiving unit, an image processing unit, and a density calculation unit. The image receiving unit receives platform image information. The image processing unit stitches the image information together using the SIFT algorithm to obtain a complete platform image and uses the YOLOv11 algorithm to perform passenger target recognition and number of people counting on the complete platform image. The density calculation unit calculates the average crowd density based on the actual size of the waiting area on the platform, which serves as real-time crowd density data. The evacuation zone determination module includes a mapping storage unit and a zone matching unit. The mapping storage unit pre-stores a mapping list of crowd density and optimal transfer channel evacuation zone lengths, which is obtained through simulation based on the classic Moore neighborhood floor field cellular automata model. The zone matching unit is used to match the appropriate evacuation zone length L for the transfer channel from the mapping list based on real-time crowd density data. c ; The evacuation zone division module is a critical value conversion unit, used to convert the evacuation interval length L... c Convert to indicator light threshold N c ; The evacuation guidance execution module includes an instruction transmission unit, a PLC control unit, and an array of ground evacuation guidance indicator lights evenly spaced along the length of the platform waiting area. Each indicator light in the array is equipped with a unique identification ID value. The instruction transmission unit is used to receive the indicator light threshold value N. c The PLC control unit is used to connect the ID values ​​of each indicator light with N. cCompare and control the indicator light direction; if the ID value is less than N... c The indicator light points to the transfer passage, and the ID value is greater than N. c The indicator lights point to the regular stairwell exits, guiding passengers in different evacuation areas to the corresponding exit channels.

[0016] As a further improvement to the above scheme: the crowd density calculation module is integrated into the DSP digital signal processor, and the mapping storage unit is the built-in memory of the DSP digital signal processor, so as to realize the calculation of real-time crowd density data and the matching of evacuation interval length.

[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. At the outset of an emergency evacuation, real-time crowd density data is accurately acquired for the waiting area on the platform between the transfer passage and the nearest regular staircase exit. This data serves as the core basis for determining the appropriate evacuation zone length for the transfer passage, rather than using a fixed area division method. This achieves dynamic matching between the evacuation zone and real-time passenger flow density, flexibly adapting to evacuation needs in different density scenarios such as daily, peak, and extreme passenger flows. Simultaneously, using the evacuation zone length of the transfer passage as a boundary, the platform waiting area is precisely divided into two evacuation zones corresponding to the transfer passage and the regular staircase exit, respectively. People from different zones are guided to evacuate to their corresponding exits, achieving precise passenger flow diversion between the two types of exits. This fully utilizes the transfer passage as an auxiliary evacuation exit, avoids evacuation bottlenecks caused by excessive crowding at a single exit, and effectively prevents cross-traffic interference caused by blindly following others, significantly reducing the probability of platform congestion. Furthermore, the execution steps are logically clear and the process is concise. From density acquisition to area division and crowd guidance, a complete evacuation decision-making closed loop is formed. No complex algorithm calculations are required, and it is easy to integrate with hardware systems to achieve engineering applications. While improving the overall efficiency of emergency evacuation at subway transfer stations, it significantly enhances the orderliness and safety of the evacuation process. From a macro perspective, it optimizes the allocation of evacuation resources in the waiting area of ​​the platform and solves the core problems of mismatched exit resources and low evacuation efficiency in traditional evacuation methods.

[0018] 2. An evacuation simulation model based on a classic Moore neighborhood cellular automata model of the floor field is constructed, perfectly matching the structure, size, and facility layout of an actual subway transfer station. This model maximizes the replication of the actual platform evacuation scenario while incorporating realistic evacuation rules, avoiding decision-making biases caused by the disconnect between the simulation model and the actual scenario. The simulation design not only clearly defines the three core safety evaluation indicators—evacuation time, crowd density, and density-speed ratio—but also sets a critical dangerous crowd density threshold of 4 ped / m². The absence of "no supercritical density areas" is used as a crucial condition for selecting the optimal evacuation interval length. This approach aims to improve evacuation efficiency while ensuring safety during the evacuation process, fundamentally preventing the risk of stampedes caused by unreasonable interval divisions leading to excessive local crowd density. Furthermore, this method uses crowd density as the independent variable and evacuation interval length as the dependent variable, setting density gradients covering various scenarios for simulation calculations. Under each density group, multiple evaluation indicators corresponding to different interval lengths are tested. By comprehensively selecting from multiple indicators, the optimal interval length with the shortest evacuation time, no supercritical density areas, and minimal congestion duration is determined, ensuring that each density value matches the best-fitting evacuation interval, making dynamic interval division more targeted. Finally, by establishing a crowd density-optimal transfer channel evacuation interval length mapping list for real-time retrieval, the extensive simulation calculation results are solidified. During emergency evacuations, the optimal interval length can be obtained directly based on the real-time detected crowd density, eliminating the need for complex on-site simulation calculations. This satisfies the real-time requirement of second-level response in emergency evacuation scenarios while ensuring the efficiency and accuracy of interval length determination. It provides solid technical support for the implementation of evacuation route planning methods, making the entire decision-making process of evacuation route planning more scientific, quantifiable, and operable. Attached Figure Description

[0019] Figure 1 This is a flowchart of the evacuation route planning method.

[0020] Figure 2 This is a three-dimensional schematic diagram of the platform level.

[0021] Figure 3 This is a schematic diagram of the evacuation zone length allocation system. Detailed Implementation

[0022] 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] This invention discloses a crowd evacuation path planning method and system based on density detection in subway stations. It is applicable to crowd path planning and guidance in emergency evacuation events such as fires, floods, and equipment failures in the platform waiting area between the transfer passage and the nearest conventional stair exit in subway transfer stations. The core of the method is to dynamically divide the appropriate evacuation area by real-time crowd density detection in the platform waiting area and guide passengers to evacuate to the corresponding exit passage, thereby mitigating congestion caused by unreasonable path selection and improving the efficiency and safety of emergency evacuation in subway stations.

[0024] I. Evacuation Route Planning Methods

[0025] like Figure 1 As shown, the evacuation route planning method of the present invention is executed sequentially through four core steps, realizing a closed loop of the entire process from density detection to crowd guidance. The specific implementation process is as follows: 1. Obtain real-time crowd density data in the platform waiting area at the beginning of an emergency evacuation event. Multiple cameras installed at key locations such as the top and pillars of the platform waiting area are used to collect full-area image information of the platform. After image stitching, target recognition, and people counting, the average crowd density is calculated as the real-time crowd density data ρ. The process is carried out in three stages: Image Acquisition: After an emergency evacuation event is triggered, the multi-camera shooting unit immediately performs a full-area, no-dead-angle scan of the waiting area on the platform, acquiring multiple local platform image information. The image transmission unit transmits the acquired image information to the crowd density calculation module in real time. Image stitching: The SIFT algorithm is used to stitch together multiple local images to obtain a complete image of the platform waiting area. The specific steps are as follows: ① Gaussian blurring and downsampling are applied to the acquired local images to construct a Gaussian pyramid, and then a Gaussian difference pyramid is generated by subtracting layers. Extreme points are selected in the neighborhood, and stable feature points are obtained after sub-pixel interpolation and edge point removal; ② The magnitude and direction of the gradient in the neighborhood of the feature points are calculated, the direction distribution is statistically analyzed, and the precise main direction is determined by interpolation based on the interval corresponding to the maximum magnitude; ③ The neighborhood of the feature points is divided into sub-regions based on the main direction, and the gradient magnitude of each sub-region is calculated to form a feature vector; ④ By matching the feature vectors of different local images, the multi-dimensional spatial image is stitched together to obtain a complete platform image. Density Calculation: The YOLOv11 algorithm is used to perform passenger target recognition and number of people counting on the stitched complete platform image. The specific steps are as follows: ① The backbone network performs hierarchical feature extraction and downsampling processing on the complete platform image to mine multi-level feature information from low-level details to high-level semantics, and outputs multi-scale original feature maps; ② The neck network performs feature fusion and refinement optimization on the multi-scale original feature maps, fusing shallow detail features and deep semantic features to generate multi-scale fused feature maps adapted for passenger target detection; ③ The head network performs target detection prediction on the multi-scale fused feature maps, and outputs a prediction tensor containing the bounding box coordinates of passenger targets and the target existence confidence, to complete the accurate identification of passenger targets and count the total number of people in the platform waiting area; Finally, the density calculation unit combines the actual size (area S) of the platform waiting area and calculates the average crowd density using the formula ρ = total number of people / S, which serves as the real-time crowd density data for this evacuation.

[0026] 2. Determine the appropriate evacuation zone length for transfer corridors based on real-time crowd density data.

[0027] This section uses the obtained real-time crowd density data ρ as a basis to retrieve the evacuation interval length L of the appropriate transfer passage from a pre-stored mapping list. c L c It satisfies the mapping relationship L with ρ c =F(ρ), the establishment of this mapping relationship is a preprocessing step in the simulation. During the actual evacuation, only the list retrieval needs to be completed. The specific implementation is divided into two parts: establishing the mapping relationship and real-time interval matching. (1) Establish mapping relationship ①Build a simulation model: Based on the classic Moore neighborhood floor field cellular automata model, construct an evacuation simulation model that is completely consistent with the structure, size, and facility layout of the actual subway transfer station, and incorporate a multi-exit selection mechanism and cross-floor transfer rules; at the same time, define evacuation time, crowd density, and density-speed ratio as the core evaluation indicators of evacuation safety, where the critical danger crowd density is set to 4 ped / m², and the density-speed ratio is used to quantitatively characterize the degree of regional congestion (the larger the ratio, the more severe the congestion).

[0028] ② Design simulation variables: The average crowd density ρ in the waiting area of ​​the target platform is used as the independent variable, and the length L of the evacuation zone in the transfer passage is used as the independent variable. c The dependent variable is , and all other simulation parameters are consistent with the actual station.

[0029] ③ Conduct simulation calculations: For the scenario of separate evacuation from the waiting area of ​​the target platform, set different crowd density data ρ gradients. The gradient range covers the density intervals under various conditions such as daily subway operation, peak passenger flow, and extreme passenger flow. Under each density ρ, test the evacuation time, maximum crowd density in the area, and congestion duration corresponding to multiple different Lc values. Through comprehensive screening of multiple indicators, determine the Lc that minimizes the evacuation time of the target platform, eliminates the supercritical density area, and minimizes the congestion duration as the optimal value under that density ρ.

[0030] ④ Establish a mapping list: Organize the optimal Lc values ​​under different combinations of population density data ρ and correction variables to form a mapping list of population density-optimal transfer channel evacuation interval length. This list is pre-stored in the built-in memory of the DSP digital signal processor for real-time retrieval and retrieval during actual evacuation.

[0031] (2) Real-time interval matching

[0032] During emergency evacuation, the interval matching unit of the evacuation interval determination module inputs the obtained real-time crowd density data ρ into a pre-stored mapping list and directly retrieves the length L of the evacuation interval for the appropriate transfer passage under that density. c The interval length is determined.

[0033] 3. Divide the evacuation zone into a first evacuation zone and a second evacuation zone based on the length of the evacuation interval.

[0034] This section uses the obtained evacuation interval length as a boundary to divide the platform waiting area into a first evacuation zone and a second evacuation zone, corresponding to two transfer passages and a regular staircase exit, respectively. The transfer passage (an evacuation staircase connecting the platform level of another subway line to the concourse level) and the regular staircase exit are located on either side of the platform waiting area, and their spatial layout and evacuation routes are as follows: Figure 2 As shown, L1, L2, and L3 are the first, second, and third underground levels, respectively. The transfer passage is an indirect evacuation exit across subway lines. Passengers can use the stairs on the platform level of another subway line to reach the concourse level via this passage. The regular staircase exit is a direct evacuation exit from the current platform level to the concourse level. These two types of passages form a dual-exit evacuation system for the platform waiting area. To achieve precise control of evacuation guidance, the length L of the evacuation section of the transfer passage is first determined. c Convert to indicator light threshold N c Then N c Based on the combination Figure 3 To accurately delineate evacuation areas, the specific conversion formula is: N c =round(L c / s), round is the rounding function, s is the evenly spaced layout distance of the evacuation guidance indicator lights on the platform floor (in this embodiment, s is 0.8~1.2m, which can be adjusted according to the actual platform width); Nc It is a positive integer, serving as the critical threshold for subsequent evacuation guidance indicator light direction control.

[0035] Figure 3 The image clearly shows that the total length of the waiting area on the platform is L. Evacuation guidance lights are evenly spaced along the entire length of the platform. The ID values ​​of the lights start from 1 and increase sequentially along the platform's length. Figure 3 The numbers 1, 2, ..., N are marked in the text. c -1、N c N c The indicator lights +1, ..., N are numbered one-to-one, and the evacuation zone length L in this section... c ,Right now Figure 3 The length of the dedicated section on the left side of the central platform for evacuation via the transfer passage is the core length basis for dividing the two evacuation zones.

[0036] In this embodiment, the indicator light spacing is selected as s=1m (compliant with the 0.8~1.2m layout specification). The appropriate evacuation zone length L for the transfer passage is obtained based on real-time crowd density. c =32.4m, substituting it into the conversion formula, we get: N c =round(32.4 / 1)=32, which gives the critical value N of the evacuation indicator light. c =32; if the match yields L c =18.7m, then N c =round(18.7 / 1)=19, and so on, to achieve precise conversion from length to indicator light threshold value.

[0037] The converted indicator light threshold N c As the dividing point, combined with Figure 3 The platform layout completes the evacuation zone division: the corresponding indicator light ID value is less than N. c The area is designated as the first evacuation zone, i.e. Figure 3 L c The corresponding area on the left side of the platform has a unique transfer passage that serves as the evacuation exit; the corresponding indicator light ID value is greater than N. c The area is designated as the second evacuation zone, i.e. Figure 3 LL c The remaining area on the right side of the corresponding platform has only one corresponding regular staircase exit as an evacuation exit. The two evacuation zones are separated by N. c The corresponding indicator lights are physically separated, with no overlapping areas and completely covering the entire platform waiting area, and... Figure 3 The spatial distribution of the transfer passage and the regular staircase exits is adapted to achieve precise matching between the platform crowd and the two exit passages.

[0038] 4. Guide passengers in different evacuation areas to the corresponding exit channels.

[0039] This section utilizes an array of evacuation guidance indicator lights deployed on the floor of the platform waiting area to achieve precise crowd guidance. The indicator lights are LED-lit and covered with transparent protective covers, ensuring passengers can clearly see the direction of the lights while protecting the indicator structure from damage caused by people stepping on or bumping into them. The specific implementation process is as follows: Basic layout of indicator lights: Evacuation guidance indicator lights are evenly spaced along the length of the platform waiting area, covering the entire platform waiting area, and each indicator light is equipped with a unique identification ID value (the ID value is a positive integer of 1, 2, 3, ..., N, which increases sequentially along the length of the platform).

[0040] Indicator light directional control: The PLC control unit of the evacuation guidance execution module sets the ID value of each indicator light against the indicator light threshold N. c Perform real-time comparisons and control the indicator light direction according to the following rules: ①ID value is less than N c The evacuation guidance indicator lights have arrows pointing to the transfer passage corresponding to the first evacuation zone.

[0041] ② Evacuation guidance indicator lights with an ID value greater than Nc point to the regular stairwell exit corresponding to the second evacuation zone.

[0042] Crowd guidance: Passengers will follow the arrows on the evacuation guidance indicator lights at their location to evacuate to the corresponding exit passage, thus achieving an orderly flow of people from the first evacuation zone to a transfer passage and people from the second evacuation zone to the regular stairwell exit, avoiding congestion caused by crowding towards a single exit passage.

[0043] II. Evacuation Route Planning System

[0044] The evacuation route planning system of the present invention is used to execute the evacuation route planning method. It includes a platform data acquisition module, a crowd density calculation module, an evacuation interval determination module, an evacuation area division module, and an evacuation guidance execution module, which are sequentially connected in communication. The crowd density calculation module is integrated into a DSP digital signal processor, and the mapping storage unit of the evacuation interval determination module is the built-in memory of the DSP digital signal processor. This enables rapid calculation of real-time crowd density data and rapid matching of evacuation interval lengths. The hardware composition, functions, and working process of each module are as follows: 1. Platform Data Acquisition Module This module serves as the system's data input, comprising a multi-camera shooting unit and an image transmission unit. Its core function is to acquire and transmit image information from the platform waiting area. Multi-camera shooting unit: High-definition industrial cameras are used and installed on the top, pillars and other unobstructed locations in the waiting area of ​​the platform. The number and placement of the cameras are determined according to the actual size of the waiting area of ​​the platform to ensure that the scanning range covers the entire waiting area of ​​the platform without any blind spots. After an emergency evacuation event is triggered, a full-area scan is immediately performed to collect multiple local platform images.

[0045] Image transmission unit: Electrically connected to the multi-camera shooting unit, with a built-in 5G wireless transmission module, it transmits the collected local platform image information to the crowd density calculation module in real time and without delay via 5G wireless signal.

[0046] 2. Crowd density calculation module

[0047] This module is integrated into the DSP digital signal processor and is the core computing unit of the system. It includes an image receiving unit, an image processing unit, and a density calculation unit. Its core functions are to perform image stitching, passenger recognition, people counting, and crowd density calculation. Image receiving unit: It has a built-in 5G wireless receiving module, which communicates with the image transmission unit of the platform data acquisition module to receive local platform image information in real time.

[0048] Image processing unit: Preloaded with SIFT image stitching algorithm and YOLOv11 target recognition algorithm, it first stitches local images to obtain complete platform images, then performs passenger target recognition on the complete platform images and counts the total number of people. The recognition results and the number of people statistics are transmitted to the density calculation unit in real time.

[0049] Density Calculation Unit: Pre-stores the actual size (area S) of the waiting area of ​​the platform. After receiving the total number of people transmitted by the image processing unit, it calculates the average crowd density using the formula ρ = total number of people / S, which is used as the real-time crowd density data ρ. Then, it transmits ρ to the evacuation zone determination module.

[0050] 3. Evacuation Zone Determination Module

[0051] This module is integrated with the crowd density calculation module in the same DSP digital signal processor, including a mapping storage unit and an interval matching unit. Its core function is to complete the matching and output of the evacuation interval length Lc. Mapping storage unit: This is the built-in high-speed memory of the DSP digital signal processor, which pre-stores the crowd density-optimal transfer channel evacuation interval length mapping list established by the above method, supporting fast data retrieval and retrieval.

[0052] Interval matching unit: Receives real-time crowd density data ρ transmitted by the crowd density calculation module, performs precise retrieval in the list of the mapping storage unit based on ρ, obtains the appropriate transfer channel evacuation interval length Lc, and transmits Lc to the evacuation area division module.

[0053] 4. Evacuation Area Division Module

[0054] This module is a threshold conversion unit, which can be a standalone hardware processing module or a subroutine integrated into a DSP digital signal processor. Its core function is to convert the evacuation interval length Lc into the indicator light threshold value Nc. The critical value conversion unit pre-stores the layout spacing s of the evacuation guidance indicator lights on the platform ground. After receiving Lc transmitted by the evacuation section determination module, it completes the calculation according to the formula Nc=round(Lc / s) to obtain the critical value Nc of the indicator lights, and transmits Nc to the evacuation guidance execution module through 5G wireless signal.

[0055] 5. Evacuation Guidance Execution Module

[0056] This module serves as the system's execution output, comprising a command transmission unit, a PLC control unit, and a ground evacuation guidance indicator array. Its core functions include receiving the indicator threshold value Nc, controlling the indicator direction, and guiding the crowd. It is the final carrier for realizing evacuation route planning. Command transmission unit: Built-in 5G wireless receiving module, which communicates with the evacuation area division module, receives the indicator light threshold value Nc in real time, and transmits Nc to the PLC control unit.

[0057] PLC control unit: It is a programmable logic controller that is electrically connected to the instruction transmission unit and the evacuation guidance indicator array. It is preloaded with a logic program that compares the indicator ID value with Nc. After receiving Nc, it immediately compares the ID value of each indicator one by one and sends a pointing control command to each indicator.

[0058] Ground evacuation guidance indicator array: It consists of multiple LED evacuation guidance indicators arranged at equal intervals. Each indicator is equipped with a unique ID value and has a built-in signal receiving module. After receiving the directional control command sent by the PLC control unit, it immediately adjusts the arrow to point to the corresponding exit channel to realize on-site evacuation guidance for passengers.

[0059] The method and system of this invention achieve dynamic path planning and orderly crowd diversion for emergency evacuation from waiting areas on subway transfer stations through fully automated implementation of "density detection - interval matching - area division - light guidance". Compared with the traditional fixed signage guidance evacuation method, this invention dynamically adjusts the evacuation area boundary according to real-time crowd density, adapting to the evacuation needs of different passenger flow scenarios, and effectively mitigating congestion of single exit channels caused by unreasonable crowd following. At the same time, each module of the system achieves fast calculation and data transmission based on 5G wireless communication and DSP digital signal processor, and the delay of the entire process from image acquisition to indicator light adjustment is controlled within seconds, ensuring the real-time nature of emergency evacuation. In addition, the evacuation guidance indicator lights are deployed on the platform ground, matching the height of passengers' line of sight, making the guidance effect more intuitive and further improving the efficiency and safety of crowd evacuation.

[0060] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for planning crowd evacuation routes in subway stations based on density detection, characterized in that, The planning steps include the following: At the outset of an emergency evacuation, real-time crowd density data is obtained for the transfer passage and the platform waiting area between it and the nearest regular staircase exit. Based on real-time crowd density data, determine the length of the evacuation zone for the appropriate transfer passage within the current platform waiting area. Based on the length of the evacuation section of the transfer passage, the current platform waiting area is divided into the first evacuation zone and the second evacuation zone, which correspond to the transfer passage and the regular staircase exit, respectively. Passengers in the first evacuation zone are guided to the transfer passage, while passengers in the second evacuation zone are guided to the regular stairwell exit.

2. The method for planning crowd evacuation routes in a subway station based on density detection according to claim 1, characterized in that, Length L of the evacuation zone of the transfer passage c The data satisfy a mapping relationship L with the crowd density data ρ in the corresponding platform waiting area. c =F(ρ), the process of establishing this mapping relationship is as follows: Simulation model construction: Based on the classic Moore neighborhood floor field cellular automaton model, an evacuation simulation model consistent with the actual subway transfer station structure, size, and facility layout is constructed; at the same time, evacuation time, crowd density, and density-velocity ratio are defined as the core evaluation indicators of evacuation safety, where the critical danger crowd density is set to 4 ped / m², and the density-velocity ratio is used to characterize the degree of regional congestion. Simulation variables are designed as follows: the crowd density data ρ in the waiting area of ​​the target platform is used as the independent variable, and the evacuation interval length L is used as the independent variable. c The dependent variable; Simulation calculations were conducted: For a separate evacuation scenario in the waiting area of ​​the target platform, different crowd density data ρ gradients were set, covering density intervals under various conditions. Different L values ​​were tested under each density group. c The values ​​corresponding to evacuation time, maximum population density in the area, and congestion duration are selected to filter out the L values ​​that minimize the evacuation time at the target platform, eliminate areas with supercritical density, and minimize congestion duration. c , which is the optimal value under this density; Establish a mapping list: This involves mapping the optimal L values ​​under different population density data ρ and correction variable combinations. c The values ​​are organized to create a mapping list of crowd density and optimal transfer corridor evacuation interval length for real-time retrieval.

3. A method for planning crowd evacuation routes in a subway station based on density detection, as described in claim 1 or 2, characterized in that, The process of acquiring real-time crowd density data is as follows: Multiple cameras installed at various locations in the waiting area of ​​the platform perform a full-area scan of the current waiting area of ​​the platform, collect platform image information, and then stitch the image information together. The image recognition algorithm is used to identify passenger targets in the stitched complete platform image and count the total number of people in the current waiting area of ​​the platform. The average crowd density is calculated by combining the actual size of the current waiting area of ​​the platform, which serves as the real-time crowd density data.

4. A method for planning crowd evacuation routes in a subway station based on density detection, as described in claim 1 or 2, characterized in that, After determining the length of the evacuation zone, use N c =round (L c / s) will be the length of the evacuation zone L c Convert to indicator light threshold N c ,round is the rounding function, and s is the spacing between the evacuation guidance lights on the platform floor.

5. A method for planning crowd evacuation routes in a subway station based on density detection, as described in claim 1 or 2, characterized in that, Passengers are guided to the corresponding exit channels by evacuation guidance lights installed on the ground in the waiting area of ​​the platform. The evacuation guidance lights are evenly spaced along the length of the platform, and each evacuation guidance light is equipped with a unique identification ID value.

6. The method for planning crowd evacuation routes in a subway station based on density detection according to claim 4, characterized in that, Based on the indicator light threshold N c The direction of the evacuation guidance indicator lights is controlled according to the following rule: the ID value of the evacuation guidance indicator light is less than N. c At that time, the indicator light points to the transfer passage; the ID value of the evacuation guidance indicator light is greater than N. c At that time, the indicator light points to the regular stairwell exit.

7. The method for planning crowd evacuation routes in a subway station based on density detection according to claim 3, characterized in that, The SIFT algorithm is used to stitch together image information. The specific steps are as follows: Gaussian blurring and downsampling are applied to the acquired platform images to construct a Gaussian pyramid, generating a difference-of-Gaussian pyramid and filtering out extreme points. Stable feature points are obtained through sub-pixel interpolation and edge point removal. The magnitude and direction of the gradient in the neighborhood of the feature points are calculated, the direction distribution is statistically analyzed, and the precise main direction is determined by interpolation. The neighborhood of the feature points is divided into sub-regions based on the main direction, and the gradient magnitude of each sub-region is statistically analyzed to form a feature vector. Multiple platform images are stitched together by matching the feature vectors to obtain a complete platform image.

8. The method for planning crowd evacuation routes in a subway station based on density detection according to claim 3, characterized in that, The image recognition algorithm uses the YOLOv11 algorithm. The specific steps are as follows: the backbone network performs hierarchical feature extraction and downsampling on the stitched complete platform image to output multi-scale original feature maps; the neck network performs feature fusion and refinement optimization on the multi-scale original feature maps, fusing shallow detail features and deep semantic features to generate multi-scale fused feature maps; the head network performs target detection and prediction on the multi-scale fused feature maps, outputting a prediction tensor containing passenger target bounding box coordinates and target existence confidence, thus completing passenger target recognition and counting the number of people.

9. A crowd evacuation path planning system for subway stations based on density detection, characterized in that, A method for planning crowd evacuation routes in a subway station based on density detection, as described in any one of claims 1-8, comprises a platform data acquisition module, a crowd density calculation module, an evacuation interval determination module, an evacuation area division module, and an evacuation guidance execution module that are sequentially connected in communication. The platform data acquisition module includes a multi-camera shooting unit and an image transmission unit installed at various locations in the platform waiting area. The multi-camera shooting unit is used to perform a full-area scan of the platform waiting area and collect platform image information at the beginning of an emergency evacuation event. The image transmission unit is used to transmit the collected platform image information to the outside. The crowd density calculation module includes an image receiving unit, an image processing unit, and a density calculation unit. The image receiving unit receives platform image information. The image processing unit stitches the image information together using the SIFT algorithm to obtain a complete platform image and uses the YOLOv11 algorithm to perform passenger target recognition and number of people counting on the complete platform image. The density calculation unit calculates the average crowd density based on the actual size of the waiting area on the platform, which serves as real-time crowd density data. The evacuation zone determination module includes a mapping storage unit and a zone matching unit. The mapping storage unit pre-stores a mapping list of crowd density and optimal transfer channel evacuation zone lengths, which is obtained through simulation based on the classic Moore neighborhood floor field cellular automata model. The zone matching unit is used to match the appropriate evacuation zone length L for the transfer channel from the mapping list based on real-time crowd density data. c ; The evacuation zone division module is a critical value conversion unit, used to convert the evacuation interval length L... c Convert to indicator light threshold N c ; The evacuation guidance execution module includes an instruction transmission unit, a PLC control unit, and an array of ground evacuation guidance indicator lights evenly spaced along the length of the platform waiting area. Each indicator light in the array is equipped with a unique identification ID value. The instruction transmission unit is used to receive the indicator light threshold value N. c The PLC control unit is used to connect the ID values ​​of each indicator light with N. c Compare and control the indicator light direction; if the ID value is less than N... c The indicator light points to the transfer passage, and the ID value is greater than N. c The indicator lights point to the regular stairwell exits, guiding passengers in different evacuation areas to the corresponding exit channels.

10. A crowd evacuation path planning system for subway stations based on density detection according to claim 9, characterized in that, The crowd density calculation module is integrated into the DSP digital signal processor, and the mapped storage unit is the built-in memory of the DSP digital signal processor, realizing the calculation of real-time crowd density data and matching of evacuation interval length.

Images